<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Luca Maria Aiello on Medium]]></title>
        <description><![CDATA[Stories by Luca Maria Aiello on Medium]]></description>
        <link>https://medium.com/@lajello?source=rss-547f52530508------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*3ryOujqQ3GmcQFHDZ-3ASw.png</url>
            <title>Stories by Luca Maria Aiello on Medium</title>
            <link>https://medium.com/@lajello?source=rss-547f52530508------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 03:44:26 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@lajello/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Ten Social Dimensions of Conversations]]></title>
            <link>https://medium.com/socialdynamics/ten-social-dimensions-of-conversations-155200fa64f6?source=rss-547f52530508------2</link>
            <guid isPermaLink="false">https://medium.com/p/155200fa64f6</guid>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[twitter]]></category>
            <category><![CDATA[reddit]]></category>
            <category><![CDATA[research]]></category>
            <category><![CDATA[computational-sociology]]></category>
            <dc:creator><![CDATA[Luca Maria Aiello]]></dc:creator>
            <pubDate>Sun, 19 Apr 2020 15:31:34 GMT</pubDate>
            <atom:updated>2020-04-24T19:27:48.616Z</atom:updated>
            <content:encoded><![CDATA[<h4>Using AI to find the building blocks of our relationships</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QthpbxRrYton3gldpB_OsA.jpeg" /></figure><p>Social relationships are paramount. They determine where we work, who we marry, and what we do. That’s why social science researchers— and more recently their “computational social science” cousins — spent considerable efforts to draw systematic categorizations of the fundamental sociological dimensions that describe human relationships. <strong>We tried to figure out what are the fundamental dimensions of social relationships and how to detect them automatically from online interaction data.</strong></p><p>By compiling an extensive review of decades’ worth of findings in sociology and social psychology, <a href="https://dl.acm.org/doi/10.1145/3274312"><strong>we identified</strong></a><strong> ten dimensions that have been widely used to categorize relationships:</strong></p><ul><li>Exchange of <strong>Knowledge </strong>📚</li><li><strong>Power </strong>dynamics (e.g., between a boss and their employee) 👑</li><li><strong>Status </strong>giving — conferring appreciation, gratitude, or admiration 🙏</li><li>Expressions of <strong>Trust </strong>🤞</li><li>Giving emotional or practical <strong>Support </strong>🤗</li><li><strong>Romance </strong>🌹</li><li><strong>Similarity </strong>— shared interests, motivations or outlooks 👯‍♂️</li><li><strong>Identity</strong> — Sense of belonging to a group or community 🏳️‍🌈</li><li><strong>Fun</strong> 😂</li><li><strong>Conflict</strong> ⚔️</li></ul><p>It turns out that people find these dimensions very relatable. Through a small crowdsourcing experiment, <strong>we asked 200 people to spell out words that described their social connections and found that all of them fitted into the 10 dimensions.</strong></p><p>Being able to quantify these 10 dimension from interaction data is quite useful because, by combining them in opportune proportions, one can draw an accurate, explainable, and intuitive description of the nature of most relationships.</p><p>In our <a href="https://arxiv.org/abs/2001.09954">work</a> that will be presented at <a href="https://www2020.thewebconf.org/">The Web Conference 2020</a>, <strong>we used Natural Language Processing to measure the 10 dimensions from conversational text. </strong>We asked crowdworkers to label 9,000 Reddit comments with the 10 dimensions. We used this data to train AI tools that can predict all dimensions rather accurately from potentially any text— <em>for </em>🤓<em> eyes: using </em><a href="https://huggingface.co/transformers/model_doc/bert.html"><em>BERT</em></a><em> and </em><a href="https://pytorch.org/tutorials/beginner/nlp/sequence_models_tutorial.html"><em>LSTM</em></a><em> we got AUCs ranging from 0.75 (Identity) to 0.98 (Romance).</em></p><p>And now, the fun part. We applied our tool to online data discover how the 10 dimensions are used in different contexts and at different levels of aggregation: individual messages, relationships, groups, and even entire States in the US!</p><h4>Social Dimensions in Individual Messages — movie scripts 🎥</h4><p>Scripted movie dialogs are fictional yet plausible representations of conversations that span a wide spectrum of human emotions and relationship types. <strong>We ran our models on all movie lines from the </strong><a href="https://www.cs.cornell.edu/~cristian/Cornell_Movie-Dialogs_Corpus.html"><strong>Cornell movie script corpus</strong></a><strong>, and found out what are the dimensions conveyed by iconic movie lines.</strong></p><p>Next, a few lines from some iconic 20th century movies. We reported the detected levels of the 10 dimensions in the inset diagrams; those above the dashed line are the most significant ones.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Lc6GWv5csCB_0q0fFF8-cA.png" /><figcaption>“Only a fully trained Jedi Knight, with The Force as his ally, will conquer Vader and his Emperor. If you end your training now, if you choose the quick and easy path, as Vader did, you will become an agent of evil “ — <strong>Ben Kenobi</strong></figcaption></figure><p>Ben transfers his knowledge of the ways of the Force to Luke. This line highlights the power relationship between the master and the pupil, and warns about impending conflict.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JjPsDYGdI_1kkpBZxBiqFQ.png" /><figcaption>“Frankie, you’re a good old man, and you’ve been loyal to my Father for years…so I hope you can explain what you mean” — <strong>Michael Corleone</strong></figcaption></figure><p>Michael likes to think of himself as a man of honor. He tributes appreciation (Status) to his friend for his services, and grants him Trust in the wake of suspicions of betrayal.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*maiKZCB__jhexBtvGD3I5g.png" /><figcaption>“Look, Dave, I know that you’re sincere and that you’re trying to do a competent job, and that you’re trying to be helpful, but I can assure the problem is with the AO-units, and with your test gear” — <strong>HAL 9000</strong></figcaption></figure><p>It seems that the supercomputer HAL 9000 is trying to provide some technical support (= knowledge + support) to an increasingly worried astronaut.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F_9i77m1aNQmlzQXae_D9A.jpeg" /><figcaption>“But, as you are so fond of observing, Doctor, I’m not human” — <strong>Spock</strong></figcaption></figure><p>The identity of a Vulcan, expressed before the extreme sacrifice 🖖.</p><h4>Social Dimensions in Relationships — Twitter 🐦</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lNXKauMwVCZL-GLIP5olHQ.png" /><figcaption>The “Isle of Ties” is one of the games from <a href="https://www.tinghy.org/">tinghy.org</a>. Players mark their Twitter friends with 3 dimensions that best describe their relationships with them.</figcaption></figure><p>Dimensions estimated at message-level can predict the dimensions that people would use to describe their social relationships. To show that, we collected data from <a href="https://www.tinghy.org/">tinghy.org</a>, a platform of games that gathers people’s perceptions about their Twitter friends (see pic on the left). We found that <strong>we can get good predictions of relationship-level labels (e.g., a relationship based on trust) by looking at 20 twitter replies or more between the two people involved.</strong></p><h4>Social Dimensions of Communities — Enron corp. emails 📧</h4><p>Enron Corporation was an American company founded in 1985 that went bankrupt in 2001, when its systematic practices of accounting fraud were exposed to the public. After the scandal and the resulting investigation, The <a href="https://link.springer.com/chapter/10.1007/978-3-540-30115-8_22">Enron Email Dataset</a> was released to the public. We looked at how expressions of different dimensions varied over time in Enron emails, on average.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/878/1*h465glBARfPbnTLZdlOgEA.png" /><figcaption>How the presence of five social dimensions in Enron employees emails changes over time, compared to traditional sentiment analysis (performed with the <a href="https://github.com/cjhutto/vaderSentiment">Vader</a> tool). Plots are marked with four significant events in Enron’s history: i) the beginning of concerns about the financial stability of the company; ii) the first round of layoffs; iii) the start of financial losses; iv) the declaration of bankruptcy.</figcaption></figure><p><strong>The evolution of conversational dimensions reveals a rich picture that matches the known stages of Enron’s downfall</strong>. As the initial concerns sparked, the exchange of status and support plummeted: panic started to spread and employees stopped celebrating their achievements, thanking each other, and offering comfort. About three months later, the frequency of knowledge exchange dropped sharply: as concerns grew, employees spent less time in dealing with their everyday duties. A few weeks before the layoffs, as it became clear that many employees would have been made redundant, conflict exploded and the power structure collapsed — fewer orders were given to the angry crowd of employees who were made aware of the impeding jobs cuts. In the aftermath of the layoffs, those who managed to stay in the company gave support to each other for a few weeks before the imminent crack.</p><h4>Social Dimensions in Society — US States statistics 🦅</h4><p>Last, we tested whether the presence of those dimensions in conversations is associated with real-world outcomes at societal level. <strong>We extracted the ten social dimensions from 160M Reddit messages posted by 1M users who we could geo-reference at the level of US States</strong>. We conducted a geographical analysis to study the relationship between the presence of the 10 dimensions and socio-economic outcomes, estimated by official statistics. In particular we checked the relationship between:</p><ul><li>Exchange of <strong>knowledge </strong>and average <strong>education level</strong></li><li>Exchange of <strong>knowledge</strong> and average <strong>income</strong></li><li>Expression of <strong>social support</strong> and <strong>suicide rates</strong></li></ul><p>As expected, Knowledge is positively associated to education levels and to wealth — knowledge exchange is a driver of innovation and economic growth. Support is positively associated with suicide rates: people affected by depression, especially those who have suicidal thoughts, do not tend to trust their peers and seek social support in different contexts, often online.</p><p>All the dimensions combined predict the official statistics quite accurately (up to an adjusted R2 of 0.52), even after discounting for confounding factors like population density.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QGcPj-N3_hEEUdHfOhmp0Q.png" /><figcaption>Expressions of social dimensions in online language correlate with relevant societal outcomes estimated by official statistics at the level of US States</figcaption></figure><h4>What is this all for?</h4><p>The ability of automatically extract fundamental dimensions of social exchange from text could contribute to creating research and analytics tools for social networks. We believe that the dynamics of a number of processes mediated by social networks could be re-interpreted with our application of the 10 dimensional model to conversation networks. For example, <strong>researchers could look at the phenomena of information diffusion, spreading of fake news, polarization, and link creation in the light of the ten social dimensions.</strong></p><p>Social media companies could monitor the prevalence of the ten dimensions in the public online discourse as a way to <strong>detect and promote positive and meaningful interactions</strong> —for example those based on support and trust — rather than just punishing misbehavior.</p><h4>Want to learn more?</h4><p>Check out the <a href="https://arxiv.org/abs/2001.09954">preprint</a> or our <a href="http://social-dynamics.net/projects/social_dimensions/index.html">data and code</a>. Research conducted by <a href="https://twitter.com/MinjeChoi1">@minje_choi</a> at <a href="https://twitter.com/umsi">@umsi</a> and <a href="https://twitter.com/lajello">@lajello</a>, <a href="https://twitter.com/varkrisz">@varkrisz</a>, and <a href="https://twitter.com/danielequercia">@danielequercia</a> at <a href="https://twitter.com/BellLabs">@BellLabs</a>. Please contact us for any questions!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=155200fa64f6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/socialdynamics/ten-social-dimensions-of-conversations-155200fa64f6">Ten Social Dimensions of Conversations</a> was originally published in <a href="https://medium.com/socialdynamics">SocialDynamics</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Summer Visions in Berlin]]></title>
            <link>https://medium.com/socialdynamics/summer-visions-in-berlin-830519bddab8?source=rss-547f52530508------2</link>
            <guid isPermaLink="false">https://medium.com/p/830519bddab8</guid>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[computational-sociology]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[summer-school]]></category>
            <dc:creator><![CDATA[Luca Maria Aiello]]></dc:creator>
            <pubDate>Tue, 06 Aug 2019 23:00:42 GMT</pubDate>
            <atom:updated>2019-08-30T09:54:12.816Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cC5RXwA5JyUTYwb6USznHQ.png" /></figure><p>The third edition of the <a href="http://summerschool.computationalsocialscience.eu/2019/">Summer School on Methods for Computational Social Science</a> took place last July in Berlin, in a historic mansion surrounded by the beautiful nature of the Wernsdorfer Lake. This year’s edition has focused on research and <strong>methods for analyzing multimedia data</strong>. The organizers <a href="http://www.claudiawagner.info/">Claudia Wagner</a> (GESIS) and <a href="http://www.nicolaperra.com/">Nicola Perra</a> (Greenwich University) did an excellent job for the third time in a row. They put together a very balanced group of 40+ excellent students and a lineup of 9 speakers I was lucky enough to be part of. Here’s my summary of that exciting week.</p><h4>Automated Image Analysis for Social Science Research. Andreu Casas (New York University)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OtvJZ2aPr0cIK3fT7fG4Ng.jpeg" /><figcaption>Andreu Casas (New York University) teaching a tutorial on deep learning for computer vision</figcaption></figure><p>The school kicked off with an excellent 101 tutorial on deep learning for machine vision by <a href="http://andreucasas.com/">Andreu Casas</a>, which featured an interesting glossary of deep learning terms “translated” to more familiar statistical terms recognizable by scientists across disciplines. Andreu is a political scientist who embraced machine learning as a tool to answer questions about societal dynamics. He told a story about <strong>the </strong><a href="https://science.sciencemag.org/content/308/5728/1623"><strong>persuasive power of images</strong></a><strong> in steering the political life of our communities</strong>, which calls for new ways of channeling this great power for social good. Andreu is working in this direction by trying to quantify the role of social media pictures in creating real-world <a href="https://www.researchgate.net/profile/Andreu_Casas/publication/326865658_Images_that_Matter_Online_Protests_and_the_Mobilizing_Role_of_Pictures/links/5b6dd642299bf14c6d98c2c2/Images-that-Matter-Online-Protests-and-the-Mobilizing-Role-of-Pictures.pdf">mobilization</a> and protests. Machine learning models could do that quite effectively but at the cost of a ludicrous amount of training data. That’s why Andreu is currently building new tools that could quench the data thirst of AI with far fewer data entries.</p><h4>Machine vision in social computing. Miriam Redi (Wikimedia Foundation)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UKtxUnANuf6pHhQ1ZBT7SA.jpeg" /><figcaption>Miriam Redi (Wikimedia Foundation) recalling the computer vision classification accuracy we were used to a few years back</figcaption></figure><p>In 1966, <a href="http://www.papert.org/">Seymour Papert</a>, who back then was a research associate at MIT, drafted a proposal for a summer internship project. The document was titled <strong>“</strong><a href="https://dspace.mit.edu/handle/1721.1/6125"><strong>The Summer Vision Project</strong></a><strong>”</strong> and it planned to solve in the span of a few months the problem of automatically detecting the shape of objects in a picture. It turns out that a few more decades of work were needed to figure out a good solution to that problem. With this short story, <a href="http://www.visionresearchwitch.com/">Miriam</a> starts recounting the history of Computer Vision, with all its ups and downs, until the recent deep learning revolution. Miriam wrote part of this glorious history with her seminal work in computational aesthetics (her survey on <a href="https://dl.acm.org/citation.cfm?id=3301299">understanding visual interestingness</a> has been published recently). At Wikimedia, she’s now tackling a challenge with even greater importance for mankind: modeling visual knowledge. <strong>Can we use images to promote knowledge equity?</strong> How do people learn from images? Can we detect images with high encyclopaedic value across cultures? Stay tuned for upcoming work that will try to answer some of these questions.</p><h4>Social Media for Lifestyle Health. Yelena Mejova (ISI Foundation)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LJnDc0h7e_B0x8rxKBywYg.jpeg" /><figcaption>Yelena Mejova (ISI Foundation) showing how words used on social media can predict obesity at country level</figcaption></figure><p>Until the <a href="https://www.forbes.com/sites/forbestechcouncil/2018/09/07/how-the-5g-revolution-will-drive-future-innovations-of-iot/#5e46cd12637e">5G revolution</a> hits us with its swarm of intelligent sensors, smartphones remain the most ubiquitous form of distributed sensing. The endless stream of geo-referenced multimedia content captured by those devices and continuously poured on the Web contains precious information about our feelings, behaviors, and even about our health. Since her early days as a researcher at Yahoo! Labs, <a href="https://sites.google.com/site/yelenamejova/">Yelena</a> has pioneered the science of mining Twitter data to measure how our lifestyle — and especially our eating habits — impacts our health. Bad nutrition is the direct cause of <a href="https://www.who.int/en/news-room/fact-sheets/detail/the-top-10-causes-of-death">most deaths</a> in high-income countries, yet it’s notoriously difficult to track at scale. Yelena explained how digital data can come to the rescue: <strong>mentions of food and drinks in tweets </strong><a href="https://dl.acm.org/citation.cfm?id=2702153"><strong>are predictive</strong></a><strong> of obesity rates at country and regional level</strong>; body weight can be inferred easily from <a href="https://dl.acm.org/citation.cfm?id=2897727">profile pictures</a>; and machine vision can inspect <a href="http://openaccess.thecvf.com/content_cvpr_2017/html/Salvador_Learning_Cross-Modal_Embeddings_CVPR_2017_paper.html">food pictures</a> to infer their ingredients and energy intake.</p><h4>The Spirit of the City. Luca Maria Aiello (Nokia Bell Labs)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kB4cM-UrPpZMMdhJUBcX-w.jpeg" /><figcaption>Luca Aiello (Nokia Bell Labs) surveying sources of geo-referenced data alternative to social media posts</figcaption></figure><p>Geo-referenced pictures uploaded on social media are also excellent descriptors of the space we live in. Five years ago, motivated by literature in <a href="https://en.wikipedia.org/wiki/The_Image_of_the_City">urbanism</a> and <a href="https://www.jstor.org/stable/1728966?seq=1#page_scan_tab_contents">social psychology</a>, I teamed up with my two pals <a href="http://www.di.unito.it/~schifane/">Rossano</a> and <a href="https://researchswinger.org/">Daniele</a> to embark on a mission to measure the <strong>“Spirit of the City”: the intangible aspects that determine our experience of urban places</strong>. We used social media (especially Flickr pictures) to draw many street-level maps of intangible properties of space (<a href="https://dl.acm.org/citation.cfm?id=2631799">beauty</a>, <a href="https://iris.unito.it/handle/2318/1579196">smell</a>, <a href="https://royalsocietypublishing.org/doi/abs/10.1098/rsos.150690">sound</a>, <a href="https://dl.acm.org/citation.cfm?id=3274413">ambiance</a>, <a href="https://www.frontiersin.org/articles/10.3389/fphy.2018.00027">culture</a>, … you name it), and collected them under <a href="http://www.goodcitylife.org/">www.goodcitylife.org</a>. In my lecture I showed how these maps can predict real-world outcomes, facilitate the work of architects and city planners, and improve the quality of our daily habits. But I also talked about the challenges of studying space through the lens of Web data after <a href="https://en.wikipedia.org/wiki/General_Data_Protection_Regulation">GDPR</a> and <a href="https://en.wikipedia.org/wiki/Christopher_Wylie">Chris Wylie</a>’s confessions inflicted a critical blow to open research on social media data. A great opportunity lies ahead for young researchers to use “unconventional” sources of data for spatial studies. In this spirit, I discussed how we combined open medical prescription data with food grocery sales statistics to measure the <a href="https://epjdatascience.springeropen.com/articles/10.1140/epjds/s13688-019-0191-y">impact of eating habits on diabetes prevalence</a> for the whole city of London (with a <a href="http://goodcitylife.org/food/index.php">cool viz</a> by <a href="https://twitter.com/tobi_vierzwo">Tobi Kauer</a>, check it out!).</p><h4>“Seeing” Spatial Contexts at Scale in Urban Sociology. Jackelyn Hwang (Stanford University)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s8tLK8B0s6E1uFc9JJk3uA.jpeg" /><figcaption>Jackelyn Hwang (Stanford University) describing the complex pipeline of image analysis to extract elements of disorder from street-view pictures</figcaption></figure><p><a href="https://sociology.stanford.edu/people/jackelyn-hwang">Jackelyn</a> provided a fascinating overview on the origin and milestones of Urban Sociology, from the early work of <a href="https://en.wikipedia.org/wiki/The_Philadelphia_Negro">William Du Bois</a> and the <a href="https://en.wikipedia.org/wiki/Chicago_school_(sociology)">Chicago School</a> up to the flourishing of <a href="https://www.jstor.org/stable/10.1086/210356?seq=1#metadata_info_tab_contents">Systematic Social Observation</a> research (SSO), an ensemble of techniques to capture sights, sounds, and feel of the streets beyond survey and ethnographic methods. As a sociologist who’s well-versed in using state-of-the-art computational methods, Jackelyn is working on bringing SSO to the next level by leveraging the power of Big Data. She provided a great example of how <strong>Google Street View images can be used to capture elements of physical disorder</strong>, decay and low maintenance that, in turn, are linked with a number of important societal outcomes. In particular, she is working on the problem of training computer vision algorithms to detect thrash left on the street — a seemingly simple exercise that, in reality, hides lots of hard challenges ranging from the collection of reliable ground-truth annotations to adapting complex neural network architectures to this specific task. The work is still in progress but she showed already very promising results for the cities of Boston, Detroit, and LA. Fascinating lecture and an excellent example of how social science can reach new heights with the help of computational methods.</p><h4>Data Exploration Through Artificial Intelligence and Dataviz. Mauro Martino (IBM Research)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JVxJuy9ro26QvxGqQ0Nxgw.jpeg" /><figcaption>Mauro Martino (IBM) providing evidence of how the creative genius of data visualization experts can compose images of prodigious persuasive power</figcaption></figure><p><a href="http://www.mamartino.com/">Mauro</a> is a dataviz rockstar. He landed in Berlin fresh from a deluge of press releases praising his latest creation: <a href="https://www.fastcompany.com/90376689/what-you-look-like-as-an-renaissance-painting-according-to-ai">AiPortaits.com</a>. On the website (which has been taken down after receiving millions of requests), you could upload a photo of your face and get back a unique version of it that looks like an old painting. The Generative Adversarial Network technology he used is so advanced it can produce really <a href="https://mymodernmet.com/ai-portraits/">stunning</a> works of art — much cooler than traditional style-transfer approaches. In his lecture, Mauro presented a selection of his finest creative production of the last few years, including the visualization of the <a href="http://www.mamartino.com/projects/rise_of_partisanship/index.html">rise of partisanship</a> in the US House and the <a href="http://formafluens.io/">Forma Fluens</a> project, a piece of art that uses doodles drawn by millions of people all over the world to unearth our cultural differences as well our universal mental models. With a breathtaking series of amazing imagery, Mauro demonstrated <strong>the extraordinary persuasive power of data visualization</strong> and reaffirmed the centrality of dataviz in the cycle of Data Science. For those interesting in learning the craft, he provided a <a href="https://drive.google.com/open?id=147gcxla7Ry6DRX06yXmqOIsfswkH3RfT#">useful list</a> of resources to start from. He finished with a word of wisdom: to learn how to create beautiful things you should <em>“get exposed to beauty” </em>first.</p><h4>The Visibility of News Sources across Media Channel. Sandra González-Bailón (University of Pennsylvania)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ciuylIX7AvwpCYvoR17A0w.jpeg" /><figcaption>Sandra González-Bailón (Pennsylvania University) explaining the complex network effects behind the process of news consumption</figcaption></figure><p>The collaboration of social sciences with the computer and physical sciences has strengthened considerably in the last few years; that’s because some of the most important questions on societal dynamics cannot be fully answered without a combined approach. <a href="https://www.asc.upenn.edu/node/648">Sandra</a>’s research is an exemplary case of how to these disciplines come together to create a full-fledged computational social science approach. Strong of her background in sociology, Sandra could put under tight scrutiny the existing models of collective news exposure as they are often backed exclusively by self-reported measures and disregard the broad diversity of news outlets. By analyzing large-scale browsing data of news websites, she could record for the first time at scale the discrepancy between reported exposure and actual consumption, thus discovering that <strong>the often-ignored underwood of low-reach news outlets plays a massive role in collective information exposure. </strong>By building <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3351334">audience networks</a> from the interactions between users and news websites she also challenged the selective exposure paradigm: unlike in social media, the news consumption patterns of readers seem not to be polarized according to political stances. Her book <strong>“Decoding the Social World”</strong> is out and <a href="https://www.amazon.com/Decoding-Social-World-Consequences-Communication/dp/0262037076">available on amazon</a>.</p><h4>From Warburg to Deep Learning. Maximilian Schich (University of Texas at Dallas)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*onG07cLtDHUgIrcfEnIt_w.jpeg" /><figcaption>Maximilian Schich (University of Texas at Dallas) demonstrating how modern art, technology, and culture descend from the rational genius of Leibniz</figcaption></figure><p>The exciting and fast-paced technological innovations we have been experiencing in the last decades makes us forget that the rich cultural scene we bask in is a direct emanation of the talent of our illustrious ancestors. <a href="http://www.schich.info/">Max Schich</a> — an art historian working at the intersection of information visualization, computer science, physics, and art — poured on us copious wisdom from the past, like a river in flood. His very inspirational talk stretched from Leibniz to Star Wars and touched recurrently upon the concept of Culture. To emphasize how much <strong>physical artifacts have a deep influence on our digital life</strong>, he proposed a striking parallel between contemporary multimedia collections (like Pinterest walls) and the <a href="https://warburg.library.cornell.edu/about">Mnemosyne Atlas</a>, a collection of symbolic visual artifacts by historian <a href="https://en.wikipedia.org/wiki/Aby_Warburg">Aby Warburg</a>. Max has worked extensively on characterizing the complex evolution of cultural processes, for example by reconstructing the <a href="https://science.sciencemag.org/content/345/6196/558">network of mobility</a> of notable individuals over the course of two millennia. But that is just a piece in the much broader puzzle of cultural complexity that he is relentlessly trying to complete, piece after piece. Max is working on a book that we hope to be able to read very soon for more enlightening lessons at the interface between history and technology.</p><h4>Social Speech Processing . Björn W. Schuller (Imperial College London)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JTHZZwfwnadZ3swWIAhJzg.jpeg" /><figcaption>Björn W. Schuller (Imperial College London) smiling at smartwatches that track your laughter instead of counting your steps</figcaption></figure><p>Björn represents one of those (rare) exceptions to the commonsense rule: “success and modesty don’t go hand in hand”. His <a href="https://www.overleaf.com/articles/publications-by-bjorn-schuller/vjyvgjfrzgtf.pdf">colossal corpus</a> of work on audio processing for affective computing and digital health is absolutely spectacular, and I was struck by the simplicity, candor, and humility he demonstrated while presenting it. Using a blend of acoustic and linguistic information, mixed together with cutting-edge deep learning, Björn builds all sorts of sound-powered applications. Just to mention a few: an impressive noise reduction technique, a laughter tracker, an emotion detector, and a conversational agent that can effectively deal with patients affected by depression. Björn works also in the private sector as the founder of A<a href="https://www.audeering.com/">udeering</a>, a company for audio intelligence technologies, yet keeping up the spirit of open science by making lots of <strong>code and toolkits for audio processing available for free</strong> (check them out: <a href="https://gitlab.com/openCoSy/CAS2T">CAS2T</a>, <a href="https://www.audeering.com/opensmile/">Opensmile</a>, <a href="https://github.com/openXBOW/openXBOW/">OpenXbow</a>, <a href="https://ihearu-play.fim.uni-passau.de/">iHearU</a>, <a href="http://deepspectrum/">DeepSpectrum</a>, <a href="https://github.com/auDeep/auDeep">AuDeep</a>, <a href="https://github.com/end2you/end2you">end2you</a>). Bravo Björn, way to go!</p><h4>…and, of course, the students!</h4><p>Last but not least, a big shout-out to all the students who participated to the school. They patiently attended all the lectures and gave heart and soul in developing their mini-projects (congrats to the <a href="https://networkdatascience.ceu.edu/article/2019-08-05/methods-analyzing-multimedia-data-css-summer-school-berlin">winners</a> and to those awarded with <a href="https://twitter.com/pushkalagarwa/status/1156605356249878528">honorable</a> <a href="https://twitter.com/lajello/status/1155054386361618432">mentions</a>). Most of all, they where the first in line when it came to the most important activities of the week: drinking, dancing and diving. That’s how you build a strong foundation for the next generation computational social scientists.</p><p>Thanks again to Claudia and Nicola for creating this unique opportunity and for letting us be part of it 💗 <em>#iwillbeback</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*1HHiLCjuJnEPLsif35Vppg.jpeg" /><figcaption>Teachers and the future generation of computational social scientists posing for a memorable group picture</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=830519bddab8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/socialdynamics/summer-visions-in-berlin-830519bddab8">Summer Visions in Berlin</a> was originally published in <a href="https://medium.com/socialdynamics">SocialDynamics</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Science Reversions in Torino: DSAA’18]]></title>
            <link>https://medium.com/socialdynamics/science-reversions-in-torino-dsaa18-dc9fdca18389?source=rss-547f52530508------2</link>
            <guid isPermaLink="false">https://medium.com/p/dc9fdca18389</guid>
            <category><![CDATA[digital-health]]></category>
            <category><![CDATA[ethics]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[bias]]></category>
            <category><![CDATA[research]]></category>
            <dc:creator><![CDATA[Luca Maria Aiello]]></dc:creator>
            <pubDate>Fri, 12 Oct 2018 10:04:02 GMT</pubDate>
            <atom:updated>2019-08-30T09:43:54.628Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rRAvYEEcLxhHUFAeB0-bPQ.jpeg" /></figure><p>Oh, sweet Torino! Far from the mainstream international routes that lead businessmen to Milan and tourists to Rome and Venice, Torino is a gem of art and architecture and a vibrant hub for science and innovation.</p><p>Last week, Torino hosted three high-profile scientific events: the 5th Conference on Data Science and Advanced Analytic (<a href="https://dsaa2018.isi.it/">DSAA</a>); an <a href="https://dsaa2018.isi.it/program/industrial-day">Industrial Day</a> sponsored and hosted by Intesa Sanpaolo in its brand-new <a href="http://www.grattacielointesasanpaolo.com/en/">skyscraper</a> designed by Renzo Piano; and “<a href="https://www.isi.it/en/news-events/meet-the-isi-fellows-2018-">Science Crossroads</a>”, a yearly event organized by the <a href="https://www.isi.it/en/home">ISI Foundation</a> to award young scientists from all over the world with a Fellowship in recognition of their scientific achievements. <a href="https://twitter.com/FrancescoBonchi">Francesco Bonchi</a> and <a href="https://twitter.com/FakeFoster">Foster Provost</a>, general chairs of DSAA (+<a href="https://dsaa2018.isi.it/organization">team</a>) and <a href="https://twitter.com/ciro">Ciro Cattuto</a>, Scientific Director of the ISI Foundation (+<a href="https://www.isi.it/en/people">team</a>) did an excellent job in putting together five full days of scientific gatherings.</p><p>Even though the the program was very interdisciplinary, most of the talks felt interconnected by a common philosophical theme: in the inchoate era of Artificial Intelligence, we may need to revise our scientific methods to gain full understanding of the technologies we develop and of their ethical implications. Sometimes, this process might entail borrowing some forgotten elements of our past to move faster towards a brighter future. Here’s the summary of these <em>“Science Reversions”</em>.</p><h3>The Role of Modeling in the Era of Big Data</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v2mG6D7bM8vPvD2X-VTtpA.jpeg" /><figcaption>A generative, mechanistic epidemic model: Big Data and Machine Learning form a predictive system wrapped around a core model tailored on the specific pandemic under study</figcaption></figure><p><strong>A Mechanistic View of Science (Alessandro Vespignani, Northeastern University). </strong>Alessandro Vespignani is a modern <em>numen</em> of computational epidemiology. He opened his talk with a brief survey on the history of numerical epidemic models, emphasizing how their advances virtually halted in the ’50s. Those data-hungry models started to work accurately only after the Big Data revolution: Multidimensional and granular data from a galaxy of public and private providers allowed epidemiologists to predict with striking accuracy the spreading of pandemics like <a href="https://www.ncbi.nlm.nih.gov/pubmed/23237460">H1N1</a>, <a href="http://currents.plos.org/outbreaks/article/assessing-the-international-spreading-risk-associated-with-the-2014-west-african-ebola-outbreak/">Ebola</a>, and <a href="http://www.pnas.org/content/114/22/E4334">Zika</a>. Inebriate by the power of Big Data, the scientific community explored new predictive models that were increasingly data-driven and less focused on understanding the underlying phenomena. That line of thinking led to glaring failures, epitomized by the infamous <a href="https://www.wired.com/2015/10/can-learn-epic-failure-google-flu-trends/">Google Flu Trends fiasco</a>. The <em>“I don’t care about understanding, as long as it works”-</em>narrative is fundamentally wrong, Vespignani argues, advocating for a <strong>reversion to a mechanistic approach</strong>. Researchers should first produce carefully-crafted models with domain knowledge baked into them and then integrate those into a larger predictive system. The use of Big Data and machine learning is still crucial for these models to work with maximum accuracy. He provided examples of how social media data can be used to bootstrap modern epidemic models and machine learning techniques are beneficial to create ensembles of different models. The Big Data enthusiasts who proclaimed <a href="https://www.wired.com/2008/06/pb-theory/">The End of Theory</a> ten years ago may have spoken too soon. His new, colorful book <em>“</em><a href="https://www.amazon.co.uk/Charting-Next-Pandemic-Infectious-Spreading/dp/3319932896"><em>Charting the Next Pandemic</em></a><em>”</em> is out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8HFPsZOP_TVb_sd62l0Cmg.jpeg" /><figcaption>Chris Bishop illustrating the results of models based on factor graphs with a toy interface for movie recommendations</figcaption></figure><p><strong>One Model to Rule Them All (Christopher Bishop, Microsoft Research).</strong> There is quite an abundance of Machine Learning methods out there. A question that must have flashed in the minds of most machine learning students at least once is: which of those approaches is <em>the absolute best</em>? Unfortunately, in the general case, <a href="https://en.wikipedia.org/wiki/No_free_lunch_theorem">there is no free lunch</a>: when averaged over all the possible problems, any algorithm is as good (or bad) as any other. But what about neural networks? Aren’t those sort of magic? Well, sadly, neural networks have a lot of assumptions and prior knowledge baked into them, which makes them excellent tools only for some (albeit very important) classes of problems. In line with Prof. Vespignani, Chris Bishop (world-renowned pioneer of AI and author of a best-selling technical <a href="https://www.amazon.co.uk/Pattern-Recognition-Learning-Information-Statistics/dp/0387310738">book</a> on machine learning) asserted that <strong>we cannot abandon domain knowledge and rely on Big Data only</strong>: to obtain good predictions we need to lay assumptions based on our knowledge of the problem. Only after embedding those assumptions into a model we can turn to Big Data for training. But what if we could automate this process? Could we have a tool that asks the analyst to write down a set of assumptions and takes care of all the rest? <a href="http://infernet.azurewebsites.net/">Infer.net</a> is Microsoft Research’s attempt to achieve this vision. Infer.net offers a compiler that converts a high-level model description directly into source code to perform inference. Its flexibility is its power: it can cope with the variety of real world problems and it has been already used in a number of real-world <a href="https://dotnet.github.io/infer/papers.html">use cases</a>. Bishop’s new book <em>“</em><a href="http://www.mbmlbook.com/"><em>Model Based Machine Learning</em></a><em>”</em> is out and in early open access. Physical copies will be sold and the proceeds will be matched my MSR and will go to fund research on cystic fibrosis. So, in Bishop’s words:<em> “please buy the book even if you don’t read it”</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MLs0uKxOkNXvgY-nDfW-fA.jpeg" /><figcaption>It is easy to tell bots apart from humans, less so to identify different types of users</figcaption></figure><p><strong>The Limit of Predictability (Claudia Perlich, Dstillery).</strong> How far can we push the accuracy of machine learning predictions? In many real-world problems, classification models can reach impressive accuracy. But this doesn’t mean that those models are particularly smart. By design, <strong>machine learning models focus on what’s easy.</strong> In classification, for example, the model focuses on those datapoints that can be easily separated from the others. Regrettably though, easily separable instances are those with the littlest value. Strong of her experience as an advisory scientist at <a href="https://dstillery.com/">Dstillery</a>, Claudia Perlich elucidated this concept with a number of examples coming from real-world ML tasks. It is easy to classify types of web users until you realize that most of your true positives are bots that leave very distinctive digital traces. Click prediction on mobile devices can achieve excellent results mostly because of people accidentally tapping on links while trying to activate the flashlight on their phone. Last, by mining geo-referenced data, you may be able to easily spot plenty of frequent-travelers only to realize later on that all them are flight attendants. To avoid frustration, we need to stop and (again) revert to the basics. Her main advice is to think carefully about the set of metrics we optimize for. Blindly using Click Trough Rate (CTR) is often not a good idea. More crucially, when machine learning is applied to critical tasks such as policing or job recommendations, we ought to make a very responsible use of the learning techniques we use: picking the wrong metric can yield disastrous societal outcomes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p1U18YrtbizAotxjuVPdKQ.jpeg" /><figcaption>Explaining machine learning predictions is ultimately a causal inference question</figcaption></figure><p><strong>Explaining High-dimensional Machine Learning Models (Foster Provost, NYU).</strong> In a machine learning prediction task, when the gain in precision given by new training instances declines, adding new features is the main way forward. With more features, a model can once more benefit from additional training points. Looping over the addition of rows and columns to a training set yields monstrously complex models that work well but whose output is hardly interpretable. Foster Provost wants to find out <em>why </em>high-dimensional model work. Prof. Provost can boast a long-lasting experience in the startup world (he founded 5 successful ones) and his faceted background gives him a good edge in the ability of addressing real-world problems by solving their underlying fundamental challenges. He rightly contended that listing the top-predictive variables (maybe out of millions) is not a good way to provide explanations. Instead, he argues that <strong>explainability should be approached as a causal inference problem</strong>: finding the minimal piece of evidence that, if not present, would have led the model to take a different decision. In an algorithmic setting, causal inference is possible because we can potentially observe the outputs of all the possible inputs. He used this reasoning to provide explanations in the task of <a href="https://dl.acm.org/citation.cfm?id=2600523">document classification</a> but the method he proposes is very general, as it is completely agnostic of the learning algorithm to focus only on the input and observed outputs. Prof. Provost’s book <em>“</em><a href="https://www.amazon.co.uk/Data-Science-Business-data-analytic-thinking/dp/1449361323"><em>Data Science for Business</em></a><em>”</em> is out. He is an artist too: his first album “<a href="https://www.meanreversionmusic.com/">Mean Reversion</a>” (which inspired this post’s title) has been recently released. Check it out!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*6t2T7IEAEk3hPRq-mTnMDQ.jpeg" /><figcaption>Scientist and musician: listen to a sample of Mean Reversion’s first single ”<a href="http://smarturl.it/blessingofmybirth">Blessing of My Birth</a>”.</figcaption></figure><h3>Enlightening the Dark Side of Technology</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i2hhZm_PCoWam8BdIz7Eug.jpeg" /><figcaption>Disabilities are mapped into their corresponding “Coolabilities” for the purpose of job matching</figcaption></figure><p><strong>Averting the Apocalypse of Intelligent Technologies (V R Ferose, SAP). </strong>A decade ago, Web technologies were all but good: they were bringing people together and information to everyone. In the last few years, This positive view of technology has changed dramatically. V R Ferose, visionary senior VP at SAP, lists some examples of the long-term negative consequences of technology ranging from the emerging behavioral issues in <a href="https://www.fastcompany.com/40588020/the-case-against-teaching-kids-to-be-polite-to-alexa">kids interacting with home assistants</a> to the deaths caused by <a href="https://www.bbc.co.uk/news/world-asia-india-44709103">fake news</a> spreading in local communities. Ferose’s mission is to learn how we can understand the long-terms consequences of technology and spin them towards good outcomes. He provided five possible ways to move forward. The first is decentralization: when the data is controlled by just a few it is harder to prevent misuse (Tim Berners Lee’s <a href="https://www.fastcompany.com/90243936/exclusive-tim-berners-lee-tells-us-his-radical-new-plan-to-upend-the-world-wide-web">Inrupt</a> is an attempt of bringing the Web back in the hands of <em>The People</em>). The second is changing narrative: he provided a powerful example of describing as <em>“coolabilities”</em> the unique strengths of people with disabilities and how those can be used to increase job offers matches. The third is automation of policies: ideally, any technological requirement decided by law should propagate automatically on every existing deployment and implementation. The fourth is changing learning metrics: the way in which <strong>machine learning</strong> methods optimize their objective functions <strong>can dangerously amplify existing anomalies</strong> (e.g., discrimination). The last is democratization: it should not be only up to a few billionaires to solve the world’s problems, especially when they have huge conflicts of interest; he exemplified this concept with a Gandhi quote: <em>“it’s not what to do with your money that is important, it is the purity of the means of what you earn that matters”</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4NICrKPQpo4_elDjiFIt6A.jpeg" /><figcaption>The Web suffers from a number of biases that are all inextricably entangled</figcaption></figure><p><strong>The Gordian Knot of Biases (Ricardo Baeza-Yates, NTENT).</strong> The Web is also plagued with biases. Ricardo Baeza-Yates, one of the fathers of <a href="https://www.amazon.co.uk/Modern-Information-Retrieval-Concepts-Technology/dp/0321416910">modern Information Retrieval</a>, has recently devoted much of his work in the study of how biases emerge online and interact with each other. There are biases of all kind. Most of the content online is produced by a tiny fraction of Web users; we consume that content through interfaces that catch our attention in predictable ways; the information we access is filtered and ranked by algorithms which, in turn, are trained on biased behavioral data. These <strong>biases</strong> have very different roots (cultural, statistical, cognitive) and <a href="https://cacm.acm.org/magazines/2018/6/228035-bias-on-the-web/fulltext"><strong>are all entangled</strong></a><strong> in a Gordian knot that is impossible to untie</strong> (and it would be probably absurd to cut it off with the swing of a sword either). Some solutions indicated by Ricardo recall Fenrose’s recipes. He advocates for going back from a <em>“human in the loop”</em> paradigm to the <em>“human at the center”</em> philosophy, which entails the adoption of better policies and a more conscientious use of Big Data (by transitioning to <a href="https://www.kdnuggets.com/2018/10/big-small-right-data.html">small data</a> when needed). He concluded with an optimistic rebuttal to Noah Harari’s <a href="https://www.amazon.co.uk/Lessons-21st-Century-Yuval-Harari/dp/1787330672">last book</a>, leaving the audience with the hope that benefits of AI greatly outweigh its potential risks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-gV2cVIBdrL_qhCYVXYDTQ.jpeg" /><figcaption>Debunking attempts of resurrecting physiognomy</figcaption></figure><p><strong>Learning Bias Mitigation (Margaret Mitchell, Google).</strong> Try to play this little game: take a look at an image of <a href="https://thewateringmouth.com/wp-content/uploads/2016/04/bananas-in-a-basket-1024x680.png">bananas</a> and come up with five terms to describe it. Done? It’s quite likely that your list of words won’t include the color of those bananas. This is because our brain has a stored representation of reality which does not necessarily reflect what happens in the world: in our mind bananas are yellow and we don’t need to mention that because it sounds obvious. Similarly, we associate the word “doctor” to a male figure and “couple” to heterosexual partners. When describing things around us, we report surprising experiences (a murder) and gloss over obvious and very frequent facts (a person blinking). <strong>The online data we use to train machine learning models are soaked in such <em>human-reporting biases</em></strong> and propagate in all the automated pipelines of our algorithmic systems (a phenomenon called <em>“bias network effect”</em> or <em>“bias laundering”</em>), ultimately leading to the amplification of injustice. Margaret Mitchell is a scientist working on AI at Google and she is interested in understanding how we can mitigate the biases in machine learning. The are glaring examples of these biases percolating into machine learning models, including the latest <a href="https://en.wikipedia.org/wiki/Cesare_Lombroso">lombrosian</a> algorithms that detect <a href="https://medium.com/@blaisea/physiognomys-new-clothes-f2d4b59fdd6a">criminals </a>and <a href="https://medium.com/@blaisea/do-algorithms-reveal-sexual-orientation-or-just-expose-our-stereotypes-d998fafdf477">homosexuals</a> from faces, which Dr. Mitchell promptly debunked. She then proposed a number of ways to mitigate biases, the most interesting of which is an <a href="https://arxiv.org/abs/1801.07593">adversarial approach</a>. The model simultaneously learns a predictor and an adversary: the objective is to maximize the predictor’s ability to predict the outcome variable while minimizing the adversary’s ability to predict the protected attribute.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r2bw5NR4nwUpsY264U8jgQ.jpeg" /><figcaption>Maximizing diversity of information in social networks</figcaption></figure><p><strong>Breaking the Filter Bubble with Algorithms (Aris Gionis, Aalto University).</strong> <a href="https://en.wikipedia.org/wiki/Echo_chamber_(media)">Echo chambers</a> are yet another facet of social media biases. Aris Gionis, professor at Aalto university and newly-nominated ISI Fellow, provided three possible algorithmic solutions to burst those filter bubbles, or at least to mitigate their effect. The first mitigation strategy is to improve awareness. He proposes to do so by mapping users and content from Twitter into a latent space which allows to automatically compute an <em>ideology score</em> that can be associated to the user. The second is to <a href="https://arxiv.org/pdf/1809.04393.pdf">maximize the diversity of content</a> users are usually exposed to in their social circles (<em>“tell me something that my friends don’t know”</em>), which turns out to be a NP-hard problem also in its approximated version. The last is selecting people who are good representatives of opposing ideologies but whose agreement is as high as possible. He concluded with a warning, though: we don’t know to which extent our attempts of <strong>reducing polarization might actually backfire and increase conflict</strong> rather than reducing it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qafd_s8za7saxNFP16QpZA.jpeg" /><figcaption>Social contagion is one of the fundamental social processes that has impact on a number of societal outcomes</figcaption></figure><p><strong>Understanding Collective Phenomena for Social Development (Marton Karsai, École Normale Supérieure de Lyon).</strong> Understanding social phenomena on the Web is the first step to turn technology into something that improves people’s life. Marton Karsai, a complex system scientist who masters data science and computational methods, has studied large-scale online data to link collective social phenomena to social development. He studied <a href="http://rsif.royalsocietypublishing.org/content/11/101/20140694">social influence</a> on the Skype social network, the relationship between the <a href="https://arxiv.org/pdf/1804.01155.pdf">use of language and socioeconomic status</a>, and how <a href="https://arxiv.org/pdf/1801.08856">homophily is impacted by economic status</a>. In this last work, he looked at an impressive dataset that matches mobile calls and credit information history and found that <strong>social classes are homophilous</strong> and the richest class is much better connected than the poorer ones. His methods and data open up to many opportunities to corroborate theories from social sciences about class mobility (<a href="https://en.wikipedia.org/wiki/Pierre_Bourdieu">Bordieu’s theory</a> is the first that comes to mind). Prof. Karsai is also one of the recipients of the ISI fellowship.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZeLGSZJ_CTGXOMVZh3qgqQ.jpeg" /><figcaption>Advancing public policy by relying on algorithmic systems can be quite a nightmare</figcaption></figure><p><strong>The Governance-by-Design Dystopia (Deirdre Mulligan, UC Berkley).</strong> Many speakers pointed out the importance of smart policies to save ourselves from the potential risks of new technologies. We were lucky enough to be able to listen to an expert in the domain: Deirdre K. Mulligan, professor of law at the UC Berkeley School of Information, an academic at the forefront of the endeavor of designing new policies for <a href="https://mitpress.mit.edu/books/privacy-ground">privacy management</a>. According to Prof. Mulligan, we live already in an era of <em>“</em><a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3175017"><em>Governance-By-Design</em></a><em>”</em>: we use and design technological systems for the advancement of public policy. Even if governing through technology is a quite seductive concept, she acknowledges that, at the moment, this approach suffers from severe limitations. To illustrate her point, she told the stories of four famous cases in which law and technology clashed in various ways: the <a href="https://en.wikipedia.org/wiki/FBI%E2%80%93Apple_encryption_dispute">Apple vs. FBI</a> encryption war; the EU General Data Protection Regulation (<a href="https://eugdpr.org/">GDPR</a>); the Stop Online Piracy Act (<a href="https://en.wikipedia.org/wiki/Stop_Online_Piracy_Act">SOPA</a>), and online voting. GDPR is a good representative example of how achieving privacy-by-design is hard: it is tough to foresee what the implications of GDPR might be on future businesses and on desirable properties orthogonal to privacy, such as fairness. Unfortunately, the inability of estimating the impact of design choices on future developments is one of the curses of design in general (read the story of <a href="https://classicnewyorkhistory.com/how-robert-moses-shaped-the-long-island-he-misunderstood/#_ftn1">Robert Moses</a> for an example in the domain of urban design). However, prof. Mulligan offered some suggestions on how governance-by-design could be improved. 1) We should <strong>design with modesty</strong> and do not dictate exactly what to do because it’s hard to predict the future; 2) We should improve the <strong>technical expertise of regulators</strong>; 3) We need to <strong>keep the public in the loop</strong> of the discussions about policies. The importance of those propositions becomes apparent when thinking that even the choice of a simple threshold in a classification system becomes a heavy political responsibility when the system has the power of impacting the life of many.</p><h3>Mapping health, culture, and well-being</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9rNt6uxlJ2sqiamHJ-g85g.jpeg" /><figcaption>A platform to monitor exposure and estimating impact on health</figcaption></figure><p><strong>Mapping Air Pollution Mortality (Francesca Dominici, Harvard University).</strong> <br>Only in US, thousands of people die every year because of air pollution. Francesca Dominici, an interdisciplinary statistician who has built a stellar career by linking Big Data with health outcomes, talked about her experience in studying this silent killer. She has developed a neural network model that uses on-the-ground air-monitoring data and satellite-based measurements to estimate daily pollution levels across US at very high granularity. By matching that information with Medicare data (460 million health records covering 97% of the population ages 65 or older) she showed that <strong>exposure to </strong><a href="https://www.nejm.org/doi/full/10.1056/NEJMoa1702747"><strong>air pollution is killing</strong></a><strong> thousands of senior citizens each year</strong>. These people could be saved by just cutting 1 microgram of fine particulate matter per cubic meter of air below current standards. The richness of the data she has studied opens wide opportunities but poses also challenges. She discussed how the threat of unmeasured confounding bias is amplified, and causality is even harder to assess with observational studies. More about her project in this <a href="https://www.hsph.harvard.edu/news/multimedia-article/harvard-chan-this-week-in-health-archive/">podcast</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aYQ30Y-ixobLxcB0lF_KFQ.jpeg" /><figcaption>The intangible layers of the urban fabric</figcaption></figure><p><strong>Spatial AI for Health and Well-being (Rossano Schifanella, University of Torino).</strong> The surrounding environment influences us in ways that we often don’t consciously perceive. Upon receiving his ISI Fellowship, Prof. Rossano Schifanella talked about his extraordinary research journey to map the urban space with the aim of improving well-being through new services and design policies. He argued that <strong>well-being is connected to intangible aspects of urban life</strong>. Over the years, he mined data from social media and a variety of other sources to build a multi-layer model of intangible aspects that define the urban space. He mapped the sensory perceptions that people have while walking in the city (<a href="https://dl.acm.org/citation.cfm?id=2631799">visual beauty</a>, <a href="https://arxiv.org/abs/1505.06851">smell</a>, <a href="http://rsos.royalsocietypublishing.org/content/3/3/150690">sound</a>), the <a href="https://dl.acm.org/citation.cfm?id=2741631">walkability</a> of a street, the <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0198441">activities</a> that happen indoor and outdoor, and even the ambiance of neighborhoods. He has also worked to measure how physical transformations of the city can impact subjective well-being. Using data from a mobile operator, he has studied the impact of the construction of large-scale urban infrastructures on the behaviour of city dwellers. He also talked about his recent work in mapping the health of city dwellers using granular and large-scale medical data. Overall, an impressive range of studies about urban life, some of which are collected on the <a href="http://goodcitylife.org/">goodcitylife.org</a> portal.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*age4Bt_UT-Yl2m_GAzX7cg.jpeg" /><figcaption>Languages in tweets reveal the level of ethnic integration of cities</figcaption></figure><p><strong>Studying Geography from the Lens of Twitter (Bruno Gonçalves, JP Morgan).</strong> Bruno Gonçalves<strong> </strong>is a physicist, computer scientist, new ISI fellow, and prolific <em>twitterologist</em>. In his talk, He provided a visually-powerful overview of how geo-referenced tweets can be used to learn about cultural and social dynamics at world-scale and with an impressive granularity. He studied the multiplicity of languages used to tweet in cities to estimate their level of <a href="https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0191612">ethnic integration</a>. He also looked at how the discussion around a topic spreads spatially to spot the trend-setting cities and to identify the urban centers that follow them (the <a href="http://advances.sciencemag.org/content/advances/2/4/e1501158.full.pdf">paper</a> cleverly uses transfer entropy to estimate influence between cities). Last, he presented a <a href="https://arxiv.org/pdf/1512.07281.pdf">comparative study</a> between Twitter and Sina Weibo to cast a cautionary tale about data representativity: the observed characteristics of a phenomenon may greatly vary depending on the social media used as a platform of analysis. So, <strong>pick the appropriate data</strong> to answer your research question.</p><h3>What else?</h3><p>There was so much going on including more parallel sessions with scientific presentations and tutorials, a great social dinner and celebrations of all sorts. In this post I reported a selection of the presentations I saw but check out the DSAA <a href="https://dsaa2018.isi.it/program/program-at-a-glance">program</a> for more pointers.</p><h3>Until next time…</h3><p>The Science Crossroads will be held next year in Torino, as usual. The next edition DSAA will be hosted in Washington, DC. I hope the organizers will keep the tradition of having an industrial day, this year it was terrific!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bnuGAnAek0Lu6MJv8_AgJg.jpeg" /><figcaption>DSAA 2019 in Washington, DC</figcaption></figure><p>In the meantime, if you happen to be in UK in December, you should consider dropping by the Complex Network Conference in Cambridge. Hope to see you there!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TqWIVnAbIWMo5zqyISSnsg.jpeg" /><figcaption>Complex Networks conference in Cambridge, UK</figcaption></figure><p>Thanks for reading! If you liked the post, please hit the applause 👏 button. You can also connect with me on <a href="https://twitter.com/lajello">Twitter</a> to share ideas or give suggestions.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dc9fdca18389" width="1" height="1" alt=""><hr><p><a href="https://medium.com/socialdynamics/science-reversions-in-torino-dsaa18-dc9fdca18389">Science Reversions in Torino: DSAA’18</a> was originally published in <a href="https://medium.com/socialdynamics">SocialDynamics</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Network Wanderings in Paris: Netsci 2018]]></title>
            <link>https://medium.com/socialdynamics/network-wanderings-in-paris-netsci-2018-df331a3fa93e?source=rss-547f52530508------2</link>
            <guid isPermaLink="false">https://medium.com/p/df331a3fa93e</guid>
            <category><![CDATA[bell-labs]]></category>
            <category><![CDATA[paris]]></category>
            <category><![CDATA[network-science]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Luca Maria Aiello]]></dc:creator>
            <pubDate>Thu, 21 Jun 2018 15:39:02 GMT</pubDate>
            <atom:updated>2018-06-22T14:54:32.418Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_pW8vhlvKOzDAYcApMybfQ.png" /></figure><p>Record edition for the <a href="https://netsci2018.com/">13th conference on Network Science</a>: almost 800 attendees and a long waiting list. Vittoria Colizza and Alain Barrat did a great job at organizing the event, that turned out to be a great success. Besides many excellent keynotes and two days of satellite meetings, the conference featured five parallel tracks with contributions on theory of complex systems, epidemics &amp; spreading, biology, economics, and socio-technical systems. I have attended the <a href="http://www.quantifysuccess.org/netsci2018/">Quantifying Success</a> and the <a href="https://mlns18.tumblr.com/">Machine Learning in Network Science</a> satellites and, during the main event, I stuck to the Social Systems track. Here’s the summary of the week’s highlights, from the perspective of a Computer Scientist.</p><h3>The brain’s wanderings</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v_WlzNXCA15rYqAi1T9bqQ.png" /><figcaption>The brain transitions more effectively between concepts that are linked in tight communities rather between those connected by bridges — Danielle Bassett’s keynote for the Erdős-Rényi Prize.</figcaption></figure><p><strong>Network Wanderings (Danielle S. Bassett, Penn Engineering) </strong>In 2017, <a href="https://en.wikipedia.org/wiki/Danielle_Bassett">Danielle Bassett</a> received the <a href="https://medium.com/penn-engineering/penn-engineerings-danielle-bassett-wins-top-prize-in-complexity-science-59943b8eca79">Lagrange prize</a> (a top prize in complexity science) for her pioneering work in bringing together network science and neuroscience. This year, she is the recipient of the The <a href="http://www.netscisociety.net/erdosrenyiprize.html">Erdős-Rényi Prize</a>, a prestigious award that is given every year by the Network Science Society to a scientist under-40 to recognize special achievements in the field of network science. In her keynote, she provided an overview of the remarkable work that her team has done in recent years to understand how the brain works (or fails to work properly under degenerative conditions) and how network science can contribute to crack longstanding questions about how to draw and interpret functional maps of the brain. But the most fascinating concept in her presentation is the one of <strong><em>Network Wanderings</em></strong>. Simply put, given a graph of concepts, the question is what is the most effective way to navigate this graph for a person to effectively learn those concepts. In her <a href="https://arxiv.org/pdf/1805.12491.pdf">recent work</a>, Prof. Bassett has shown that some network topologies can ease the learning process. She conducted experiments with participants exposed to sequences of visual stimula. For each stimulus, subjects were asked to respond by pressing a combination of one or two buttons on a keyboard. The sequences were produced based on a pre-determined graph of transitions between stimula. They found that pairs of sequences connected by highly embedded edges were reproduced more effectively than pairs across bridging edges, even when the node degree in the network was constant. This is a very powerful idea that applies not only to the brain but might apply to online learning, recommendations, and consumption of information. <em>Congratulations, Danielle!</em></p><h3>Misinformation, biases, inequality</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x-95F703apDOYC2ZQ_0LIw.jpeg" /><figcaption>Social influence + a tiny bit of unfriending people with different views dooms the social network to eternal segregation (and misinformation) — Fil Menczer’s keynote</figcaption></figure><p><strong>Top 5 reasons why misinformation spreads — #1 will (not) shock you! (Fil Menczer Indiana University).</strong> <a href="http://cnets.indiana.edu/fil/">Fil Menczer</a> has studied the phenomenon of (mis)information spreading online for more than a decade. The research work he has led, and the tools his team has developed to build a more trustworthy Web are collected under the <a href="https://osome.iuni.iu.edu/">OSoMe</a> project page. In his very engaging keynote, Fil has shown that the interplay between cognitive, social, and algorithmic biases makes us vulnerable to misinformation. Segregation is a key ingredient; he showed that, adding a low probability of homophily-driven unfriending to a social influence model, people segregate in groups and ideas stop spreading. He also showed that high-quality content emerges in scenarios where<em> i) </em>information load is low <em>ii) </em>the cost of accessing to new information is high, and <em>iii)</em> the content is consumed proportionally to its popularity. On top of those factors, the #1 reason why social networks effectively spread misinformation is that Bots can exploit very efficiently all these factors at once, thus being able to amplify the echo of fake news in the online ecosystem. When asked about the generality of Twitter-based studies and the actual impact of Twitter on the real-word societal dynamics, he provided two quite interesting points: many social platforms provide unmonitored, Twitter-like methods to spread misinformation (see Facebook pages) and Twitter is very often used by journalists to write news that are then broadcast to a much wider audience.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4UqYXbqQ4Eo-dGXGpxbpug.jpeg" /><figcaption><a href="https://academic.oup.com/joc/article-abstract/65/6/932/4082320?redirectedFrom=PDF">#myNYPD</a>: Mainstream vs. counterpublic perception of the relationship between police officers and residents — Brooke Foucault Welles’ plenary talk</figcaption></figure><p><strong>The power of networked counterpublics (Brooke Foucalt Welles, Northeastern University). </strong>If on one hand fake news pollute the social and political discourse on social media, on the other hand there are communities of online activism that provide critical mechanisms to overcome mainstream news narratives and empower marginalized groups. In her plenary talk, <a href="https://twitter.com/foucaultwelles/status/1006585318211031040">freshly tenured</a> professor <a href="https://camd.northeastern.edu/commstudies/people/brooke-foucault-welles/">Brooke Foucalt Welles</a> builds on Warner’s theory of <a href="https://mitpress.mit.edu/books/publics-and-counterpublics">counterpublics</a> and talks about networked counterpublics in online communities. Starting from the tragic story of 18-year-old Michael Brown, shot and killed by a police officer in Ferguson, Missouri, Prof. Welles tells the tale of how people started tweeting about it. Contributing to hashtags of social awareness is often labeled as “lazy activism”, but she showed that networked counterpublics, organized by hashtag, can actually change drive social change by providing a collective perspective on facts that would not emerge otherwise. Her final recommendation is to “Keep tweeting!” because it makes the difference. Her book “#HashtagActivism: Race and Gender in America’s Networked Counterpublics” will be out in 2019.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lmXXjO4TOAD4u4LusxMRLQ.jpeg" /><figcaption>Algorithmic biases and limited access to social capital impact the online visibility of women and minority groups — Claudia Wagner’s plenary talk</figcaption></figure><p><strong>Minorities in social networks (Claudia Wagner, GESIS).</strong> Studies about inequality and biases are trending the computational social science community, not least because their importance in explaining (and alleviating) much of the difficulties that minorities and underrepresented categories experience everyday. <a href="http://www.claudiawagner.info/">Claudia Wagner</a> and her team at <a href="https://www.gesis.org/">GESIS</a> are working to measure how algorithmic biases and limited access to social capital impact the online visibility of women and minority groups. They are doing it looking at multiple settings and datasets. A new <a href="https://arxiv.org/abs/1702.00150">network model</a> they developed proves how homophily and preferential attachment confines minorities in a condition of low visibility. An analysis of <a href="https://link.springer.com/content/pdf/10.1140%2Fepjds%2Fs13688-016-0066-4.pdf">Wikipedia biographies</a> reveals an impressive gap between men and women. Last, women have weaker and less embedded ties in networks of <a href="https://www.worldscientific.com/doi/pdfplus/10.1142/S0219525917500114">scientific collaborations</a>. The road to gender equality is uphill…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/516/1*ptYP59TS0rCRsBj_SkmoRw.jpeg" /><figcaption>Visual content to tear down barriers— Miriam Redi at the Machine Learning satellite</figcaption></figure><p><strong>Machine learning for (visual) knowledge equity (Miriam Redi, Wikimedia Foundation).</strong> Similar to the real word, the online world is full of barriers that are hard to overcome. The barriers arising from the differences in culture and language are those that <a href="http://www.visionresearchwitch.com/">Miriam Redi</a>, together with her fellow teammates at the <a href="https://wikimediafoundation.org/wiki/Home">Wikimedia Foundation</a>, are trying to overcome. She gave us an overview of a number of projects currently under development at Wikimedia. Two of them struck me the most. The first is about automatically learning when a statement <a href="https://meta.wikimedia.org/wiki/Research:Identification_of_Unsourced_Statements#Citation_Need_Modeling">needs a citation</a>; this apparently narrow task hides a number of fundamental questions, for example understanding what can be considered “common knowledge” in a culture. The latter, still in its infancy, is to learn how to use new forms of visual language to enrich and complement knowledge and overcome language barriers. Given Dr. Redi’s expertise in measuring <a href="http://www.aaai.org/ocs/index.php/ICWSM/ICWSM15/paper/download/10547/10522">intangible properties</a> of images and training <a href="https://arxiv.org/pdf/1508.03868.pdf">cross-cultural models</a> for visual sentiment concept ontologies, my bet is that the project has very good chances to turn into a success.</p><h3>Structure and evolution of social networks</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vet-5hLtHk0w_R9IrOJg6Q.jpeg" /><figcaption>Normality as a new way to measure the quality of communities and to identify anomalous ones— Leman Akoglu at the Machine Learning satellite</figcaption></figure><p><strong>Attributed networks: Social Circles, summarization, comparison (Leman Akoglu, Carnegie Mellon).</strong> Given an attributed subgraph representing a community (a set of nodes annotated with labels), how do we evaluate its quality? <a href="http://www.andrew.cmu.edu/user/lakoglu/">Leman Akoglu</a> introduces the concept of <a href="https://arxiv.org/abs/1601.06711">Normality</a> for network communities. Normality takes into account the network structure and attributes together to quantify both internal consistency and external separability. This approach allows many boundary-edges as long as they come from external hubs (which are connected to many nodes, so they can be “discounted”).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pGikIkNuMerWcJLhWT6fIQ.jpeg" /><figcaption>Discovering roles in networks — Tina Eliassi-Rad and collaborators</figcaption></figure><p><strong>Role discovery in networks (Tina Eliassi-Rad, Northeastern University). </strong>Community detection on graphs focuses on finding groups of highly connected nodes. Role discovery on graphs goes beyond grouping nodes based on their connectivity and embeddedness and finds groups of nodes that share similar graph topological structure. The idea makes a lot of sense but the first practical methods to implement it have been proposed only very <a href="https://aaltodoc.aalto.fi/bitstream/handle/123456789/24447/master_Arockiasamy_Albert_2017.pdf?sequence=1&amp;isAllowed=y">recently</a>. <a href="http://eliassi.org/">Tina Eliassi-Rad</a> and her collaborators propose a <a href="https://arxiv.org/pdf/1609.02646">new method</a> to perform this task in a supervised fashion. The results look quite compelling.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WiAFiSpbqpjp7PDy0HYCgg.jpeg" /><figcaption>Strength or embeddedness: which comes first? — Esteban Moro on strong ties</figcaption></figure><p><strong>Tie strength precedes social embeddedness (Esteban Moro, Carlos III University). </strong>Considering the humongous amount of work done inthe past to predict missing/future links in social networks, it is surprising that comparatively little work has been done on predicting ties strength. <a href="http://estebanmoro.org/">Esteban Moro</a> took on this challenge starting from its underlying theoretical foundations. A few decades ago, Mark Granovetter theorized that tie strength and network embeddedness (i.e., the number of common neighbors between the two linked individuals) are deeply intertwined concepts. Prof. Moro analyzed traces from call data records to determine if tie strength comes before or after embeddedness. It <a href="https://link.springer.com/article/10.1140/epjds/s13688-017-0127-3">turns out</a> that, usually, strong links are created first and common neighbors grow around them in a later stage. He then introduced some results about link strength prediction, showing that classic features used in link prediction are way less effective for this task.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wspq9kX1KXgdvRpQyFfp1w.jpeg" /><figcaption>Network model of innovation processes reproduces Heap’s Law — Iacopo Iacopini</figcaption></figure><p><strong>Network dynamics of innovation processes (Iacopo Iacopini, Queen Mary University London).</strong> Processes that operate on semantic spaces oscillate between states of exploration and exploitation. For example, when authors write books, they use a recurrent core set of words and, occasionally, they use new ones. This type of processes originate the so-called <a href="https://en.wikipedia.org/wiki/Heaps%27_law">Heap’s Law</a>, which describes the scaling of the number of distinct elements as a function of the total number of elements used in the process. Empirically, in most scenarios this is an exponential law with an exponent lower than 1. <a href="https://iaciac.github.io/">Iacopo Iacopini</a> introduced a <a href="https://arxiv.org/abs/1707.04239">model</a> for the emergence of innovations, in which cognitive processes are described as edge-reinforced random walks on the network of links among concepts, and an innovation corresponds to the first visit of a node. The model is simple and elegant and it is able to reproduce Heap’s laws for the emergence of new topics in scientific papers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F1oA6Es20SLoe6bPl0UYZw.jpeg" /><figcaption>Discovering how people perceive their social interactions can help us build better social network models — Luca Aiello at the Social Systems track</figcaption></figure><p><strong>The human perception of social relationships (Luca Maria Aiello, Nokia Bell Labs).</strong> Graphs are convenient abstractions to model social interactions but every time we draw a social graph we make an implicit assumption about what links between nodes mean. The meaning of social links can be represented at different <a href="http://www.lajello.com/papers/snam17.pdf">levels of abstraction</a>: they can encode tie strength, polarity, sentiment, topics of conversation, and so on. In <a href="https://arxiv.org/abs/1407.5547">previous work</a>, stemming from Peter Blau’s theory of <a href="https://www.amazon.co.uk/Exchange-Power-Social-Life-Peter/dp/0887386288">social exchange</a>, we have have proposed a method that pushes the representation of social links to a new level. Given a conversation graph (a social network with text messages on the edges) the method could estimate whether the relationship is about social support, knowledge exchange, or status exchange. Those three aspects are not exhaustive of all types of social interactions though. Recently, we have conducted a literature review and a crowdsourcing study to discover what are the fundamental relationship types that people and that sociologists has studied. It turns out that there are 10 and that they can help us to cast better predictions on how social networks evolve. <a href="http://www.lajello.com/">Stay tuned </a>for the upcoming paper.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q2BEkXZhYGEcBnwWzoSs7Q.jpeg" /><figcaption>Long-term product adoption is a function of several factors that go beyond social contagion — Shankar Iyer</figcaption></figure><p><strong>Overseeding in social networks (Shankar Iyer, Facebook). </strong>Cracking the problem of influence maximization in real social systems remains the Holy Grail of all provides of social Web services. <a href="https://www.linkedin.com/in/shankar-iyer-b2065b53/">Shankar Iyer</a> and <a href="http://www.ladamic.com/">Lada Adamic</a> propose a more nuanced approach to the problem that makes an attempt to go beyond some of the oversimplified network models used in the past. Factors for the adoption of a new product or idea cannot be exhaustively captured by the network topology only. The need of supporting ties, the cold-start process of adjustment to the novelty, and the risk of permanent churn are contextual factors that might play a major role too. If those elements are not taken into account, massively seeding a new product in the social network might actually reduce its long-term adoption.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mW8j0sRirQP1OiadwZwg3A.jpeg" /><figcaption>Contagion models: no model rules them all — Meysam Alizadeh</figcaption></figure><p><strong>Social contagion of political extremism (Meysam Alizadeh, Indiana University).</strong> There are several models of network contagion. <a href="https://www.jstor.org/stable/10.1086/521848?seq=1#page_scan_tab_contents">Complex contagion</a>, <a href="http://www.pnas.org/content/109/16/5962.full.pdf">structural diversity</a>, and <a href="https://pubsonline.informs.org/doi/abs/10.1287/mnsc.2014.1936">embeddedness </a>are among the most established ones. They are partially conflicting as they rely on different structural indicators to model contagion. <a href="http://cnets.indiana.edu/groups/santo-lab/meysam-alizadeh/">Meysam Alizadeh</a> (together with <a href="http://cnets.indiana.edu/people/santo-fortunato/">Santo Fortunato</a>, <a href="https://ingmarweber.de/">Ingmar Weber</a> and <a href="http://infosci.cornell.edu/faculty/michael-macy">Michael Macy</a>) compared them in the setting of online political discussion, with a special attention to supporters of extremist parties. The result is that there’s no single model that is best suited for all settings. This finding suggests that contagion dynamics might be driven by different mechanisms, depending on the context.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j7_klQfch3vrgOCQo9saow.jpeg" /><figcaption>Bringing the backboning algorithms to the extreme — Luis M Rocha at the plenary session</figcaption></figure><p><strong>The Metric Backbone of Contact Networks in Epidemic Spread Models (Luis M Rocha, Indiana University).</strong> A backbone graph is a subgraph that preserves the main structural properties of the original graph but number of edges is much smaller than in the original one. The <a href="http://www.pnas.org/content/106/16/6483">disparity filter</a> and the <a href="https://arxiv.org/abs/1701.07336">noise-corrected backbone</a> are some of the backboning algorithms that have been proposed in the past. <a href="https://www.informatics.indiana.edu/rocha">Luis Rocha</a> and collaborators proposed a new approach that brings network backboning for weighted networks to its extreme. They have tested it on several dataset ranging from networks of axonal connections to graphs of social interactions from the <a href="http://www.sociopatterns.org/">sociopatterns </a>project and they shown amazing results: the resulting reduced graphs preserve shortest paths but erase up to ~90% of the nodes. Details and code will be available <a href="https://www.informatics.indiana.edu/rocha/publications/SM/index.html">here</a>.</p><h3>Mobility</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H6XvPLogbR5aVEE7G-RIsg.jpeg" /><figcaption>Discovering classes of users from mobility data — Markus Strohmaier’s keynote at the Machine Learning satellite</figcaption></figure><p><strong>Understanding markovian network dynamics (Markus Strohmaier, Aachen University).</strong> In his opening keynote at the Machine Learning satellite, Prof. <a href="http://markusstrohmaier.info/">Markus Strohmaier</a> gave an overview of an approach to discover classes of users that navigate graphs. The <a href="https://arxiv.org/abs/1411.2844">approach</a> uses the concept of exceptional model mining and it applies to any type of navigation, from web browsing, business reviews and online music played. The most compelling use study presented is on human mobility from temporal Flickr data. The technique not only extracts user classes but it also uses the attributes of the nodes to provide meaningful explanations of those output classes. A great approach for everyone who’s working with mobility data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1wT6VaU9SiNr_tmQRa5cSA.jpeg" /><figcaption>Field theory applied to human mobility — José Ramasco</figcaption></figure><p><strong>Field theory for recurrent mobility (José Ramasco, IFISC).</strong> Two among the most established frameworks to model mobility are the gravity model and the radiation model. <a href="https://ifisc.uib-csic.es/~jramasco">José Ramasco</a> introduced a new perspective on the problem by using field theory formalism. In a nutshell, he models each agent that moves into space with a vector that expresses its average mobility direction and intensity. The approach naturally lends itself to find centers on mobility, also in polycentric cities. A very nice theoretical take on the theme of urban mobility.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HmZmPUKjiy-Jnrjj7gLt5A.jpeg" /><figcaption>The Gini Index of places of gathering in Boston — Esteban Moro</figcaption></figure><p><strong>Economical Segregation of Encounter Networks in cities (Esteban Moro, Carlos III University).</strong> There are many things you can do when you have fine grained mobility data and accurate estimates of socio-economic indicators at individual level. <a href="http://estebanmoro.org/">Esteban Moro</a> and other folks at MIT use that type of data to estimate how much places of gathering mix people who have different economic conditions. The Gini index of hotspots vary greatly across the city. The paper is not yet public but will be coming out soon, looking forward to it.</p><h3>Networks and Culture</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fAKW5cV4ccaVP3wwbdgJWA.jpeg" /><figcaption>The perfume smell wheel — Vaiva Vasiliauskaite at the plenary session</figcaption></figure><p><strong>Network analysis of perfumes (Vaiva Vasiliauskaite, Imperial College London).</strong> Remember the <a href="https://www.nature.com/articles/srep00196">flavor network</a> of food pairings? Vaiva Vasiliauskaite and <a href="https://www.imperial.ac.uk/people/t.evans">Tim Evans</a> worked on a similar concept but with perfumes. Using <a href="http://www.thegoodscentscompany.com/">data</a> of perfume descriptions, they built and analyze the network of similarities between perfumes that share common descriptors. Using additional <a href="https://www.fragrantica.com/">data</a> on perfume composition, they study which components are those that best enhance perfumes. They also produced a nice smell wheel that reminded me the categorization of urban smells from the <a href="https://arxiv.org/abs/1505.06851">Smelly Maps</a> paper I co-authored. Well done, Vaiva. Very nice work!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wbrz7QNaBNf9LHuEB2xKeA.jpeg" /><figcaption>Party Cannon don’t give no f**cks — Peter Klimelek on cultural elites</figcaption></figure><p><strong>Evolution of cultural elites: quantitative evidence from fashion (Peter Klimek, Robert Kreuzbauer, Stefan Thurner). </strong>The nature of social dynamics that underlies changes in art and fashion has been <a href="https://www.jstor.org/stable/2118231?seq=1#page_scan_tab_contents">debated </a>for long time. Using a large dataset of 8 million musical albums release over 60 years, Peter Klimek and colleagues found a strong evidence that fashion cycles in musical styles are driven by an evolutionary process that can be explained using <a href="http://www.ceacb.ucl.ac.uk/cultureclub/files/2004-2005/CC2005-06-02-Bird_and_Smith_2005.pdf">signaling theory</a>: individuals with high status who can afford sending strong signals to the majority of low-status individuals are trend-setters. Very fun examples from logos of metal bands. The paper will be out soon on the arXiv.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AS28R6GLAnBS4PVMJ72bog.jpeg" /></figure><p><strong>Social Influence in Music Listenership (Taha Yasseri, Oxford Internet Institute). </strong>Can live music events generate complex contagion in music streaming? <a href="https://www.oii.ox.ac.uk/people/taha-yasseri/">Taha Yasseri</a> found positive evidence by studying a large dataset from Last.fm. Using regression discontinuity analysis to estimate causal patterns, he <a href="https://arxiv.org/abs/1711.05701">observed</a> that concert attendance has an effect on music listenership not only among attendees but also in their network of friends. One of the most entertaining talks of the conference, with a legendary <a href="https://twitter.com/TahaYasseri/status/1007252930033061888">selfie</a> with the audience and a spicy remark to the infamous Reviewer#2 who rejected his submission.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6_i_lj75E1e-LNEPW4Qyxg.jpeg" /></figure><p><strong>Political coalitions and divisions on Twitter in 18 countries (Livia Teernstra, Justus Uitermark). </strong>The polarization and fragmentation of political discourse between political party members varies quite a lot across countries. Livia Teernstra and Justus Uitermark have found four types of network configuations. At the two extremes, Greece (with no dialogue between parties) and Finland (a virtuous example of dialogue and cooperation between parties).</p><h3>Quantifying quality and success</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jB8bNiMggYpoF3EN2Y2f4Q.jpeg" /><figcaption>Machine vision algorithms can assess visual beauty similar to how the human eye would do — Rossano Schifanella at the Quantifying Success satellite</figcaption></figure><p><strong>Quantifying the effect of performance, success and social structures on user engagement (Rossano Schifanella, University of Turin).</strong> Web users have an intuitive understanding about how popularity of content is not always associated with quality. <a href="http://www.di.unito.it/~schifane/">Rossano Schifanella</a> takes a systematic approach to measure the distance between the two by using algorithmic tools that can <a href="http://www.di.unito.it/~schifane/papers/icwsm15beauty.pdf">measure visual beauty</a> of pictures. He showed that modern deep learning pipelines can estimate very reliably the visual appeal of a picture as perceived by human observers. By applying the algorithm at scale on the Flickr photo sharing site, he found that popularity of pictures is broadly distributed whereas picture beauty is normally distributed. This originates a strong mismatch: the vast majority of high-quality pictures enjoy little or no attention.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fZpwKbONlH9FUH9mqKdv6Q.jpeg" /><figcaption>People tend to increase the quality of their content after connecting with peers who produce high quality content — Luca Maria Aiello at the Machine Learning satellite</figcaption></figure><p><strong>The impact of social recommender systems (Luca Maria Aiello, Nokia Bell Labs).</strong> The struggle between popularity and quality can be the bane of recommendation systems too. In my invited talk at the Machine Learning satellite, <a href="http://www.lajello.com/">I</a> described how it might be convenient for companies to reinforce the rich-get-richer phenomenon of popular content as an easy strategy for monetization. But I also show that better alternatives are possible.</p><p>Simple <a href="http://www.lajello.com/papers/recsys14.pdf">contextualization strategies</a> or content recommendations based on quality, rather than popularity, are effective methods that can trigger also positive side effects. In a recently published <a href="http://www.lajello.com/papers/tkde.pdf">paper</a> we used causal inference from longitudinal data to observe how social recommendations that suggest individuals with a high-quality production on Flickr might be effective in increasing the future content production. There’s a catch though: being recommended only people who produce exceptionally good content increases our fear of inadequacy and, as a consequence, boosts the probability of disengagement and churn.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uecDxay7YybwidBKRFVr-g.jpeg" /><figcaption>The newtork of faculty hiring — Aaron Clauset at the Quantifying Success satellite</figcaption></figure><p><strong>Faculty hiring and the spread of scientific ideas (Aaron Clauset, Univeristy of Colorado at Boulder).</strong> <a href="http://tuvalu.santafe.edu/~aaronc/">Aaron Clauset</a> took a deep dive into the question of whether science is a real meritocracy. He did that by investigating the role of the faculty hiring process in <a href="https://arxiv.org/pdf/1805.09966.pdf">spreading of ideas</a> across communities. His analysis provides strong evidence that the faculty hiring network is the main factor that determines which ideas survive and prosper. He also shown that research from prestigious universities spreads more quickly and broadly than that from less prestigious ones. Compared to prestige, quality plays a little role in spreading: his models indicate that renowned institutions are more effective in diffusing low-quality ideas than minor institutions are in spreading high-quality ones.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*huNSZJ_InT-RrTwI6G7CNw.jpeg" /><figcaption>Research published in papers cited by multiple disciplines (the inner triangle) do not receive Nobel prizes — Roberta Sinatra</figcaption></figure><p><strong>Quantifying success in science and arts (Roberta Sinatra, Central European University).</strong> <a href="http://www.robertasinatra.com/">Roberta Sinatra</a> and her team have been studying the dynamics of success for quite some time and this year they have presented several studies that look at success in different disciplines and from multiple angles. Prof. Sinatra (whose <a href="https://twitter.com/robysinatra/status/1006158443844169729">newborn </a>daugther is perhaps the youngest attendee in the history of NetSci) presented her latest work about interdisciplinarity in science. Looking at how many citations a paper receives from different disciplines, she was able to place papers in an “interdisciplinarity space”. Most of the papers and all those which won Nobel prizes are placed at the border of this space (i.e., they’re cited by only one or two disciplines). Future will tell if this trend is bound to change anytime soon.</p><h4>Information is beautiful</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VvpOxqDNBUYYl3TTkdoeoQ.jpeg" /><figcaption>Sometimes, statistics alone does not tell the full story. Visualization comes to the rescue — Miriah Meyer’s plenary talk</figcaption></figure><p><strong>The process of designing effective and beautiful visualizations (Miriah Meyer, Univeristy of Utah, and Paolo Ciuccarelli, Density Design Lab).</strong> Visualization tools have become essential for extracting meaning from the immense amount of data we are faced with today. <a href="https://www.cs.utah.edu/~miriah/">Miriah Meyer</a> and <a href="https://densitydesign.org/person/paolo-ciuccarelli/">Paolo Ciuccarelli</a> talked about their experience of interacting with stakeholders whose work got great benefit from the use of visualization. In her keynote, Miriah gave the example of <a href="http://www.cs.utah.edu/~miriah/mizbee/Overview.html">MizBee</a>, a tool she built to study the co-localization of genes on the same chromosome within/across individuals or species. Before that tool, domain experts used to look at their data by drawing co-occurrence matrices that turned out to be largely ineffective, if not misleading. Using the tool, the project PI was able to get to new scientific discoveries. When asked:<em> “How long it would have taken to get there with no viz”</em> he replied: <em>“Honestly… I don’t think I would ever have gotten here.”</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ffeNGoVH7vDWUoYrTKxuzg.jpeg" /><figcaption>Making sense of “hairball” networks — Paolo Ciuccarelli’s keynote in the amazing social dinner venue</figcaption></figure><p>Paolo, in his inspiring dinner keynote, described a similar experience of working with network scientist stakeholders who badly wanted to produce an effective visualization a network structure only to realize after multiple iterations that the temporal processed over the network — and not the graph itself — was what they were interested in. The two main takeaways: 1) take a peek outside the set of tools you’re used to and don’t be afraid to shift visualization paradigm 2) The work of visualization experts is not only important to get appealing plots and interfaces but it is an integral part of the scientific method.</p><h3>What else?</h3><p>A lot of stuff! The event was huge and there was <a href="https://www.netsci2018.com/photos">so much going on</a>, including the great social events and celebrations. Here I’ve reported a selection of the presentation I saw but check out the <a href="https://www.netsci2018.com/conference-program">program </a>for more pointers.</p><h3>Until next time…</h3><p>The NetSci saga will continue next year in Chile and Vermont.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*W3FVnfNPD06KpegCDEN6rQ.jpeg" /><figcaption>NetSci and NetSciX in 2019</figcaption></figure><p>In the meantime, if you happen to be in UK in December, you should consider dropping by the <a href="https://www.complexnetworks.org/">Complex Network conference</a> in Cambridge. Hope to see you there!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b6iPcjcBR-DpDxY7PsFASw.jpeg" /><figcaption>Complex Networks conference in Cambridge, UK</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=df331a3fa93e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/socialdynamics/network-wanderings-in-paris-netsci-2018-df331a3fa93e">Network Wanderings in Paris: Netsci 2018</a> was originally published in <a href="https://medium.com/socialdynamics">SocialDynamics</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Complex Networks Conference Marseille]]></title>
            <link>https://medium.com/socialdynamics/complex-networks-conference-marseille-3d6b10e128b3?source=rss-547f52530508------2</link>
            <guid isPermaLink="false">https://medium.com/p/3d6b10e128b3</guid>
            <category><![CDATA[complex-systems]]></category>
            <dc:creator><![CDATA[Luca Maria Aiello]]></dc:creator>
            <pubDate>Fri, 15 Jul 2016 18:08:11 GMT</pubDate>
            <atom:updated>2016-07-18T10:45:44.172Z</atom:updated>
            <content:encoded><![CDATA[<p>Some non-exhaustive notes from the meeting.</p><p><strong>Alessandro Vespignani [Keynote].</strong></p><p>Comparison between the weather forecast and epidemics forecast: the reason why the first is at a level of commercial exploitation and the latter is not is because of lack of data, not lack of models.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WgD_3tSzJxncAXmvzmUwIw.jpeg" /></figure><p>Key factor is to monitor mobility effectively. Mobility was wave-like in the past (Black death in 1347 moved at 4 miles per day) and network-like today, with properties of multiplexity (interaction in different places, school workplace; multiple transportation systems, etc.).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Qna17446F6dBWt5BaODHoA.jpeg" /></figure><p>Do we have enough data to build a forecast for epidemics? <a href="http://www.gleamviz.org/">www.gleamviz.org</a> tries to do that by ingesting a model that are specific to each single epidemic outbreak (each outbreak has unique properties). The architecture is quite complex and mostly based on a network representation of metapopulations and their connections.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tnruLGhUhwQfCg20KQ17Tw.jpeg" /></figure><p>In 2014 the model was calibrated for Ebola, adding parameters about the impact funeral in the contagion process. The parameters are fitted to the existing data and then used to do forecast, with periodical recalibrations that have to deal with 1) lack of data and to 2) the impact that the forecast has on the behaviour of the population. So, besides the prediction, these models are useful for situational awareness and intervention planning.</p><p>In the case of Zika, there was an epidemic happening but no one knew until the microcephaly cases emerged. The predictive model for Zika is even more complex as the disease is conveyed by mosquitoes and metrics of population exposure risk like mosquito abundance, temperature/climate need to be considered. Gross domestic product (GECON) turns out to be an important feature as well. The Zika case shows how modeling might be useful to understand what happened in the past (but we didn’t notice) besides informing interventions (e.g., the opportunity to rush for a vaccine campaign).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C-xokTLWp0jIuLeAWdifgQ.jpeg" /></figure><p><strong>Johan Bollen [Keynote].</strong></p><p>Humanity is a social species that more than other species requires extensive network of social relationships. Social media is trying to satisfy this need online but, does social media make us happier? He argues that that’s likely not the case. This could be because of the “relative standing” phenomenon (the grass of your neighbor is always greener) which originates directly from the friendship paradox. If popularity and happiness are correlated, we will have an happiness paradox (my friends are on average happier than myself) which might lead to sort of collective depressive phenomena.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*I8dod9Dg7HnItA3rzP7-Pg.jpeg" /><figcaption>The “Menczer anticipatory vacation envy paradox” :)</figcaption></figure><p>Used 5M random twitter users (reduced to 40k after filtering) and their follow relations (reciprocal = friendship) and a well-being analysis over their tweets with Opinion Finder Lexicon. Pollyanna effect is found in data (the lexicon is almost always positive). Friendship paradox is very strong (95%). Happiness is assortative and also exhibits a strong (yet weaker) happiness paradox (60%). The correlation between happiness and degree is low though, meaning that the friendship paradox is not the only cause of the happiness paradox. Unhappy users are particularly exposed to the paradox (“if you’re unhappy you’d better just shut down you social media account”).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f9VmZpd_3zxmidGdhbIWTw.jpeg" /></figure><p>They also studied depression through twitter. Searched for twitter accounts publicly reporting their depression diagnosis. They did some textual feature extraction from those tweets (<strong>check this out:</strong> PCA from GloVe: word to vec developed at Stanford <a href="http://nlp.stanford.edu/projects/glove/">http://nlp.stanford.edu/projects/glove/</a>) and try to predict depression from those features. Accuracy is extremely high (it could be because of the very small test set — 60 positive examples — so the language of the positives might not overlap much with the language of negatives). An interesting observation is that sentiment analysis done on large textual corpora measures basically the sentiment of a whole language, so one should correct for that when studying sentiment in large populations. Open question is about the actual perception that people have of the happiness paradox.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U9P7bTLY3rmauEHFi7SUAA.jpeg" /></figure><p><strong>Daniela Paolotti [Keynote].</strong> Forecasting infulenza in real-time with participatory approaches. Mobility for instance can be estimated through Flickr geo traces. Twitter sentiment can correlate with the vaccine coverage for the pandemic (www.ZikaTracking.org). Wikipedia click counts on articles related to flu can predict the unfolding of an epidemics. However, the famous Google flu failure example reveals that passive data sources do not capture well the actual epidemics: a community of self-reported symptoms of flu can be the solution. <a href="http://www.influenzanet.eu/">www.influenzanet.eu</a> does that: participants across 10 countries in Europe fill general surveys (socio-economic, general health, demographics) and have weekly symptoms questionnaires. The crowdsourced data fits well with the actual spreading data, and correlates very well with data from self-collected bio samples in UK. Their data allows for and estimation of the different socio-demographic risk factors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ayZiceAppM3LKAccWtjatg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Tw4XfmpUCVMoOQM3IMsIbg.jpeg" /></figure><p>All this self-reported data can be used for prediction for seasonal flu when combined with the GLEAM model. Traditional surveillance models can do a forecast 2 weeks in advance, this works 1 month in advance. Results at <a href="http://www.fluoutlook.org./">www.fluoutlook.org.</a></p><p><strong>Maxime Lenormand. </strong>To rebalance spatial inequality in the city the idea is to use shopping trips to redistribute the flow of trips in the city and also the wealth flow. BBVA data of bi-partite network of purchases &lt;user,shop&gt;, the objective is to minimize skew in spending per area and maximize mixing of people going around in different areas. So the idea is to recommend shops to match these constraints but respecting the distribution of distance for individual mobility pattern (rewiring of the bipartite network). Inequality is reduced by 98% by rewiring 10% of shopping trips. The catch is that the *global* overall distance is preserved but not the ones at the level of individuals. Paper here: <a href="http://arxiv.org/pdf/1604.08394.pdf">http://arxiv.org/pdf/1604.08394.pdf</a></p><p><strong>Claire Lagesse.</strong> The analysis of a graph of street segments (segments are nodes, the connections between them are the edges) can reveal interesting structural patterns of the city. They argue that the segment perspective is too local and that a better way of representing this graph would be to have a graph of <em>ways</em> instead of segments. Ways are defined as sequences of segments with angles that are less than 60 degrees. In other words, the basic idea is to extend space syntax from the segment level to the way level. They build the way graph for four cities (Manhattan, London, Paris, Barcelona) and compute some structural properties such as centrality. Using ways instead of segments the correlation between length, degree, and betweenness are very high (not sure why this is a positive thing). They say that the structural maps they output are useful to inform interventions. For example, they helped town planners in Avignon to decide where the tram line should pass.</p><p><strong>Vittoria Colizza [Keynote].</strong> Again about disease spread but this time focused on animal vectors. Brucellosis is a livestock bacteria-induced disease that can transmit through contact and environmental contamination in between animals and to human. To model the transmission they use a vulnerability model trained on 6 years of data with a temporal network of transition of cattle in between farms. They estimate a threshold for the epidemic model in some way. Model to estimate disease spreading in temporal network.</p><p><strong>Victor Jirsa [Keynote].</strong> Modeling of the full brain as a network. Monumental work done by a large team, very technical details about specific technologies and models. An interesting info was his take on the FMRI “scandal” that recently emerged (basically, standard software in FMRI tools used an assumption that has been revealed as wrong). He said that if that is confirmed (and he seemed quite positive about it), the results of around 40,000 papers published in that area would be deeply impacted.</p><p><strong>Michele Tizzoni.</strong> Sociopatterns project deployed in African village. People tend to meet the same people every day. The distribution of contact duration is very stable over conditions, with high assortativity in demographic factors. Data available at www.sociopatterns.org/datasets</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SJ4fa_sXdUD4m9Zsc410ng.jpeg" /></figure><p><strong>Michele Starnini.</strong> Model for f2f interaction exist to reproduce statistical properties of duration and other descriptors. They extend those for groups. You can also measure how in this model an opinion can spread based on how moving agents collide.</p><p><strong>Luca Aiello.</strong> In online social networks, not all ties are created equal. The outcome of models and analysis from network science can change dramatically when the type of social structures (links, nodes, groups) is factored in. Three examples. <strong>First</strong>. structural properties of links created spontaneously are very different from the properties of links created as effect of an algorithmic recommendation. Matching experiments can be done to understand if recommenders lead to a “bubble” in which users exposed to recommendations tend to follow the same set of people (it turns out it is not the case). <strong>Second</strong>. Social links are carriers of non-material resources (as per Peter Blau’s exchange theory). A method is proposed to measure from text which resources are exchanged on a tie (e.g., affection, status, knowledge). If the overall graph is decomposed in resource-specific graphs, those exhibit very different properties (e.g., different assortativity). <strong>Third</strong>. The same reasoning can be extended to groups. The common identity and common bond theory can be operationalized to distinguish topical from social groups. The two group types behave differently in terms of external percolation of content diffusing. Also, the content percolation could be very high in some groups (e.g., adult communities) that are usually considered in isolation. Slides available here: <a href="http://www.lajello.com/slides/complenets16.pptx">http://www.lajello.com/slides/complenets16.pptx</a></p><p><strong>Jari Saramaki [Keynote].</strong> Studying dynamics of ego-networks with communication darta. Ego networks are maintained at a cost of time and brainpower (when you have 200 friends you are usually also aware of roughly 20k relationships between them). They studied the ego networks of 24 Oxford students with surveys and tracking of their phone calls. How do people divide their attention between contacts?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*iEzQ0ymHPo9nQBE8tHMlGg.jpeg" /></figure><p>They try to avoid averages when describing individuals so they extract a typical signature of interaction for each user. Findings are that we only have a few good friends with whom we communicate a lot and all individuals more or less the same signature (see pic above). The signatures are pretty much the same in other datasets. Matching the call data with the Big5 personality traits for all the participants, no relation emerges between personality traits and the signature. However, openness to experience and extraversion correlate with network turnover.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kyUsyISSr4Z12nBQc0oIhQ.jpeg" /></figure><p>They have also looked at circadian rythms. Those are determined by physiological internal cycles but can change from individual to individual. You can compare the daily activity of an individual with the population average</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7NpJ0oL77qwjx_XBOW2tWQ.jpeg" /><figcaption>Look at the paper cited for a way to build typical signatures of the temporal behavior</figcaption></figure><p>They then looked a Sune Lehman’s data of ~1000 students to track circadian rythms, using NMF to cluster people in traditional chronotypes (normative, morningness, eveningness). Chronotypes tend to correlate with some Big5 traits. If you look at the size of social network, morning people tend to be less social, social networks are larger during the summer.</p><p>As open question: they looked at signatures over months, weekly and daily patterns, but what about lower granularity such as minutes or seconds? A possible application would be to use these daily patterns to identify bots. Another application would be that those could be effective to spot dangerous phases of bipolar disorders.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*T5cMzJ0yic1ujad_4a9S3w.jpeg" /><figcaption>Some refs</figcaption></figure><p><strong>Tina Eliassi-Rad [Keynote].</strong> In network science, you have a lot of choices but not many guidelines. they developed this axiomatic approach to check if the strength measure has a certain property (including things like isomorphism, frequency, intimacy…). Details here: <a href="https://arxiv.org/pdf/1112.2774.pdf">https://arxiv.org/pdf/1112.2774.pdf</a>. You can do the same for metrics of network similarity <a href="http://www.research.rutgers.edu/~ss2078/papers/NetSim.pdf">http://www.research.rutgers.edu/~ss2078/papers/NetSim.pdf</a>. Overall, good ways to produce axiomatic frameworks.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3d6b10e128b3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/socialdynamics/complex-networks-conference-marseille-3d6b10e128b3">Complex Networks Conference Marseille</a> was originally published in <a href="https://medium.com/socialdynamics">SocialDynamics</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Urban Analytics” event at the Alan Turing Institute (day2)]]></title>
            <link>https://medium.com/socialdynamics/urban-analytics-event-at-the-alan-turing-institute-day2-e0b2c5405504?source=rss-547f52530508------2</link>
            <guid isPermaLink="false">https://medium.com/p/e0b2c5405504</guid>
            <category><![CDATA[smart-cities]]></category>
            <dc:creator><![CDATA[Luca Maria Aiello]]></dc:creator>
            <pubDate>Fri, 08 Jul 2016 11:44:42 GMT</pubDate>
            <atom:updated>2016-07-16T11:14:10.998Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Two short presentations.</strong></p><p><a href="http://www.commsp.ee.ic.ac.uk/~icore/GavinPearson.html">Gavin Pearson</a> from <a href="https://www.gov.uk/government/organisations/defence-science-and-technology-laboratory">Defence Science and Technology Laboratory</a>. Expressed the need to understand how governmental (and military) bodies co-evolve and affect each other. Expressed the need to active learning techniques to take fast, context-aware decisions. The laboratory will announce a new long-term program (5–10 years) on these topics.</p><p><a href="https://www.geog.leeds.ac.uk/people/m.birkin">Mark Birkin</a> from <a href="http://lida.leeds.ac.uk/">Leeds institute for Data Analytics</a> presented the institutions that collaborate within the institute: Consumer Data Research Centre, Medical Research Council, InnovateUK, Virtuosity (VR in urban context), NHS. Involved in H2020 and EPSRC projects. Partnership with several private entities in the consumer market(Sainsbury, Barclays, Telefonica...).</p><p><strong>Group activity.</strong></p><p>Short plenary summary on the research priorities in urban analytics identified during the first day group activity. Emphasis of 1) need to identify use-cases from stakeholders, 2) find effective ways to involve/engage citizens 3) exploration of plausible/implausible scenarios for the future of cities.</p><p>Group activity to write examples of ambitious proposal that can be use to inform the draft of a future ATI call for proposals. A list of titles and project descriptions follows (<em>list of keywords for future proposal draft</em>):</p><p><strong>Group1 (ours).</strong><em> Urban resilience</em> in relation with changes in the city (intentional or unintentional): new technologies deployed (uber), natural disasters (floodings), planned extraordinary events (visit from the pope), planned ordinary events (change a toll on a bridge). How <em>data-driven simulations</em> can be visualized to decision/policy-makers to understand who is impacted and why. Group 3 proposed something similar: Urban resilience in the domain of transport, disasters, perturbations.</p><p><strong>Group2: </strong><em>“Feynman diagrams for urban systems”</em>; <em>“Multiscale systems”</em> combinations of physical, social, spatial, temporal dimensions; <em>“Economic complexity”</em>: network of industries on which a city is projected on, try to understand which co-location of activities drive different outcomes in cities.</p><p><strong>Group4: </strong><em>“Ecology of cities”</em> how climate change or other perturbations affect the ecology of cities, have a digital platform to model uncertainty . Power of centrality vs. power of dispersion/decentralization. <em>“Public infrastructure for urban modeling”</em> tools to support stakeholders to access expertise around urban modeling (including citizen science), how to build better comprehensive data collection tools. A little bit on the lines of: <a href="http://organicity.eu/">http://organicity.eu/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e0b2c5405504" width="1" height="1" alt=""><hr><p><a href="https://medium.com/socialdynamics/urban-analytics-event-at-the-alan-turing-institute-day2-e0b2c5405504">“Urban Analytics” event at the Alan Turing Institute (day2)</a> was originally published in <a href="https://medium.com/socialdynamics">SocialDynamics</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>