<?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 Raffaele Cannizzaro on Medium]]></title>
        <description><![CDATA[Stories by Raffaele Cannizzaro on Medium]]></description>
        <link>https://medium.com/@raffaele.cannizzaro?source=rss-c803167a7bf8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*2wdLaGVDREQKj18la4KN3w.jpeg</url>
            <title>Stories by Raffaele Cannizzaro on Medium</title>
            <link>https://medium.com/@raffaele.cannizzaro?source=rss-c803167a7bf8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 08:58:09 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@raffaele.cannizzaro/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[About Echo Fucking Chambers]]></title>
            <link>https://medium.com/@raffaele.cannizzaro/about-echo-fucking-chambers-9f67bfe2d3cc?source=rss-c803167a7bf8------2</link>
            <guid isPermaLink="false">https://medium.com/p/9f67bfe2d3cc</guid>
            <category><![CDATA[fake-news]]></category>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[politics]]></category>
            <category><![CDATA[donald-trump]]></category>
            <dc:creator><![CDATA[Raffaele Cannizzaro]]></dc:creator>
            <pubDate>Wed, 23 Nov 2016 19:15:41 GMT</pubDate>
            <atom:updated>2016-11-24T07:06:59.316Z</atom:updated>
            <content:encoded><![CDATA[<p>Today I read the full transcript of Trump&#39;s <a href="http://www.nytimes.com/2016/11/23/us/politics/trump-new-york-times-interview-transcript.html?smid=tw-share&amp;_r=0">interview</a> with the New York Times. While I still found some of his positions quite alarming (for example, him still hanging on to a now debunked fake news story about scientists fabricating Global Warming evidence, or the absence of any statement on the conditions of various minorities or women in the US), sections of the interview were instead reassuring. For example, it&#39;s a good thing that Trump is actually backing away from some vicious promises that he made during the presidential campaign, including repealing Obamacare, prosecuting Hillary Clinton, or doubling down on waterboarding. He also disavowed the Nazi incidents.</p><p>Good? Not so fast. I was touched by the news (&quot;maybe it’s not as bad as it looks&quot;), and my immediate thought was to share my findings with my friends on Facebook. But, guess what, I realized that my position would not be aligned with the majority of the shared posts that populate my wall, and could spur controversy. So, I ended up not sharing it. Then I felt guilty. Guilty about not having the courage to share something I though was interesting. Guilty about being scared about other peoples&#39; opinions. What is wrong with me!? So I started analyzing what was going on in my mind. Here is what I think I found. Spoiler: it has to do with Facebook algorithms <a href="https://medium.com/@raffaele.cannizzaro/dear-facebook-now-do-something-21b7bb45734c#.lfntcxeza">again</a> and how they spread polarizing content at the expense of balanced and nuanced opinions.</p><p><strong>Facebook algorithms help create a unilateral Democrat&#39;s narrative. </strong>In my feed, as soon as something happens, there are 2–3 major left-leaning news outlets that shape how the events get reported: usually, the Huffington Post, the Washington Post, and the New York Times. One may think of them as objective news outlets, and they indeed are miles away more objective from their right counterparts, but due to <a href="https://medium.com/@raffaele.cannizzaro/dear-facebook-now-do-something-21b7bb45734c#.lfntcxeza">Facebook algorithms privileging shares and clicks</a>, I end up seeing the subset of articles that is more polarizing, and, given the composition of my friends, have a left-leaning twist. For example, in the case of the Trump speech, I saw a WaPo article titled &quot;<a href="https://www.washingtonpost.com/blogs/plum-line/wp/2016/11/23/trumps-new-interview-with-the-new-york-times-isnt-reassuring-its-deeply-alarming/">Trump’s new interview with the New York Times isn’t reassuring. It’s deeply alarming.</a>&quot; which only focused on the bad, and dismissed the good as &quot;there are glimpses of hopes&quot; in the subtitle. Similarly, the Huffington Post reported <a href="http://www.huffingtonpost.com/entry/donald-trump-conflict-of-interest_us_58349547e4b09b6055ff367d">Donald Trump Basically Says Conflicts Of Interest Aren’t Illegal If The President Has Them</a>, when it&#39;s more complex than that (read the actual excerpt and you&#39;ll find out). I am not saying the news is false, I am not even saying I necessarily disagree with it, but it&#39;s definitely edited to adhere to a certain narrative. This is, by the way, what I have always accused the right of doing, and I still think they do it way more than the left.</p><p>What’s particularly interesting here is that, had these news outlets posted more diverse or nuanced opinions on the Trump interview, these news wouldn’t be as likely to get to my feed because they would not get as many shares/clicks. So, again, it’s not necessarily a problem with the news outlets, certainly not with NYTimes at least, but it’s a problem with how Facebook privileges polarizing news over anything else, in the name of shares/clicks.</p><p>The effect is that the Facebook algorithms end up shaping a democratic narrative that basically says &quot;Trump bad, Obama good&quot; while impoverishing it of nuanced opinions. Same goes on the conservative side where everything bad is about Obama. Even if balanced reporting existed (and it does), it would get surfaced less because it doesn’t get as many clicks/shares. So we don’t see it as much. It really drives me nuts.</p><p><strong>Facebook is such a public place. </strong>I have friends on Facebook, sure. But I also have coworkers, bosses, and bosses of my bosses. Or potential employers. Or potential friends. Others may have potential romantic partners. Guess what, they are all watching. And we know how it’s in human nature to judge fast. One post that’s outside of the narrative, or a post that’s poorly written or unclear, and I can be immediately catalogued negatively. So, what the heck? Why go through this major risk and jeopardize my future employment or relationships? Why wouldn’t I just post baby pictures or fully-aligned opinions and get tons of likes? Win — win!</p><p>Also, as I have already mentioned elsewhere, Facebook doesn’t really make it easy to be &quot;interested&quot; in a topic or figure, without necessarily agreeing with what it has to say. I don’t have any data but I bet $100 that the majority of people that &quot;follow&quot; Hillary Clinton also support her, and the majority of people who &quot;follow&quot; Donal Trump also support him. It takes a special kind of courage to &quot;follow&quot; Donald Trump if you are supporting Hillary (&quot;what if my friends know?&quot;). You can be interested in something without necessarily agreeing with it. But, really, how many people do this on Facebook?</p><p>In summary, I believe that incentives are stacked AGAINST the spread of balanced news and nuanced opinions, from all sides of the equation: the content providers are incentivized to create polarizing content, otherwise they don’t get traffic, and the end users are incentivized to only share and follow aligned content. That’s fucked up.</p><p><strong>So, is Facebook evil? </strong>It may seem, from this post, that I am mad at Facebook and I consider them the worst thing that has ever happened to mankind. Quite the contrary. First, I believe the phenomenons that I described here also exist elsewhere: we are all victims of confirmation bias; traditional news corporations have always exploited such human vulnerability to the max (Fox News, anyone?). Also, TV holds the same power to spread fake news and exploit controversy than Facebook does: all it takes is put a credible scientist who has studied a topic for 30 years and is a reputable authority in the field, and have her debate with some populist pseudo science nutjob. In the eye of the public, both opinions acquire equal weight even if one is garbage, thus leading to falsehoods spreading. Again, it’s a matter of incentives: nutjobs and controversy increase audience, science is boring.</p><p>On the other hand, I think that Facebook has the unique power, the means, and therefore the responsibility to make things better. Changes in the way algorithms work can have profound effects on how news is spread, and we have now the tech capability to improve our lives. Technology is a wonderful thing, let’s put it to good use. Because, let’s face it, algorithms are not neutral, let’s not pretend they are, especially as they are currently engineered and when a for profit organization owns them. Let’s do our best to make them serve us in the most neutral way as possible.</p><p><strong>What can I do? </strong>While Facebook or <a href="https://docs.google.com/document/d/1OPghC4ra6QLhaHhW8QvPJRMKGEXT7KaZtG_7s5-UQrw/edit">somebody else </a>fixes this, here are a couple of suggestions on how to get out of the echo chamber, without hurting your public image.</p><ol><li>Ruthlessly &quot;Follow&quot; people and news outlets you disagree with on Facebook. (Follow is different than &quot;like&quot;) and read the news that will now get to your feed. Brace yourself for some pain. Use <a href="https://www.theguardian.com/media/2016/nov/22/conservative-websites-media-bubble">this article from the Guardian</a> as a start for inspiration.</li><li>Don’t unfriend supporters of the opposing party or block dissenting opinions.</li><li>Use Google News. Download the app and start using it. Again, be interested in a wide variety of topics. Google news is less incentivized by shares than Facebook is and has already made progress against bad content (they’ve had this problem for longer)</li><li>If you have opinions, write them down and spend the time to research what you write about. You&#39;ll find yourself less polarized than you think</li><li>Go on the news website directly. What you get on facebook is just a subset of what gets published. There is more on the NYTimes than anti-trump articles. Or, there is more on the WSJ than anti Hillary articles.</li><li>Of course, read before you share.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9f67bfe2d3cc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dear Facebook, now do something]]></title>
            <link>https://medium.com/@raffaele.cannizzaro/dear-facebook-now-do-something-21b7bb45734c?source=rss-c803167a7bf8------2</link>
            <guid isPermaLink="false">https://medium.com/p/21b7bb45734c</guid>
            <category><![CDATA[donald-trump]]></category>
            <category><![CDATA[journalism]]></category>
            <category><![CDATA[2016-election]]></category>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[elections]]></category>
            <dc:creator><![CDATA[Raffaele Cannizzaro]]></dc:creator>
            <pubDate>Sat, 12 Nov 2016 02:47:43 GMT</pubDate>
            <atom:updated>2016-11-12T20:12:47.049Z</atom:updated>
            <content:encoded><![CDATA[<p>Like all of us, I woke up to the news that Trump was going to be our next president. Columnists everywhere are advancing hypotheses on why this happened, mentioning social, economic, political issues. As I look back to my experience on social media, I can’t help but thinking that Facebook may have influenced the outcome of this election, and not in a good way. [Update: many news outlets have quickly voiced their concerns over Facebook’s influence, including the <a href="http://nymag.com/selectall/2016/11/donald-trump-won-because-of-facebook.html">New York Times</a>,<a href="http://www.latimes.com/business/la-fi-election-media-20161109-story.html"> LA Times</a>. Zuckerberg has promptly <a href="https://www.google.com/url?sa=t&amp;rct=j&amp;q=&amp;esrc=s&amp;source=web&amp;cd=1&amp;cad=rja&amp;uact=8&amp;ved=0ahUKEwjC766_76HQAhXMx4MKHazdCroQqQIIHjAA&amp;url=http%3A%2F%2Fmoney.cnn.com%2F2016%2F11%2F10%2Ftechnology%2Ffacebook-mark-zuckerberg-fake-news%2F&amp;usg=AFQjCNFte10mh9e2eVAGWBEcfMtEoT5mKg&amp;sig2=HEe14XsVKROW5ARyyUJcQg">rejected</a> such accuses].</p><p>According to the <a href="http://www.journalism.org/2016/05/26/news-use-across-social-media-platforms-2016/">latest report from PewResearch</a>, the majority (62%) of US people gets their news from social media. Also, of all Facebook users, about 66% read the news there. This means that about 44% of the US adult population (almost <a href="https://en.wikipedia.org/wiki/Demography_of_the_United_States">100MM people</a>!) get their news from Facebook. Also, about “20% of social media users changed their views on a political or social issue because of something they read on social media”. This means that it’s quite reasonable to think that Facebook may have affected electoral preferences. I won’t venture here into trying to quantify this effect (I don’t have enough data), but I’ll simply try my best to guess how Facebook algorithms work, and discuss what’s wrong about them.</p><h3>How Facebook chooses what to show you</h3><p>Facebook doesn’t let you see every single piece of content that your friends share, or that is popular in the news. This is because it would be highly impractical: imagine your wall being flooded with hundreds of updates every minute, many of them highly irrelevant to you. Instead, Facebook acts as a filter and only shows a subset of the content that’s available, selected based on the output of <strong>machine learning algorithms</strong> (for the <a href="https://www.theguardian.com/technology/2016/may/12/facebook-trending-news-leaked-documents-editor-guidelines">most part</a>). These algorithms decide whether you get to see a pro-Trump or a pro-Hillary piece of news, or whether you get to see more updates related to the Lakers vs the Warriors. Generally, these algorithms serve a useful purpose: they show you what’s relevant to you, and discard what’s not. However, while it’s easy (and fairly inconsequential) to determine that the weather report of a town 1000 miles away isn’t relevant to you, when it comes to complex political, economic, or social issues, determining what’s relevant is much more difficult and ambiguous. To understand why, we first need to have a closer look under the hood of these machine learning algorithms.</p><h4>How do these machine learning algorithms work?</h4><p>In very simple terms, these algorithms are <em>optimizers</em>. They take a bunch of <em>inputs</em> (often called features) and produce an <em>output,</em> which maximizes the chances of triggering an <em>aspired behavior or metric</em>. As an example, let’s look at Amazon’s recommendation engine: if I buy a tennis racket (<em>input</em>), then the machine learning algorithm ensures that I get served suggestions about related items, such as tennis balls, tennis shorts, racket strings etc…(<em>output</em>), because this will increase the chances that I end up purchasing them (<em>aspired behavior/metric</em>). Similarly, on Facebook, algorithms scan every single piece of content that I or my friends shared and liked, and rank it in what they believe to be the order of how likely I am to find each post worthwhile and therefore <strong>trigger an aspired behavior</strong>.</p><p>While we don’t really know how these algorithms work in detail, it’s reasonable to think that they optimize for actions that keep you engaged, such as clicks and shares. This is because, if you share or click, you stay engaged longer, which is great for Facebook: the longer you stay engaged, the more likely you are to click on a piece of advertised content (which pays Facebook money), and, perhaps most importantly, you are sharing behavioral data that can be used to refine the targeting algorithms that advertisers use to reach you, which is the main source of revenue for Facebook. Also, once you share something, there is a higher chance for your friends to see such content and act on it themselves, thus potentially driving more shares and clicks. The cycle never ends.</p><p>As far as <em>inputs</em>, these algorithms probably scan everything you may imagine, and more. Not only what you click, like, and share, but also how long you stare at something on your mobile news feed, what sentiment you have reacted more to, your relationship status, your affiliations, your mobile/desktop usage, your travel schedule. You name it.</p><h3><strong>Why is there a problem then?</strong></h3><p>First, optimizing for clicks is backwards. The fact that you clicked on a piece of news doesn’t imply that you will go on and read it. In fact, it’s likely <a href="https://www.washingtonpost.com/news/the-fix/wp/2014/03/19/americans-read-headlines-and-not-much-else/">you won’t</a>. This is a major problem as it incentivizes the wrong behavior from publishers: publishers have to prioritize catchy titles, controversial subjects, and polarizing matters, because those are the ones with the highest chances of getting clicked on. Clickbait titles like “<a href="https://www.google.com/search?q=you+won%27t+believe+what+hillary&amp;oq=you+won%27t+believe+what+hillary&amp;gs_l=serp.3...13193.15977.0.16127.11.8.0.0.0.0.885.885.6-1.1.0....0...1c.1.64.serp..10.0.0.oHmLqi8OTrk">You won’t believe what Hillary has done</a>”, “<a href="https://www.google.com/search?q=bombshell+hillary&amp;oq=bombshell+hillary&amp;gs_l=serp.3...12148.13766.0.14420.9.8.0.0.0.0.651.651.5-1.1.0....0...1c.1.64.serp..8.1.650...0j0i7i10i30k1j0i7i30k1j0i8i7i30k1j0i8i30k1.yvHuxNvRfbI">Bombshell! Hillary…</a>” have been proliferating. According to <a href="https://www.theguardian.com/technology/2016/aug/24/facebook-clickbait-political-news-sites-us-election-trump">the Guardian</a>, dozens of Macedonia-run, fake news sites gathered millions of clicks and, I assume, money, simply by fabricating great controversial headlines. The fact that a bunch of non-journalists can set up shop and profitably run a site of completely made-up news on Facebook is a clear testament that the incentives are poorly designed.</p><p>Second, optimizing for shares is also backward. 6 out of 10 people <a href="https://www.washingtonpost.com/news/the-intersect/wp/2016/06/16/six-in-10-of-you-will-share-this-link-without-reading-it-according-to-a-new-and-depressing-study/">don’t even click on a piece of content</a> when the title is compelling enough to be shared. If something rings true or confirms our biases, we just share it, <em>without even reading the article. </em>This, again, has ripple effects on the way news are produced and shared that are very similar to what clicks incentives produce. Apparently “<a href="http://www.urbandictionary.com/define.php?term=sharebait">Sharebait</a>” is now a thing.</p><p>Third, shares are, de-facto, a public endorsement. Unless you go out of your way to caveat that you don’t agree with the content you are sharing, you are likely to be sharing content you agree with. I have no data on this, but I think it’s reasonable to assume. This is what usually creates the so-called “bubble”. You share stuff that you agree with, then the algorithms, based on that info, show you more stuff that you agree with, which you’ll share in a recursive loop that ends you up with a full-blue wall, when half of America is red.</p><p>Finally, Facebook makes it incredibly hard to be exposed to arguments that disprove my bias because doing so would publicly alter the public persona that I have built for my self.<strong> </strong>I think that Facebook has become such public place that our actions on it are not genuine anymore. The Facebook profile shows our most visible online identity, which we carefully shape through what we click, share, and comment on. Each action that we take has an immediate consequence on such identity. This results in some actions being a lot harder to do than others: for example, assuming I am a democrat (and likely to have democrat friends) I could still be interested in Donald Trump policies without agreeing with them. Will I share them? Probably not, unless I am prepared to have a discussion with my pro-Hillary friends. Or, I could be wanting to learn the arguments against gun control, without necessarily agreeing with them. Will I share them? Only if I am be prepared to argue with my pro-gun-control friends. Who has the energy of doing that? My guess is, not many (I don’t). The low friction path is just sharing stuff that’s safe and that I know my circle will approve. At the end of the day, Facebook has become a place where the majority of the actions we take imply approval.</p><h3>Now what, Facebook?</h3><p>It’s hard to say whether the above has steered the election in one sense or another. I’ll go out on a limb and say that the algorithms DID help Trump: we know for a fact that the Trump electorate was more likely to respond to click-bait and share-bait articles, because the Macedonian team who fabricated those fake news sites said they experimented with both blue and red sides, and concluded that pro-Trump fake news had the biggest return. Therefore, it is more likely that clickbait articles with partial, incorrect, or malicious information, were spread across Trump supporters and undecided voters more easily than amongst Hillary’s.</p><p>As far as what to do, here are a couple of ideas:</p><ol><li><strong>Start optimizing for “reads”.</strong> Instead of optimizing for clicks and shares, Facebook could design the algorithms to optimize for “reads” or “time spent”. Facebook knows exactly what my eyes are looking at (think about the autoplay function for mobile video), so it should be relatively easy to show more of the stuff that I have actually read, vs the stuff that I simply shared/clicked on without reading.</li><li><strong>Copy Google.</strong> Google’s initial pagerank algorithm had very similar problems. Search was plagued with bad content for years. Over the years, however, Google was able to get it together and provide an excellent service, deprived of spam. Facebook should find a way to continuously invest in prizing quality over shady content, independently of shares/clicks.</li><li><strong>Vet news sites.</strong> As excellently described on this <a href="http://www.businessinsider.com/facebook-blamed-trump-election-mark-zuckerberg-response-tone-deaf-2016-11">Business Insider article</a>, Google evaluates sites based on a number of factors, including “authority”, “journalistic standards”, or “accountability”. Facebook Instant News doesn’t vet its partners at all, aside from requiring partners to adhere to standard community guidelines, which bans things like sexual or violent content)</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=21b7bb45734c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>