<?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 Ramya Sethuraman on Medium]]></title>
        <description><![CDATA[Stories by Ramya Sethuraman on Medium]]></description>
        <link>https://medium.com/@ramyasethu?source=rss-afb860b991e2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*1ldtAK7SeyrBtMdT0nTh1g.jpeg</url>
            <title>Stories by Ramya Sethuraman on Medium</title>
            <link>https://medium.com/@ramyasethu?source=rss-afb860b991e2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 18:27:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ramyasethu/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[Built a GenAI chatbot trained on my blog posts]]></title>
            <link>https://medium.com/@ramyasethu/built-a-genai-chatbot-trained-on-my-blog-posts-290b1eb7938e?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/290b1eb7938e</guid>
            <category><![CDATA[chatbot-development]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[custom-chatbot]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Wed, 07 May 2025 20:18:38 GMT</pubDate>
            <atom:updated>2025-05-07T20:39:45.703Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B8ODEe9PcPDB2lGyiP8WtQ.png" /><figcaption>Custom Chatbot trained on my Medium posts</figcaption></figure><p>It’s 2025, how quickly can you build a GenAI chatbot trained on your content? Turns out, less than an hour! Give or take a few more hours, depending on how much customization you want.</p><ol><li>I asked ChatGPT to teach me how to build a simple chatbot trained on my Medium blog posts. It pointed me to chatbase.co.</li><li>I used a Chrome extension, called Instant Data Scraper, to scrape my Medium blog and exported the results as a CSV. I then opened the CSV in Google Sheets</li><li>I asked ChatGPT for a function to apply in Google Sheets to shorten the URL so it doesn’t have unnecessary parameters at the end of the URL.</li><li>I then manually scanned the URLs to make sure they were indeed my blog posts. Next, I entered these URLs (39 of them related to product) to chatbase.co and trained an AI agent on these links. I also gave it a few other public URLs mentioning my work and my website, <a href="http://www.ramyasethuraman.com/">www.ramyasethuraman.com</a></li><li>Next, I played with a few different models on chatbase.co and felt like GPT 40-mini was performing reasonably well. I also played with the user and system prompts until the chatbot gave me reasonable responses.</li><li>I now had a <a href="https://www.chatbase.co/chatbot-iframe/rync90b88HdUIK6s0MUWY">functioning custom chatbot</a>! (Cost $40/month for chatbase.co)</li><li>Next step was to embed the chatbot into my website and Medium blog. Medium didn’t have a way to showcase the custom chatbot, so I just included it in my Bio. My website, build on Squarespace, wanted me to upgrade to include custom javascript. I didn’t want to do that, so chose the embed option to embed code and viola, had my custom chatbot on my website as well. You can try it out here: <a href="http://www.ramyasethuraman.com/">www.ramyasethuraman.com</a>!</li><li>You can play with the chatbot <a href="https://www.chatbase.co/chatbot-iframe/rync90b88HdUIK6s0MUWY">here</a> and on my website: <a href="https://www.ramyasethuraman.com/aichatbot">https://www.ramyasethuraman.com/aichatbot</a>!</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=290b1eb7938e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How I’m s̶t̶r̶u̶g̶g̶l̶i̶n̶g̶ ̶w̶i̶t̶h̶ using AI to create a comic book on product management]]></title>
            <link>https://medium.com/@ramyasethu/since-my-last-creative-project-a-few-years-ago-ive-been-noodling-on-various-ideas-to-work-on-as-1ad4a408f539?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/1ad4a408f539</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[comics]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Sun, 27 Apr 2025 06:29:37 GMT</pubDate>
            <atom:updated>2025-05-06T02:49:40.827Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sDisDemUlmYQJT-zLaWXCg.png" /><figcaption>How I’m struggling with using AI to create a comic book on product management</figcaption></figure><p>Since my last creative project a few years ago, I’ve been noodling on various ideas to work on, as a side creative project. My previous creative projects have been <a href="https://amazon.com/author/ramya">writing books</a> on various topics (short stories in English, narrative on Indian immigrant dilemma living in the US, children’s books with cute illustrations, a <a href="https://www.youtube.com/watch?v=AgAz-BRN6x8">short film</a> based on one of my stories). This time, I wanted to leverage AI to help me with the project and recently, the idea has begun taking shape in a concrete way. I thought I’d share my struggles and lessons along the way so you don’t get discouraged seeing how others are using AI in a day to build a viral app, 1000x themselves and save the world all at once!</p><h3>The inspiration</h3><p>For a few years now, I’ve been noodling on how I tie my professional and personal interests closer together to do something creative. I felt like I had a unique background for a few reasons:</p><ol><li>I switched from engineering to product management a decade ago and felt like the transition was a struggle (well worth it though) and wanted to share some lessons along the way</li><li>I felt like I could share my experience working at Meta for 11+ years and how to navigate big tech, starting out fresh as a bright-eyed, naive IC product manager</li><li>Finally, I transitioned a year back from <a href="https://medium.com/@ramyasethu/from-product-meta-to-cpo-the-suite-e4188e2a7e3a">Meta to The Suite, a startup</a> focused on building a white-label community for C-Suite executives and have been sharing advice with those exploring a similar transition.</li></ol><p>I knew I wanted to write a book about product, incorporating some of my content from Medium but also trying something visual. I’m reasonably good at sketching and designing characters for stories but not that great with Procreate and other design tools. I‘ve always been deeply inspired by Scott Adams and his magic with cubicle humor &amp; Dilbert and, more recently, <a href="https://workchronicles.com/">workchronicles</a>. I also wanted to write a playbook of sorts detailing what I learned working with ChatGPT on this project. Putting all of these ideas together, I started playing with Studio Ghibli-inspired image generation using ChatGPT, GPT4o. Perplexity wasn’t generating anywhere close to the visuals I wanted and so, I switched to ChatGPT Plus.</p><p>Enter Reya, a product manager figuring out the ropes of how to work with people and build products. From the beginning, I knew Reya would not be someone who has already “made it”. She had to be someone who was actively making mistakes, learning and figuring it out the hard way. My hope was we would all find some of ourselves in Reya and her struggles. Maybe she would remind us of a younger “us”, bright-eyed and hopeful and insufferably optimistic or perhaps she is you today, reeling from a recent failure and wondering what to do. Now, I had the core idea, I started playing with ChatGPT image generation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/944/1*Dm6032HNfN6BLzmUdks3JA.png" /><figcaption>Reya, PM</figcaption></figure><h3>The journey: prompting, hallucinations &amp; more</h3><p><strong>tl;dr — </strong>1/<strong> </strong>prompting is time-consuming, 2/ ChatGPT always tries its best to be helpful though is often misguided, forgetful and makes you go, “What were you thinking really?” and 3/ Sometimes, I wonder if life would have been easier if I had just hired someone from Upwork and given them clear specs for the comics (like I did for my previous children’s books). But, I learned a ton about the quirks of prompting, what tactics are likely to get a better response and, when the AI is just not going to get me over the finish line i.e. time to invoke the human in the loop!</p><p><strong>The messy details:</strong></p><p>I’ll probably write a detailed post once I get the book out, on everything I learned about prompting and ChatGPT-generated comics. For now, in no particular order, here are some observations on creating comics using ChatGPT:</p><ol><li><strong>Sticking to a consistent character sketch is a pain. </strong>ChatGPT often forgets what the character looks like. You can see me trying to update its memory of a character sketch we had previously generated, Tom in this case:</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sOgS0p8N8XUCkAqGBy0UHw.png" /></figure><p>Here is an example where it forgot what Reya and other characters looked like and I’m reminding it:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HzWvuxG6_7hu1OoZO5-2hw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lQju3CVjWg6Kz_Tw3_dfCA.png" /></figure><p>You can see how it forgot what Reya looked like below (and made up random text for the user quotes!):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/392/1*8wjekj78Vp13FS9KqKzcjw.png" /></figure><p>Sometimes, Reya became a kid 🤷🏼‍♀️ and also notice the text being cut-off; that was a constant struggle too. I had to ask my daughter to edit the image in Procreate to recreate the text bubbles!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*v-RqdYXvYy1pWgDMLFOipw.png" /></figure><p>Sometimes in the same comic strip, the same saved character would be generated differently:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CYEiWZQCsbKnN7uHJ68FGA.png" /></figure><p>2. <strong>ChatGPT is not great with editing intricate details in images</strong>, things like font sizes. Here’s an example:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7wZX9LEI0np5kKpbCdQGFg.png" /></figure><p>It gave me the image below saying it had adjusted the font sizes. It immediately agreed and understood the problems with the image when I told it the image was very wrong 🙂</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PvCvRZ6MkuZAwLWFXSCdSA.png" /></figure><p>3. I couldn’t quite understand some of <strong>ChatGPT’s quirks</strong>. Every now and then, it would generate panels like this, no speech bubbles, borders cut off, like it just got exhausted with all the comic creation work 😁!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1016/1*iZgJpVApwIXBR6PRlzeDPQ.png" /></figure><p>It’s relentlessly <strong>optimistic though </strong>and always tried its best:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nI50iwAP9wnfNJ8Mtxdv3g.png" /></figure><p>I mean, it really really did try its best even when I was losing my patience a little 😠</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gd7VplKGKXecwc8JR2UAAg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Hh2SpHXwQGLhd6usw4Ok0g.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ESw1JJteWfP-MolO4T-klg.png" /></figure><p>Some days, it was just time to quit and start fresh the next day!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mFlwKY5BfNhLJ4pIcwxbtA.png" /></figure><p>4. Sometimes, <strong>ChatGPT completely impressed me with its ingenuity</strong>. Here is an example of when it knew we were not getting anywhere and suggested a different approach to solve the problem, which ended up working really well 💡❤️</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1ebR3QPzNxWd9KGZJT4lYw.png" /></figure><p>And then it would <strong>immediately forget it had suggested a smarter approach</strong>!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*G7UTvswgB-Sm9UQqPXTxZA.png" /></figure><p>Also, it referred to itself in the third person as if distancing itself from the problem 😄</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ErdWxUX-MYuqVk_WH1Ju_Q.png" /></figure><p>There were so many more of these back-and-forth conversations! There’s probably enough content here for an entire series on these learnings! If you found this interesting, please leave a comment, and I’ll post more about my journey creating comics with ChatGPT. For now, back to Reya’s world of product! Thanks for reading 📓!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1ad4a408f539" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Product @ Meta to CPO @ The Suite]]></title>
            <link>https://medium.com/@ramyasethu/from-product-meta-to-cpo-the-suite-e4188e2a7e3a?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/e4188e2a7e3a</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[faang]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Fri, 22 Nov 2024 00:54:19 GMT</pubDate>
            <atom:updated>2024-11-22T00:54:19.547Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OIOzuxT5-4qJTbhhuyJ-cw.png" /></figure><p>A question I often get asked is what my experience has been like having worked as an engineer and product lead at Meta for 11 years and then transitioning to being the CPO at a startup. <a href="https://www.thesuite.com/">The Suite</a> is a startup focused on building a community platform for C Suite executives. A huge draw for me was the mission of the company: to build a community and platform for executives to connect with their most valuable peers and relevant content via conferences, events, and, of course, the product platform. The transition allowed me to draw on my deep experience with News Feed personalization, Facebook groups, and new user onboarding to Facebook to apply it to my new role and hit the ground running. But, I also see several differences and thought I’d pen down my early observations before I hit my 6-month mark:</p><p>My time at Meta taught me to think critically about scale, systems, organizational dynamics, processes and of course the <a href="https://medium.com/@ramyasethu/product-management-frameworks-3p-the-art-science-of-product-72a66faafb42">Art &amp; Science of product</a>. At <a href="https://www.thesuite.com/">The Suite</a>, I’ve leveraged these skills but also learned to go broad, think critically about what makes a startup succeed, and explore a rich, diverse set of new skills. So, here are my three observations going from Big Tech → Startup world:</p><ol><li><strong>Building at scale vs building for each user touch point</strong>: My work at Meta impacted millions of users, whether it was thinking through the <a href="https://about.fb.com/news/2019/05/more-personalized-experiences/">ranking system</a> for <a href="https://about.fb.com/news/2019/03/why-am-i-seeing-this/">News Feed</a> or the governance platform for all Facebook groups. At The Suite, I have the privilege of engaging directly with C-suite executives who use our product and I’ve learned that each touch point with our members is fairly high stakes and needs to provide value immediately. It’s critically important for us to get our members in the ‘right room at the right time’, with the right set of their peers. On the member platform, this means figuring out the right product strategy and features to allow intimacy at scale: whether this is group chats to enable members to have a private conversation with a small group of their peers, ensuring email notifications convey the most relevant information to members or figuring out a strategy for building a personalized, ML-based platform for our members. We don’t have the luxury of billions of behavioral data and millions of members, so we need to make every interaction count.</li><li><strong>Balancing pace and quality</strong>: Working in a startup environment is a constant dance between moving fast while shipping quality products to optimize for each user touch point (#1 above). This is true of product work at all companies (balancing pace and quality) but is even more critical when it comes to startups. Do we wait a few weeks to add more bells and whistles to the product or launch now so we can quickly iterate if the feature doesn’t land? The classic product question but one that comes into hyper focus at startups. We recently launched the ability for our members to <a href="https://www.linkedin.com/feed/update/urn:li:activity:7222311838867234816/">post polls</a> and <a href="https://www.linkedin.com/posts/thelsuite_thelsuite-braintrust-featurelaunch-activity-7203767209523879936-nLWY/">chat with each other privately 1–1</a> and often discussed whether we are comfortable with the MVP version of this product or did we want to delay launch to get to an MLP (Minimum lovable product)? For example, we could have built intelligent ways for our members to slice and dice the data from our polls but adding this feature means we take longer to go to market and collect data on whether the feature resonates.</li><li><strong>Time to market</strong>: At Meta, I had the luxury to go deep and spend months conceiving and launching a product. The very first time we launched a <a href="https://about.fb.com/news/2019/03/why-am-i-seeing-this/">Facebook product focused on News Feed algorithmic transparency</a>, it took us more than a year from conception, UXR testing, cross-functional alignment to building and launching. The time frame to launch is much shorter in startups. One of my favorite product initiatives I worked on at Meta was figuring out a <a href="https://medium.com/@ramyasethu/supporting-human-values-and-human-flourishing-c1553de1a0ce">framework to align our AI systems with human values</a>. This was before the time of LLMs and this felt like ground-breaking, cutting-edge technology to me although fairly academic and I loved digging deeper into this technical space (think days spent reading <a href="https://intelligence.org/files/CEV.pdf">academic papers</a>) to figure out how it might shape Facebook News Feed ranking. We don’t have this luxury at smaller companies. It’s more about prioritizing really well to take our best bets to market rapidly and then iterating or pivoting until the product feature has amazing organic traction with our members.</li></ol><p>I have more observations around going deep vs broad and the diversity of experiences at a startup vs larger companies, process-driven vs scrappiness, and several other dimensions, but this is already a much longer post that I had planned, so thank you for your time ⏰ and message me for a virtual coffee chat if you’d like to talk about product, big tech or learn more about what we do at The Suite!</p><p>For today, I’m truly grateful for the chance to be <a href="https://www.businesswire.com/news/home/20240927515993/en/The-Suite-Announces-Hire-of-Ramya-Sethuraman-as-First-Chief-Product-Officer">part of a company</a> committed to building a trusted community for executives and I can’t wait to see where this journey takes us 🙏🩷.</p><p>Onward and upward!</p><p><em>Original post with comments on LinkedIn </em><a href="https://www.linkedin.com/pulse/from-product-meta-cpo-suite-handful-learnings-ramya-sethuraman-bmudc/"><em>here</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e4188e2a7e3a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Look for the helpers]]></title>
            <link>https://medium.com/@ramyasethu/look-for-the-helpers-171343af6e71?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/171343af6e71</guid>
            <category><![CDATA[autobiographical]]></category>
            <category><![CDATA[humanity]]></category>
            <category><![CDATA[uplifting]]></category>
            <category><![CDATA[immigrant-stories]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Sat, 24 Aug 2024 18:44:05 GMT</pubDate>
            <atom:updated>2024-08-24T18:50:20.011Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*aHvl2bKnVLNAieUy.jpg" /></figure><p>Life has this tendency about it. A tendency to be relentless and untenable. A tendency that wears you down over time. We all have felt it and survived it. Through all of it though, life also has this magical ability to keep producing its share of helpers. Stare hard enough at whatever challenging situation you’ve been through and you will recall the people who helped, their <a href="https://www.forbes.com/quotes/10270/">little nameless acts of kindness</a> that made a particularly hard day a little easier to get through. This post is about those helpers in my life, right from the moment I landed in this country, 24 years ago.</p><h3>1. “Dirty, old luggage”</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UZr9t7j9tbBNI2lS1fnvzw.png" /><figcaption><em>“You got a dirty, old luggage from India, it broke and now I suppose you want to sue us?”</em></figcaption></figure><p>“<em>You got a dirty, old luggage from India, it broke and now I suppose you want to sue us?” </em>the stern woman behind the Delta Airlines counter at the Atlanta International Airport glowered at me. My <a href="https://www.amazon.com/r2i-dreams-Here-Parth-Pandya/dp/1500705802">introduction to America</a> was clearly not going too well 🧳 😔.</p><p>I didn’t think my suitcase was dirty or old and, I definitely didn’t want to sue the airport. I was overwhelmed as it is, having completed my first 20 hr+ international flight and trying to figure out how to navigate the terminals and gates and inter-airport travel at the Atlanta International Airport. I don’t remember the gentleman’s name but he seemed to materialize out of nowhere with a large adhesive tape and smiled reassuringly as I was busy making myself as small as possible. Someone had run a razor through my suitcase and it had a gaping hole on its side. He helped tape it together, showed me a thumbs-up sign, and left. His kindness made my welcome to America much better than it otherwise would have been.</p><h3>2. “Could you at least inform me before class starts?”</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zfZXcVgNn1zDGuZhh4HrLg.png" /><figcaption>Image from ebay</figcaption></figure><p>So, it turns out snow is only exciting when you’ve never seen it and are just reading about it in Archie comics ❄️ ☃! I remember devouring those comics as a teenager in India and wondering when I’d have a chance to experience snow in my life! The very first day it snowed after I landed in Lexington, KY, I spent the whole day playing outside in the snow with my friends and then quickly after that day, my fascination with snow died down.</p><p>A few weeks after my enchantment with the first snowfall, I woke up late on a snowy morning, hurriedly dressed up and started plodding through the snow, having missed the University bus that would take me to campus. After five minutes in the snow, realizing I was already ten minutes late to a 45 minute Statistics class that I was teaching (!), I headed back home, emailed my professor that I couldn’t make it to class and went back to bed. I was the Teaching Assistant for an Undergraduate class in Statistics. I received an immediate reply from my Professor, <em>“Next time, could you at least inform me before class starts?”</em> I was mortified and thought I’d lost my job as a TA. The next day, I met him in person to apologize and given how diligent I usually am, he smiled and forgave my mistake. Apparently, the students all came to class early in the morning, waited around for fifteen minutes and then left. I went on to successfully complete my teaching assistantship and am so grateful, Prof. Molzon gave me another chance.</p><h3>3. “You can stay with us.”</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gnPdjZafgKQ-DvkeZ-VzBw.png" /><figcaption>Midjourney generated.</figcaption></figure><p>It so transpired that I became apartment-less as a Master&#39;s student at the University of Kentucky 🏡😥. My room-mates in student housing moved on to other apartments over the summer and I couldn’t afford a single bedroom apartment by myself. So, I turned to my seniors and friends for help and they either were already with roommates or just said no. It’s a pretty miserable feeling to be without a home as a <a href="https://en.wikipedia.org/wiki/Fresh_off_the_boat#:~:text=The%20phrase%20fresh%20off%20the,their%20ethnic%20ideas%20and%20practices.">FOB</a> so far away from “home”. Just as I started feeling desperate, one of my friends who lived off-campus with her brother and sister-in-law offered help. <em>“You can stay with us, don’t worry about it.”</em> So, I packed my life in a suitcase and went to stay with her. Her act of gratitude warms my heart to this day. I stayed at their place for two weeks, living out of a suitcase until I found a roommate with an apartment to share!</p><h3>4. “I’ll make sure you don’t graduate this semester.”</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/432/1*AT9wBsxqk2w_HE-Em6nKFA.gif" /><figcaption>Umbridge (From Harry Potter books)</figcaption></figure><p>I had heard rumors about this particular staff member on the Computer Science admin team, that she (let’s call her Ms. W) was tough, especially on international students, and had left more than a few in tears. I had no fears though, I’d always been a good student, academically strong and hard-working, and had rarely landed in trouble with teachers or staff (save the missed Statistics class 😬). I had joined for a double major in Mathematics and Computer Science and with permission from my CS director, had dropped the Masters in Math to focus on CS. Come time to graduate, I stepped into the admin office to submit papers and confirm that I will be graduating that fall and Ms. W glanced at my paper work, looked up calmly and said, “You don’t qualify because you don’t have the required Computer Science courses.”</p><p>I was shocked. I had already accepted a job offer which was contingent on my graduation. I tried explaining to Ms. W that I had permission from both the Math and CS departments and she would hear none of it. I walked out dazed, racked my brain to solve this problem, and then walked to the Math department, asked my Math professor if he had access to the email he had sent confirming that I could transfer without losing credits to the CS department. I did not have access to that email since it was a few years ago and the mail had been archived. My Math professor, seeing how distraught I was, searched for the right archived file on his computer, unarchived it, and found the email I was looking for. So then, I headed back to Ms. W in the CS department to show her a printout of the email. She took one look at it, waved her hand towards the door, and said, <em>“I’ll make sure you don’t graduate this semester.”</em></p><p>I felt numb, I walked out again and having run out of options, used the university hallway phone to call my Computer Science Thesis advisor. I tried to explain the situation and then overwhelmed, broke down in tears 😭. <em>“Stay there”</em>, that’s all he said on the phone. In ten minutes, he walked up the stairs, listened to the rest of my story and quietly walked into Ms. W’s office, gesturing for me to stay outside. I could hear raised voices inside the room and then he walked out and said, <em>“You can graduate this semester”</em>. I walked into her office tentatively. She didn’t make eye contact, roughly handed a paper that confirmed I would be graduating with a Masters in Computer Science. I took the paper and ran out as quickly as I could. Thank you, Dr. Finkel 🙏.</p><p>There are so many more of these moments where I found the people who helped and so will you. Look for the helpers ❤️.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=171343af6e71" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Product Management Frameworks 3P: The Art & Science of ‘Product’]]></title>
            <link>https://medium.com/@ramyasethu/product-management-frameworks-3p-the-art-science-of-product-72a66faafb42?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/72a66faafb42</guid>
            <category><![CDATA[product]]></category>
            <category><![CDATA[product-management]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Sun, 24 Dec 2023 06:49:09 GMT</pubDate>
            <atom:updated>2023-12-24T21:38:59.977Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KI1F9ltvQYBzOAGjO0lVsA.jpeg" /></figure><p>This post is part 3 of my 3P: Product Management Framework series. You can find the first post on People <a href="https://medium.com/@ramyasethu/product-management-frameworks-563969cdfed5">here</a> and the post on Process <a href="https://medium.com/@ramyasethu/product-management-framework-process-9a19a7880834">here</a>. This final post is about the Art &amp; Science of the final P, ‘Product’. I’ve captured <em>some</em> of the important aspects of Product here and mainly through the lens of what I’d have appreciated knowing when I started my journey in product back in 2016. There are probably several other ways to slice and dice this subject and go deeper, I’ve picked my homebrewed <em>3W1H</em> method<em>,</em> to unpack the 3rd P: Product — <em>the What, Why, Who and How.</em></p><h3><strong>What are we doing? Why does it matter?</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u3C8-qvP4zF7UzcSuQMXxA.jpeg" /></figure><p>I’ve picked key ‘<em>What’</em> concepts to highlight here and clarified a few areas that tend to throw teams off.</p><p><strong>Vision: </strong><em>‘Imagine a world where…’ </em>— That’s your vision statement. Your vivid description of a long-term, future state where your product is wildly successful for the users you were targeting in a market or portion of the market that you’ve clearly captured and why it matters. It’s the glorious end state, the big win! Here’s one of my favorite notes, <a href="https://medium.com/@johnzimmer/the-third-transportation-revolution-27860f05fa91">Lyft’s vision over the next 10 years</a>. Although not all the predictions in this fairly long note came true or are likely to come true by 2025, I do love the depth, the data, the ambitious vision and the emotional appeal of the article!</p><p><strong>Strategy: ‘</strong><em>That’s great, I love this world. How do I get there?’</em> What resources should I allocate and when? What should I not allocate resources to?</p><blockquote>What is the journey to that vision I aspire to?</blockquote><p>How are we leveraging the product’s unique value prop for the identified target users, given the competitive landscape of the product? The answers to these questions come together in the product strategy. This is where we make the hard decisions of what we will do and, more importantly, won’t do.</p><p>I’ll take an example I’m familiar with to describe some of the nuances of product strategy. In 2019, I was in charge of incorporating sentiment in ranking to make Facebook News Feed more personal. There are several ways you can go about this. You can gather people’s sentiment in research sessions to understand and improve how people feel about News Feed, you can use human raters to rate content and use that as a basis to improve sentiment, you can try to understand brand sentiment and improve it, you can leverage external panels of experts to understand and improve sentiment, you can use existing engagement ranking signals and figure out which ones are good proxies for sentiment, you can use surveys and ask people questions to understand sentiment and then improve it, the list goes on!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SpS0HwfiZs0pyz6rUU6fJQ.jpeg" /></figure><p>In this scenario, there are so many one-off decisions to make, unless we have a cohesive product strategy for the problem area, we would be shooting in the dark. Based on deep UXR to define various sentiment constructs and data from the initial surveys we ran, we developed a long-term vision and strategy for how we think about sentiment when it comes to ranking. Based on the strategy, we were able to align on a first set of hypotheses and experiments to begin testing our strategy. We narrowed down our experiment to focus on link content on Facebook, targeting users who regularly consume link content on the platform and we decided to run surveys to understand whether people think the link content they see is worth their time. We then incorporated the survey data to predict whether our users would think a particular piece of content was worth their time and <a href="https://about.fb.com/news/2019/05/more-personalized-experiences/">thoughtfully scaled the approach</a>. Each decision was backed by clear data or reasoning and hypotheses so we learn irrespective of whether the test moved the metric or not. Note that this was not just one experiment that launched surveys and trained a model but a thoughtful strategy to figure out how we might incorporate sentiment to personalize News Feed better and ways to learn from the initial approach and scale. Having a holistic strategy here helps us understand the various paths and tradeoffs to the long-term vision and we can then pick a path and start learning!</p><p><strong>User problems: </strong>Whether your company prefers to call these pain points, user journey to identify problems or JTBD, the core idea is the same.</p><blockquote>Have you identified an unmet need or problem that a user has and why that’s meaningful to solve? That’s a user problem.</blockquote><p>If you are developing a product with the main goal of helping the company make more money, for example, that’s a &lt;company&gt; problem (e.g. Facebook/Meta problem vs User problem). If you have a &lt;company&gt; problem, you can then break it down to uncover user problems that help you solve the &lt;company&gt; problem. The key idea here is to be very clear-eyed about what problem you are solving.</p><p><strong>Problem space and Solution space: </strong>I’ve sometimes seen teams spin a bit while brainstorming or discussing a user problem because of confusion between the problem space and solution space. There’s a time for each. Problems arise when we think we are brainstorming a problem space to play in and drawing the boundaries, and in the midst of it, start digressing into the solution space and constraints. So, the structured approach here should be:</p><ul><li>Be clear about whether the group is discussing the problem space and define the boundaries of the problem space</li><li><strong>First, go broad</strong> and lay out the overall problem space and the specific user problems, then switch to the solution space and draw the boundaries around the solution space, and <strong>finally,</strong> <strong>go deeper</strong> into specific solutions.</li></ul><h3>Who is doing what?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AUjNW2D5NjqEAy1aMPmm0g.jpeg" /></figure><p>It actually doesn’t matter who is doing what as long as everyone is aligned on the who and the what 🙃. See the sidenote below where I call out how different parts of the product vision and strategy could be developed by different functions; it doesn’t have to be a maverick PM who develops a grand, bold vision generated in her head! What’s important is that the final product vision and plan are developed thoughtfully, and leadership and the team are aligned on the plan. I think of the PM function as the glue that holds the team together and accountable for chugging relentlessly toward the goal. What matters is that you make progress together, not who does what.</p><p><strong>✍🏼 A side note on the importance of clear documentation</strong>: A canonical document that lays out the vision, strategy, problem space, solution levers, and execution details is pretty important to ensure everyone is rowing in the same direction.</p><ul><li>The product roadmap usually reflects most of these pieces (vision, strategy, etc.)</li><li>Sometimes, a function (like design) may be in charge of laying out a vision for a major product redesign that might then be used as an instrument to align and inspire people toward that vision.</li><li>Sometimes, UX Research might conduct foundational research that clarifies the problem space and where we should play and capture that in a document.</li><li>I’ve seen several ways to effectively craft and leverage these documents. The important principle here is to get these written down in a canonical document that people can access and ensure the team is aligned with the vision and plan.</li></ul><h3>The How</h3><p>I’ve covered a good chunk of <em>the How, </em>especially org-specific dynamics, in <a href="https://medium.com/@ramyasethu/product-management-framework-process-9a19a7880834">this post on Process</a>. What I call out below are a set of product-specific suspects that tend to derail the team if not handled thoughtfully.</p><p><strong>I. Being in the Weeds</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JAxDbUn1yhxTxqlTHvqXLw.jpeg" /></figure><p>The leaders in the org set the tone for the culture of the org. I’ve worked in organizations where the leaders operate at a much higher level and did not have the bandwidth or preferred not to be in the weeds. I’ve often been the pillar lead ‘in the details’. Caring about the details is important, no matter where you sit in the org. Sometimes, this gets conflated with micro-management or ‘being in the weeds’. Learning the fine balancing art of caring about the details and nuances, while not frustrating your team is something every leader should learn to navigate their own way, through trial and error. Sometimes, the product is of an extremely sensitive nature where the weeds are literally the sauce (definitely need a better metaphor here but you get my point!) and sometimes, the team is kicking ass as they are and we need to gracefully step out of the kitchen until it’s time to check back in again! 👩🏼‍🍳🥘🍳</p><p><strong>II. Try and then try again after a while to (maybe) succeed</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VQ2X04JxuMPqHIhYbbb1-w.jpeg" /></figure><p>Over my time at Meta, I often used to hear people push back on a product idea because <em>‘We tried that and it didn’t work because…’ </em>I’ve sometimes been the wise old product person explaining to someone new why we should steer clear of a particular product idea because it’s been tried before. I then witnessed idea after idea being tried out again by newer folks who didn’t have the history, didn’t know about previous failures, and went on to launch successful products based on the very same idea! Timing matters a lot in product. The takeaway here is to seek out and keep oneself informed of the history and learnings from the past but to not let it constrain you from trying it out again, with the benefit of the learnings from before! While several examples come to mind, a notable product launch that was debated, canned, and then considered and launched successfully at Meta was <a href="https://about.fb.com/news/2022/07/home-and-feeds-on-facebook/">The Most Recent feed</a>.</p><p><strong>III. The one with too many cooks and spoilt broth</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wl86QekPQ1NJFpNQn2ad3w.jpeg" /></figure><p>The ‘too many cooks’ problem usually manifests when it is not abundantly clear who owns what part of the process and who owns the final decision and the steps leading up to it. A typical example is when teams spin on a complicated product decision unable to align while not recognizing that the ultimate decision maker is the VP of the org. All the team is responsible for is to deeply investigate the problem, come up with options for solutions, lay out pros and cons for each option and <a href="https://naomi.com/the-traffic-light-approach-to-problem-solving-7b3d6e42acc2">present the recommended option</a> to leadership for the final decision. Clarifying who the decision maker is and what the team’s role is earlier on in the process keeps egos at bay and keeps the discussion productive and forward-moving. For extremely complex projects, a more formal process such as a <a href="https://www.forbes.com/advisor/business/raci-chart/">RACI matrix</a> or a <a href="https://naomi.com/people-and-process-d5a72341a5be">People &amp; Process</a> approach may be needed.</p><p><strong>Reopening a can of worms and when it’s worth it</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FiGpiGPWd0i0aYtv4PzPYA.jpeg" /></figure><p>A complicated facet of product is figuring out when it makes sense to go back and reopen a can of worms! A typical pattern I’ve observed is when the definition of a ‘thing’ is not clear or not aligned upon explicitly, but the team has already begun to execute on a solution. For example, consider a really complex problem area such as figuring out <a href="https://about.fb.com/news/2021/02/reducing-political-content-in-news-feed/">how to rank political content</a> on the Facebook platform. If the definition of what political content is is not clear or the team and leadership is not aligned on the definition, there will be a lot of lower-level ‘execution issues’ with the product. A PM’s instinct might be to plow through and push on the execution to move fast but sometimes, it might save us a lot of time and effort if we take a step back, reopen the scary can of worms and ask how we are defining aspects of the problem. The flip side to this argument is sometimes, to avoid continuously spinning on a problem, a PM needs to be firm and clarify that the phase where we question definitions and clarify the problem space has already been completed and it’s now time to disagree and commit. So, when do you reopen the mucky can of worms and when do you firmly keep it closed? That’s equal parts art and science!</p><p><strong>What is not measured cannot be improved. Sometimes what is measured cannot be improved</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H1GazkZcggvORvE-aF6W4Q.jpeg" /></figure><p>Silicon Valley big tech is famous for its metric-focused culture. Often, being able to measure something or even proxies to the actual metric is hugely beneficial to track progress and establish a culture where data helps inform product decisions rather than subjective opinions and based on the <a href="https://www.reddit.com/r/ProductManagement/comments/wdd5p7/what_is_the_hippo_effect/">HiPPO effect</a>. Being data-driven is a great best practice. However, if this approach is taken to too much of an extreme, it has the unintentional consequence of thwarting innovative product ideas. In my experience, I’ve seen the data-driven approach be super effective when you are making <strong>incremental improvements</strong> to product and when there are established, time-tested metrics to measure progress.</p><p>When dealing with <strong>new products, </strong>expecting to see the same set of movements in already established metrics can be detrimental. For new product innovations (say when the Facebook app launched Marketplace tab or when Dating was introduced), expecting to see the same app top line metrics move at the same rate is not realistic and we may end up killing an otherwise valuable product. One way to preserve innovation could be to negotiate a longer runway to show topline metric impact for the new product area or to show proxy metrics that indicate the product is moving metrics in the right direction and eventually the magnitude will follow.</p><p><strong>The Goldilocks principle of looping in XFN</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S0cUl5Mb--D4ZTtVfACJkw.jpeg" /></figure><p>Diverse teams where the opinions and expertise of subject matter experts and cross-functional partners (UXR, DS, Content, Design, Legal, Policy, Privacy, PMM…) is included always lead to better products for users.</p><p>There is a right way to loop in cross-functional (XFN) partners in the process. I’ve seen this fail in two common ways. One is when everyone is in the room all the time, weighing in with opinions and the product seems to move at a weighted, slow pace. This happens when the PM is not doing a great job of looping in the right partners at the right time. Looping in your half a dozen IC designers to weigh in on detailed product UX mocks when the basic idea is not validated or has directional approval from leadership is an example of ‘<strong>too early</strong>’. Looping in your XFN partners after the product feature has been halfway defined (with just the PM and the engineer in the room) and then asking XFN partners to fill in the details is an example of looping in your XFN partners ‘<strong>too late</strong>’. Another example of ‘too late’ is asking your legal team for last-minute risks and mitigations for a product that has been completely defined and engineering is halfway through execution!</p><p>As with all the other product issues I discuss in this note, a thoughtful PM who has established a solid rapport with her team and XFN partners will intuitively know the Goldilocks rule of looping in XFN. As with all other things in product, it might seem natural and easy but has been carefully learned and crafted over years of trial and error and hard work!</p><p><em>This concludes my 3 posts on the 3Ps of Product Management: </em><a href="https://medium.com/@ramyasethu/product-management-frameworks-563969cdfed5"><em>People</em></a><em>, </em><a href="https://medium.com/@ramyasethu/product-management-framework-process-9a19a7880834"><em>Process</em></a><em> &amp; Product. I hope you enjoyed it and let me know if you have other ideas to add to the mix!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=72a66faafb42" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Product Management Framework, the 3Ps: Process]]></title>
            <link>https://medium.com/@ramyasethu/product-management-framework-process-9a19a7880834?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/9a19a7880834</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-management-tool]]></category>
            <category><![CDATA[product-organizations]]></category>
            <category><![CDATA[process-improvement]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Wed, 06 Dec 2023 07:03:24 GMT</pubDate>
            <atom:updated>2025-05-06T22:19:45.970Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EK1qSqdVNYyOnqt4ezSVyA.jpeg" /><figcaption>Generated by AI</figcaption></figure><p><em>This post is part 2 of my product management framework series. You can find </em><a href="https://medium.com/@ramyasethu/product-management-frameworks-563969cdfed5"><em>Part 1 on People here</em></a><em>.</em></p><p>‘Process’ is a loaded word. I’ve worked with people (*cough* engineers) who think Process keeps them from doing the real work and I’ve collaborated with senior leaders who have insisted on more Process than was required for a large project. Finding the right balance for Process while keeping stakeholders aligned and chugging along a project can be challenging.</p><p>Sometimes spending upfront time on a clear process saves us several weeks of disagreements, lost time, and effort. Here are some common problems I’ve come across that usually indicate a new process or process update is needed. Broadly, these problems can be categorized into 3 buckets that<strong> </strong>impede<strong>: Execution | Clarity | Alignment </strong>(<em>I’ve called out the usual sources of the problem and/or solutions in italics below</em>):</p><h3><strong>I Diagnosing Execution problems</strong>:</h3><ol><li>Frequent disagreements on the priority of features or implementation details that slow down execution ( → <em>Roadmap and priorities not clear or not aligned indicating the need for a better process</em>)</li><li>Team spinning for several weeks assuming they need to align on a decision when the real decision maker is the VP of the org and the team’s job is to just present options and their best recommendation in a leadership review ( → <em>Purpose of the leadership review not defined well, process for reviews not smooth, who the decision-makers are, is not clear</em>)</li><li>Slow or bloated product development (Metrics, Roadmapping, Product brief, Hypothesis-driven experimentation, Launch workback timeline etc.) <em>I cover this in </em><a href="https://medium.com/@ramyasethu/product-management-frameworks-3p-the-art-science-of-product-72a66faafb42"><em>Part 3</em></a><em> of this 3P series on Product!</em></li><li>Delays in product launches because the cross-functional team (legal, policy, privacy, etc.) is not looped in at the right time ( → <em>Need clear process for alignment with XFN. </em><a href="https://medium.com/@ramyasethu/product-management-frameworks-3p-the-art-science-of-product-72a66faafb42"><em>More in my post on Product.</em></a>)</li><li>Too much process: E.g. roadmap planning takes forever, leadership reviews and reporting take forever ( <em>→ (i) add structure to the planning process, explicitly timebox roadmap planning, use retros to determine the right cadence for roadmap planning, and explicitly debug the most time-consuming parts of the process, (ii) set explicit expectations with leadership on the cadence of reporting &amp; reviews and (iii) establish the purpose and expected outcome of each review, and (iv) incentivize/recognize process efficiency improvements</em>)</li></ol><h3>II Lack of clarity:</h3><ol><li>Lack of Team rituals and processes — manifests in delays and costs in several ways. ( → <em>Do people understand the purpose of team meetings, is there a smooth flow to each meeting? Do you have established forums for different themes/problems such as all hands, off-sites, hackathons, war-rooms, retrospectives, and premortems?</em>)</li><li>Team not sure what the source of truth for a project is aka too many documents ( → <em>Need to establish a </em><a href="https://naomi.com/canonical-everything-c85441a84e70"><em>canonical source of truth</em></a><em> for key documents and socialize the document instead of each person spawning their own version of the truth</em>)</li><li>Ineffective leadership reviews i.e. team or leadership does not get what they want out of the review ( → <em>Establish purpose of review, a recommended review format (update/decision needed/escalation?), required attendees and how/where people can learn more</em>). See the <a href="https://naomi.com/the-traffic-light-approach-to-problem-solving-7b3d6e42acc2">Traffic Light</a> approach.</li><li>People not rowing in the same direction — this usually manifests as a lack of alignment on the vision, strategy, goals, or priorities of projects and work happening in silos. ( → <em>Do clear communication channels exist for specific purposes? Where can you find the product roadmaps? Leadership decks and feedback? Weekly team updates? How can people keep up with the latest, most relevant information?</em>)</li></ol><h3>III Lack of alignment/Frequent disagreements</h3><ol><li>Lack of alignment up the chain, information in silos, product pivots or priorities not clear ( <em>→ The basic principle needed here is Communication of the right information on the right channels at the right time. Some examples are lack of clear upward comms through the different levels, lack of a consistent set of channels to post product updates, no process around how leadership feedback is communicated or how teams catch up on product work in adjacent teams to avoid overlap, etc.)</em></li><li>Ineffective escalations or escalations that are unpleasant surprises (<em>→ Some best practices are: (i) to jointly agree with the opposing team or person that it is unlikely they will come up with a common solution and escalate together — this way no party is taken by surprise (ii) to escalate in a timely manner and not treat escalation as a bad thing to do (iii) to ensure escalation meetings have the necessary decision-makers in the room)</em></li><li>Team talking past each other/Low trust/Lack of ‘connection’ ( → <em>While there are many reasons for low trust, establishing a regular forum for people to connect informally, a culture that shares gratitude and props, and teams that connect and have fun help establish trust which makes aligning on decisions easier</em>)</li></ol><p>Have your favorite process improvement to add to this list? Let me know! You can find part 1 on People <a href="https://medium.com/@ramyasethu/product-management-frameworks-563969cdfed5">here</a> and part 3 on Product <a href="https://medium.com/@ramyasethu/product-management-frameworks-3p-the-art-science-of-product-72a66faafb42">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9a19a7880834" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Product Management Frameworks:]]></title>
            <link>https://medium.com/@ramyasethu/product-management-frameworks-563969cdfed5?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/563969cdfed5</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-management-skills]]></category>
            <category><![CDATA[product-framework]]></category>
            <category><![CDATA[first-principles-thinking]]></category>
            <category><![CDATA[product-manager-career]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Mon, 20 Nov 2023 01:31:43 GMT</pubDate>
            <atom:updated>2023-12-24T19:26:23.438Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/928/1*Wk9_NdQFd-jOBZA0moidXA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/938/1*1H7g0X1Dtz1Q8rhCoD5NUA.png" /></figure><h3>Product Management Frameworks, the 3Ps: People</h3><p>A decade at Meta taught me a lot about <a href="https://medium.com/agileinsider/simple-lessons-i-learned-as-a-product-manager-7a8f8611bbdc">better ways to think</a> using frameworks. I also <a href="https://medium.com/agileinsider/everything-i-read-on-my-way-to-a-product-management-role-bb039bea453e">read a lot</a> and am fascinated by people who can switch on their higher modes of thinking to make better decisions. This post is a compilation of a set of frameworks and principles that I’ve learned both by building products and by observing how smart people around me make decisions. I’ve categorized the frameworks into the 3Ps: People, Process, and Product. This post is about People. Stay tuned for future posts on Process and Product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TTNILTSKtW3MHKzpBxo6XA.png" /></figure><h3>People</h3><p><em>I share below a set of themes I regularly rely on as a set of guiding principles to navigate the first P, People. This is by no means comprehensive but it’s a good set to start with and I’m hoping I’ll keep adding to this post as and when I find time.</em></p><p><strong>Hiring: </strong>The one key <em>first principle</em> I’ve kept in mind while hiring for my team is <a href="https://fs.blog/warren-buffett-the-three-things-i-look-for-in-a-person/">finding people who demonstrate three things</a>: integrity, energy, and intelligence, the last of the three being the least important criteria, and the first two being non-negotiables. Energy and intelligence are easier to discern in an interview but it’s possible to dig deep enough during behavioral interviews to get a glimpse into integrity-driven decision-making.</p><p><strong>Culture: </strong>Culture is the set of norms, beliefs, and values that determine what people do and how decisions are made when you or the leader is not in the room. It’s the <em>intangible</em> that helps scale people’s behavior and decision-making without you having to drive it. Whether it’s consistent dogfooding, understanding the details of a product, giving props to the givers and the helpers in a company, observing user research sessions to deeply understand the user feedback or regularly working over weekends, the norms are set by the culture we have built together. Usually, leadership and initial employees at a company set the tone for culture. A framework I’ve developed to understand/shape a company or org’s culture is to (i) First, deeply understand/establish incentives that drive decisions (ii) Learn what the company values and allocates or does not allocate resources to (design? data and metrics? research?) (iii) Study who/what gets rewarded or punished. These are the invisible pointers to the culture of a company or an org. There’s a lot to unpack here, maybe in a future post.</p><p><strong>Finding your tribe: </strong>No matter where you are in the pecking order, work is a lot more fun with friends. Sometimes, this has been my PM circle with peers or people I’ve coached that made my day a little better. If nothing else went great that day, I still have the satisfaction of having helped someone in a PM circle. Other times, it’s been friends within my team that I work closely with or people within my extended network that I’ve worked with in the past. These ‘pockets of connection’ moments have often cheered me up on tough days! So my basic framework here is to establish connections and reach out to people, find your tribe, and set up pockets of sanity in your otherwise crazy day!</p><p><strong>Upward Management: </strong>Yes, upward management is a thing. Learn it and practice it. At its essence, it boils down to making your manager’s job (and thus your life) easier. Here are some simple rules to get you started:</p><ol><li>Learn your manager’s preferred communication style and adapt to it.</li><li>Understand her management style. Does she like frequent chat updates, wants you to loop her in only when you are blocked, prefers early escalation? Adapt accordingly. Note that your manager’s style may differ from your skip manager’s style and you need to adapt accordingly.</li><li>Assume that she has more context than you when she makes a recommendation (unless you have reason to believe otherwise). Understand why she’s making a recommendation, ask, and confirm (Is she casually suggesting you try this approach? Or does she have context from other meetings? Based on some other leadership guidance that you might not be aware of?). This is true of leaders up the chain as well. They have broader context and you can assume their recommendations incorporate the considerations and constraints of this broader context.</li></ol><p><strong>On being a leader: </strong>One leadership quote that’s resonated with me is to <a href="https://www.youtube.com/watch?v=xKi-lf4dVzY">learn to be the leader you wish you’d had</a>. You can always act in a way that’s worthy of a leader and that’s in your hands. A couple of things that have helped shape my leadership style:</p><ol><li><a href="https://themanagershandbook.com/coaching-and-feedback/radical-candor">Radical candor</a>: Challenge directly, care personally.</li><li>Build trust: Be truthful, say what you will do, do what you’ve said, do this consistently.</li><li>Have people’s backs. When the going gets tough, do this even more. Don’t let the chaos and politics of a situation compromise your values. In other words, remember to also have your back.</li></ol><p><strong>Feedback: </strong>Knowing how to give feedback effectively and receive feedback thoughtfully is a key skill to continuously learn and evolve. Through experimentation and missed opportunities, I’ve learned a few basics to keep in mind while giving or receiving feedback:</p><ol><li><strong>Receiving constructive feedback</strong>, especially if you are a high performer, is never easy. Your mind tends to focus on the one constructive element in an assessment full of songs and praises about you. Some principles to handle this well: (i) <em>In the moment</em>: Resist the reactive impulse to push back on the feedback or make excuses for it. The only response in the moment is, ‘Tell me more so I understand better’ or if you do understand the feedback, thank the other person for sharing and say you will keep the feedback in mind. (ii) <em>After you receive the feedback</em>: If you want to: dwell on it for a reasonable amount of time, call a friend and talk about it, or get your spirits up with a sweet indulgence, go for it. Don’t resist it or it tends to fester in the background. After this period of dwelling on it, try to understand the feedback and dig for the kernel of truth in it (it’s always there), ask a few trusted folks who know you well, if some themes resonate with them or how they might interpret them. (iii) <em>The next time you address the feedback: </em>Go with a positive attitude to thank the person again for the feedback and pick what was valuable about it and how you’ve thought of addressing the feedback. Do they have any other recommendations to try? This demonstrates a mature, thoughtful approach to taking feedback making it more likely that people share feedback with you and there are no surprises.</li><li><strong>Receiving positive feedback</strong>: You know you worked hard and completely deserve the positive feedback. Resist the impulse to brush it off, diminish it or praise the other person in return. Thank the person who shared the feedback and show your gratitude for it. If in a public forum, this is a great opportunity to recognize your team or people who helped you accomplish the goal.</li><li><strong>Being nice vs kind</strong>: A culture that confuses kindness with niceness will make people hesitate having hard conversations, sharing honest feedback, and they might resort to surface-level niceties all the time. This helps no-one. <em>We all make each other better if we share real feedback in an honest but kind and caring manner, in the moment, privately and directly when possible.</em> A best practice that’s worked well for me is to share the feedback directly when possible rather than surprising the person and sharing feedback with their manager. As a manager, I’ve often thanked people who have shared feedback to me about the people I support but have also encouraged them to share the feedback directly with the person.</li><li><strong>Sharing feedback upwards</strong>: Some questions to ask before you share feedback upwards: (i) Have you earned the respect of the manager or leader you want to give feedback to? (ii) Is the time right? (i.e. not: right before their review with the CEO, board meeting, right after a messy reorg they are still handling — you get the drift) (iii) Are you ready to tackle a hard conversation if this becomes one? A rubric to understand when to have a <a href="https://brightspotcdn.byui.edu/2b/0d/f1cf5d02420b9ce0716b042b7b31/summary-of-crucial-conversations.pdf">crucial conversation</a> that I’ve found super useful: (i) Are emotions high (e.g. your feelings were hurt)? (ii) Are the stakes high (e.g. a major setback that could lead to team attrition)? (iii) Do you have reason to believe you are misaligned and will continue to be in the future?</li></ol><p><strong>Performance assessment (as a manager): </strong>Ah, every manager’s favorite time of the year! Performance assessments are an inherently stress-inducing time for managers because these conversations are high-stakes situations. People take these assessments seriously and so, as a manager, you want to be as thorough, thoughtful, and precise as possible and ensure you are offering actionable, reasonable, valuable, and concrete feedback, and rewarding people for their hard work. Some principles to keep in mind:</p><ol><li><em>Care and invest in the people you support.</em> Their career matters to them a lot and so, it should matter to you, their manager. What do they care about? What motivates them? What enervates them? What are their career aspirations? Personal goals? Spend time and learn about these things. Connect them to people who might be helpful (coaches, mentors, sponsors, friends). When you write their performance assessment, consider this broader picture of them.</li><li><em>Avoid unpleasant surprises</em>: As much as possible, share timely feedback and dimensions to improve upon so that people are not surprised during their performance assessment conversations. If they are surprised, you, as their manager, have failed to support them as best as you could.</li><li><em>Support, guide, and help your people and team succeed</em>: As a manager: it’s your job to <em>support</em> the people you manage, this means helping them grow and thrive (see Radical Candor above) and doing what you can to help them towards their aspirations, <em>guide</em> them by sharing broader context and leadership feedback, as transparently as possible. And remember, the small things matter: did you remember their work anniversary? A personal accomplishment they shared with you? Recognize the givers in your team? Share well-deserved praise for a smooth review. The big things also matter: Did you spend enough time on their promotion packet? Did you socialize it with the powers that be? Did you fight for a better rating because you know they deserved it?</li></ol><p><em>Let me know if you found this useful or have your own frameworks to add to this . You can find part 2 on Process </em><a href="https://medium.com/@ramyasethu/product-management-framework-process-9a19a7880834"><em>here</em></a><em> and part 3 on Product </em><a href="https://medium.com/@ramyasethu/product-management-frameworks-3p-the-art-science-of-product-72a66faafb42"><em>here</em></a><em>!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=563969cdfed5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A list of my favorite books]]></title>
            <link>https://medium.com/@ramyasethu/my-favorite-books-work-in-progress-4d09ca73b8f4?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/4d09ca73b8f4</guid>
            <category><![CDATA[books]]></category>
            <category><![CDATA[favorite-books]]></category>
            <category><![CDATA[list-of-books]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Sat, 28 Oct 2023 02:09:17 GMT</pubDate>
            <atom:updated>2024-01-06T22:08:23.673Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/509/1*RUZLBxSZOccdMA9XWusv9g.jpeg" /><figcaption>Generated using AI</figcaption></figure><p>Every once in a while, I imagine being asked for a list of my favorite books and I hear my mouth blurting out names of a few inconsequential books I’ve read. There’s always a collar mike or a handheld microphone involved. I have no idea why. I’m in my mid-forties and no one has ever asked me for a list of my favorite books, let alone with a microphone involved. But, I thought it would be fun to come up with a list anyway and speak into the imaginary microphone!</p><p>As a lifelong lover of books and author, I found it daunting to choose and leave out some beloved books, like choosing who to invite to your birthday party. I figured this post could be a continuous work-in-progress, I could keep building on it and that made me feel better☺️.</p><p>I’ve jotted down a line next to each book with no particular theme in mind, whatever captured my fancy about the story. I’ve picked generated AI images that capture the essence of the book. Perhaps, you will read these books one day and carry a piece of the story in your heart, like I do.</p><h3><strong>Fiction:</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/473/1*wLqJrYG_Bjc_BvyHgjmFhw.png" /><figcaption>Generated by Starryai</figcaption></figure><p><strong><em>Love story, Erich Segal</em></strong>: Ah! The unforgettable opening line, “<em>What can you say about a 25-year-old girl who died?”!</em> Recently, my teenage daughter asked me for a book that would move her enough to make her cry and I gave her this book. I’m so grateful that I could incorporate a tiny bit of this book in my <a href="https://www.youtube.com/watch?v=AgAz-BRN6x8&amp;t=21s">short film</a> 💔</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/473/1*0J2sc2eCtFwyMzNvbNtj4g.png" /></figure><p><strong><em>Sphere, Michael Crichton</em></strong><em>: </em>I was probably a teenager when I read this book. I remember being scared of the dark and going to bed, imagining sequences from the book! This is probably my favorite of all the Crichton books I’ve read.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/270/0*qPJ9IIfOKI3iURQz.jpg" /></figure><p><strong><em>Kane and Abel, Jeffrey Archer</em></strong><em>: </em>There is that one unforgettable scene described in the book, where the arch-enemies walk by each other after all the feuds and struggles, and gently, one of them tips his hat and acknowledges his nemesis. Classic!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/477/1*k4-TZq7v6CGnZZeAdSmhLA.png" /></figure><p><strong><em>Palace of Illusions, Chitra Divakaruni</em></strong><em>: </em>I remember being enthralled and a bit shocked at this beautiful reimagination of the relationship between Karna and Draupadi. I remember the palace, the grandma, the angst and tragedy of unfulfilled love in an unforgivable situation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/464/1*8s0HIsolmhT7dkfcQRq1pw.png" /></figure><p><strong><em>Yargo, Jacqueline Susann</em></strong><em>:</em> My cousin recommended this book to me and it was my first glimpse of a new world, bound by its own power hierarchies and rules, tested and stretched by the pull of human emotions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/245/0*eaSD6WMs17a92eYP.jpg" /></figure><p><strong><em>Doctors, Erich Segal</em></strong>, also in the same genre, <strong><em>Hotel, Arthur Hailey</em></strong><em>: </em>I remember staying up late into the night as a teen, drawn to the worlds we have built within our world, the politics, processes, and mechanics of what makes a hotel run or an airport work or the life of doctors. Each is a deep dive into a world I’m unfamiliar with but impossibly interesting and of course, the kdrama like pull of a strong love story throughout the plot!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/468/1*J74LJm0UudZNzvUoHWFh-w.png" /></figure><p><strong><em>Emma, Pride and Prejudice, Jane Austen</em></strong>: I absolutely adore Jane Austen’s style of writing and her witty language. Here are two that are just perfect:</p><blockquote>“If I loved you less, I might be able to talk about it more.”</blockquote><blockquote>“It is a truth universally acknowledged, that a single man in possession of a good fortune, must be in want of a wife.”</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/470/1*0M4Xh9a4962Q0YT8FUu4FA.png" /></figure><p><strong><em>Monsoon Diary, A memoir with recipes, Shoba Narayanan: </em></strong>As a fresh off the boat Indian immigrant in the US (a few decades ago), I did not often find books that captured my experience or the confusion I felt. I eventually did write a book on my immigrant experiences (<a href="https://www.amazon.com/r2i-dreams-Here-Parth-Pandya/dp/1500705802/">Return to India dreams</a>) but long before that Shoba Narayan’s book offered me a reassuring journey into someone’s experience that felt similar to mine. The book filled me with nostalgia, colors and smells that felt like home.</p><h3>Young Adults and Kids’ Fiction:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/473/1*B1CM2mNqgFJ_uBTnPL4moA.png" /></figure><p><strong><em>Gossamer, Lois Lowry</em></strong><em>: </em>What a lovely word, gossamer! And what an imaginative and delightful story about the littlest one and the dreams we dream!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/454/1*dJTYhCUqpzY0N4EL3YHGew.png" /></figure><p><em>The probability of everything, Sarah Everett: </em>I started reading this book with my 10 year old who is interested in Math and the story turned out to be so much more: beauty, love, grief and justice. I loved the innocent and liminal beauty navigating past and present in the book. Fair warning though, the book did make my daughter cry in the end!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/311/1*OH7jP0ncjpwHYtoQ0OadBg.jpeg" /></figure><p><em>Little Women, Louisa May Alcott: </em>Who can forget the line, <em>“Christmas won’t be Christmas without presents”</em> and the story of Meg, Jo, Amy and Beth! Enough said.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/464/1*senoAytmqWrenIMUP9oq6g.png" /></figure><p><em>Stargirl, Jerry Spinelli: </em>I loved the wonder and beauty in this book and how one little girl was brave enough to be unabashedly herself, leaving an indelible mark on the lives of people around her.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/336/1*NiYIj7K_ahU5l90rh9g9aQ.jpeg" /></figure><p><em>Little house in the Big Woods, Laura Ingalls Wilder: </em>In our house, we are gently addicted to both this book series and the TV series on Amazon. A life so much slower, so very different, with difficult choices we can only imagine, and yet filled with universal human emotions and struggles we are all intimately familiar with.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/464/1*NViuDRouDKIFDN8iuHLY8g.png" /></figure><p><em>The possibility of fireflies, Dominique Paul: </em>This little book, like Stargirl, is one that holds beauty in its words and the way the story unfolds. A heavy read though for younger audiences.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/463/1*HR7ewCURx2Z872UziRzzJw.png" /></figure><p><em>Farenheit 451, Ray Bradbury: </em>One of those books that makes you do a double-take and go back to re-read parts of it to really take in the story. Startling, brilliant and a caution for the future!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/476/1*hmaTxzSOxz9ID8c__psoiA.png" /></figure><p><em>The giver of stars, Jojo Moyes</em>: I finished this book reading late into the night, transported to earlier times, deep in the Eastern Kentucky hillside communities, imagining courageous and spirited women on horses, transporting library books to people that didn’t have access to books, facts, knowledge. So removed from my life and so drawn was I to the events and characters in the book. An absolutely compelling, gripping, courageous tale of brave women, unforgiving times and the power of community.</p><h3><strong>Non-Fiction:</strong></h3><p><em>A Swim in a Pond in the Rain</em>, <em>George Saunders: </em>If you are a writer, you will absolutely love this book! I share a quote from the book below and you will know why I picked this book,</p><blockquote>Your reading energy did something at that phrase. Were you in or out? With me or against me? Compelled slightly forward or held slightly back? By that little tussle, you know I’m here. And I know you’re there. That phrase is a little corridor connecting us, giving us a fragment of the world over which to tussle, i.e., connect. There are many versions of you, in you. To which one am I speaking, when I write? The best one. The one most like my best one. Those two best versions of us, in a moment of reading, exit our usual selves and, at a location created by mutual respect, become one. That’s a pretty hopeful model of human interaction: two people, mutually respectful, leaning in, one speaking so as to compel, the other listening, willing to be charmed. That a person can work with.</blockquote><p><em>Managing Oneself</em>, <em>Peter Drucker: </em>I find myself referring to it often, reading from it to teach my kids. My dad used to refer to Peter Drucker as well. A timeless classic and one of the original self-help books.</p><p><em>Many Masters Many Lives</em>, <em>Brian Weiss: </em>I remember the emotions this book left me with, that there are threads beyond this life that connect us, that leave an imprint on us and there is more to life than meets the eye. It made me think of the Universe, Grace, God.</p><p><em>Fingerprints of God</em>, <em>Barbara Bradley Haggerty: </em>I love Barbara’s thorough, investigative style of tackling a deep topic. Does God exist? What can we learn from various domains and disciplines about ‘God’? What does Science tell us, what do spiritual experiences teach us, what do near death experiences have in common? Tying these various threads together, Barbara takes us on an eye-opening journey of discovering the fingerprints of God.</p><p><em>How to Fail at Almost Everything and Still Win Big</em>, <em>Scott Adams: </em>I’ve always admired how Scott Adams was able to take what he loved doing, drawing whimsical cartoons based on his own career and life experiences, and shape it into a way to make a living and shine brightly while doing so! A great read for anyone trying to bring closer together, their creative and career-driven lives.</p><p><em>The Celestine Prophecy</em>, <em>James Redfield: </em>I remember this one scene from the book, where the author describes feeling physically blown back by the energy and vibration that someone brought into the room. I experience this often, the ability to minutely feel the energy people bring into my space, and finally found a book that articulated this concept of energy and vibration the way I experience it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/475/1*muLbPs15YzlGEkZuHgwehw.png" /></figure><p><em>A Grief Observed, C.S. Lewis:</em></p><p>I found this book comforting after I lost my dad a few years ago. I’ll share a few quotes from the book to give you a glimpse of the power of Lewis’ words.</p><blockquote>We have seen the faces of those we know best so variously, from so many angles, in so many lights, with so many expressions — walking, sleeping, laughing, crying, eating, talking, thinking — that all the impressions crowd into our memory together and cancel out into a mere blur. But her voice is still vivid. The remembered voice — that can turn me at any moment to a whimpering child.</blockquote><blockquote>Slowly, quietly, like snow-flakes — like the small flakes that come when it is going to snow all night == little flakes of me, my impressions, my selections, are settling down on the image of her.</blockquote><blockquote>And suddenly at the very moment when, so far, I mourned H. least, I remembered her best. Indeed it was something (almost) better than memory; an instantaneous, unanswerable impression. To say it was like a meeting would be going too far. Yet there was that in it which tempts one to use those words.</blockquote><p><em>How Will You Measure Your Life?</em>, <em>Brene Brown: </em>The first set of books that touch on difficult and important topics like vulnerability and shame and how they shape your life experience and decisions.</p><p><em>Heart: A Story</em>, <em>Sandeep Jauhar: </em>After reading this book, I understood how the heart is resilient, capable of holding so much love and yet so delicate, beating now, gone in an instant.</p><p><em>Breath</em>, <em>James Nestor: </em>I’ve always understood mind and body work is as important as cognitive work but hadn’t thought much about how crucial the breath is to our physical and mental health, mindset and life. This book taught me various, concrete techniques, with scientific reasoning and experimental data, on how to breathe right and why it is so very critical in our modern lives.</p><p><em>Decisive, Chip &amp; Dan Heath: </em>I still remember the <a href="https://heathbrothers.com/wp-content/uploads/2013/03/The_WRAP_Process_one_pager.pdf">WRAP</a> framework (from the book) and use it when I need to make tough decisions!</p><p><em>Home safe home, Debra Lynn Dadd: </em>After reading this book a few decades ago, I got deeply interested in understanding the ways toxins and chemicals enter our bodies, and to this day, I try to follow a lifestyle that mainly incorporates organic and natural food and products.</p><p><em>Transformative Experience, L.A.Paul: </em>This book taught me that I couldn’t always make a logical, pros/cons based decision if the world created by the decision was so dramatically different from now, that I become a different person altogether. In this new world, the “I” of today will not be able to predict what the “I” of the future would prefer and what decisions would make sense for her. A fascinating philosophy-rich book!</p><p><em>The happiness hypothesis, Jonathan Haidt: </em>The one thing I will always remember from this book is the visual of the human trying to ride an elephant half submerged in a turbulent ocean, the human representing the rational, cognitive brain and the wild elephant, the emotional, reactive, amygdala tossing us every which way, per its whim. All we can do is to direct the wild animal and push to progress in the right direction, fully recognizing we don’t always control how and where the elephant wants to romp!</p><p><em>How will you measure your life? Clayton Christensen: </em>This is another classic that I often refer to to help me with my life and career decisions. It has some memorable definitions: resume vs eulogy virtues, “likeness” of who you want to become, theory of motivation etc. Every now and then, I re-read sections of the book to see if I’m indeed following my true, intrinsic motivations or chasing behind some glittery, external reward. It’s a great book to reflect on your life and career.</p><h3><strong>Just for fun or I need to relax and de-stress and need a summer read:</strong></h3><p><em>Books by Jennifer Weiner, Jane Green, Sophie Kinsella’s Shopaholic series, the P.G. Wodehouse and Jeeves series. As a kid, I also loved the Nancy Drew series and was addicted to Archies and Richie Rich comics! My dad would often take my brother and me to the local library and we would get a fresh new copy of the latest double digest series — that was life!</em></p><p><em>I aim for this to be a work in progress and maybe, I’ll come back and add new books every now and then! If you have your absolute favorites I should read, add to this list, let me know! Thank you for reading.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4d09ca73b8f4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Can AI do better than a 14 year old?]]></title>
            <link>https://medium.com/@ramyasethu/can-ai-do-better-than-my-14-year-old-18eb7eeed9c?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/18eb7eeed9c</guid>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[openai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[creativity]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Tue, 06 Dec 2022 02:19:34 GMT</pubDate>
            <atom:updated>2022-12-06T05:23:18.818Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*i523Fzpd60T-K3XNrnznlQ.png" /><figcaption>Generated by Midjourney bot</figcaption></figure><p>So my 14 year old daughter has a <a href="https://medium.com/@radhika934">medium blog</a> where she ties her two favorite hobbies together and generates original content: a love for photography and her favorite books. When she puts her mind to it, she comes up with pretty good stuff. For one of her recent posts, she decided to be more vulnerable and her commentary is more raw than her other more polished posts. This was a hard thing for her to do, for example, to admit that she cried after reading ‘The fault in our stars’ — you can read her post <a href="https://medium.com/@radhika934/love-the-fault-in-our-stars-6d39faeb3af">here</a>.</p><p>I’ve spent a bunch of time thinking about <a href="https://medium.com/@ramyasethu/supporting-human-values-and-human-flourishing-c1553de1a0ce">AI and human values</a>, part of it because it related to my work and part of it because I’m obsessed with the topic and more recently, have been following all the fun generative AI developments and wondered if <a href="https://chat.openai.com/chat">ChatGPT</a> and Dall-E could beat her in sounding human, being creative and generating authentic sounding, original content :)</p><p>So, here goes my experiment:</p><p>First, I went to ChatGPT and asked it to connect ‘The Fault in our stars’ to a photograph:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*swUbUF4ikvTqMfG7MjkvRQ.png" /></figure><p>Hmm, maybe the question was too abstract for it to generate something interesting. It did correct my typo though!</p><p>So then, I asked it an easier question: Summarize the book.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AiboQRf1itzS-kCjRMWCog.png" /></figure><p>Ok, that was good and expected. I tried something a bit more complicated — about feelings and that didn’t generate interesting results either.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ECjFEJOyVR85EZ9lpULW5w.png" /></figure><p>Ok fine. I then wanted to see if Dall E or Midjourney would do better than the photos she took that reminded her of the book. This is her photo:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7dQ-4pmMmAMsDupv.jpeg" /></figure><p>And this is what craiyon.com generated for the prompt: ‘a photograph showing light in the sea surrounded by darkness’</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rbiv-kLqLE6oQiPvdhbYdg.png" /></figure><p>Not bad. Then, I tried Midjourney using the /imagine bot in their discord server:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-r12QBm51DwLS4CtdSxb0w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/938/1*-SIXy2FOKp1zEMuyI40Z_w.png" /></figure><p>Pretty cool! But now, who plays the role of the human brain seeing the connection between the photo and the story in the book and giving it that “human touch”? Who’s going to connect the dots and say <em>“The carefully thought-out words woven into this incredible, incredible story have a way of making you laugh and bawl your eyes out in the same sentence” </em>and through those very words find a way to connect to the reader?</p><p>This experiment reminded me of something one of my favorite authors, George Saunders, had said in his book, ‘A swim in the pond in the rain’. I tried looking it up to no avail (at least not yet)…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5w8BKmKnn0lZewAiSOUGRw.png" /></figure><p>But here’s the actual quote from his <a href="https://www.amazon.com/Swim-Pond-Rain-Russians-Writing/dp/1984856022">book</a>:</p><p>“<em>By that little tussle, you know I’m here. And I know you’re there. That phrase is a little corridor connecting us, giving us a fragment of the world over which to tussle i.e. connect. There are many versions of you, in you. To which one am I speaking, when I write? The best one. The one most like my best one. Those two best versions of us, in a moment of reading, exit our usual selves and, at a location created by mutual respect, become one. That’s a pretty hopeful model of human interaction: two people, mutually respectful, leaning in, one speaking so as to compel, the other listening, willing to be charmed. That, a person can work with”</em></p><p>I don’t think that technology layer exists today to create that invisible connection between the writer and the reader, making you see the world in a way you had not just moments earlier and making you laugh and cry together. At least, not yet. When that layer is also replaced by technology, wow, would that be an incredible technological advancement or human kind preparing for true <a href="https://www.scientificamerican.com/article/artificial-general-intelligence-is-not-as-imminent-as-you-might-think1/">artificial general intelligence</a>, leaving them far behind?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=18eb7eeed9c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Human Value Alignment: Supporting human values and human flourishing.]]></title>
            <link>https://medium.com/@ramyasethu/supporting-human-values-and-human-flourishing-c1553de1a0ce?source=rss-afb860b991e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1553de1a0ce</guid>
            <category><![CDATA[human-values]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[recommender-systems]]></category>
            <dc:creator><![CDATA[Ramya Sethuraman]]></dc:creator>
            <pubDate>Thu, 24 Mar 2022 16:38:01 GMT</pubDate>
            <atom:updated>2023-07-02T17:23:21.923Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*s0_n05IAqygaiGDX" /><figcaption>Image from futureoflife.org</figcaption></figure><p>(This is a modified version of an internal note I posted within Meta)</p><p><em>Grab a cup of coffee, prepare for story time and let’s chat a bit about AI, human values and human flourishing, shall we?</em></p><h3>Some prescient stories first</h3><p>Have you heard of the story of King Midas in Greek mythology? The King asked for a wish for everything he touches to turn into gold. The next day, he touched his beloved daughter who turned into gold! And there is the story of the <a href="https://en.wikipedia.org/wiki/The_Sorcerer%27s_Apprentice">Sorcerer’s Apprentice</a> who enchants a broom to fetch water because he is bored to do it, the broom keeps fetching water until it floods the room and overpowers the apprentice. A modern-day version of this story is the<a href="https://www.lesswrong.com/tag/paperclip-maximizer"> paperclip maximizer </a>thought experiment, an artificial general intelligence that has been tasked with the objective of maximizing paperclips and ends up making itself super-intelligent and covering the earth and space with paperclips. This thought experiment seems silly to us because even if we are given an objective to maximize paperclips, we would pursue it in the context of human values such as learning, love, well-being and so on. The AI however is not constrained by such values. So, unlike what <a href="https://en.wikipedia.org/wiki/The_Terminator">Terminator</a> teaches us, the problem seems to be not that AI will take over humans but that they’d do “too good a job” of doing precisely what we asked it to do.</p><p><em>“Robots </em><a href="https://futureoflife.org/2017/02/03/align-artificial-intelligence-with-human-values/"><em>aren’t going to try to revolt against humanity</em></a><em>, they’ll just try to optimize whatever we tell them to do. So we need to make sure to tell them to optimize for the world we actually want.”</em></p><p>These stories lead us to the question of how we align AI with human values and leads us to the AI value alignment problem, pioneered by Stuart Russell. Before we jump into AI alignment, we should further establish why we need to think about AI in the context of human values. Sure, a super-intelligent AI might optimize for its objective with a single minded focus with disastrous consequences,<a href="https://en.wikipedia.org/wiki/AI_alignment"> disable its off switch</a> and so on. But, we are perhaps a few decades away from such intelligence and we have reasonable ways to track whether we are providing some form of benefit to people, so, why worry about this now?</p><p>For example, if people come back to the platform, comment and interact more with others, watch videos and spend time on the platform and so on, isn’t that a reasonably good measure of the value we provide to people? While these metrics may be a reasonable proxy to whether we are providing value to people, they are far from perfect. Before diving deeper into value alignment, I want to make the case for why it is essential for us to take a very close look at our current recommender systems.</p><h3>Unintended Consequences</h3><p>A recommender system (for e.g. a Feed of any sort) is an AI system that presents users with a set of items based on a set of factors such as past user behavior, user and content attributes (e.g. you often liked posts from NY Times, you watch a lot of videos, this post is a live video). You can read more about <a href="https://participatoryml.github.io/papers/2020/42.pdf">recommender systems and the current state here</a>. I want to touch upon some concrete problems we have with our recommender systems and then make a case for why a fundamental solution is needed.</p><p>There are <a href="https://web.stanford.edu/~gentzkow/research/facebook.pdf">studies</a> published on effects of social media on people’s subjective well being and others that make claims about social media use and effect on sleep, health, work performance etc. I won’t get into those debates here. The point I am trying to make is any <a href="https://participatoryml.github.io/papers/2020/42.pdf">metric that our ML systems optimize for</a> has to be carefully scrutinized and evaluated periodically to see if it maps back to what it intended to solve for: whether we are creating value for people. Any metric is a proxy for value created. When we optimize for this metric blindly, we lose track of the original intent.</p><p>For example, moving from optimizing for clicks to optimizing for time-in-app <a href="https://participatoryml.github.io/papers/2020/42.pdf">may reduce clickbait but could increase addiction</a>. We exchange an existing problem with a proxy metric with a newer problem because we are trying to reactively fix a problem created by the first proxy metric. The bigger takeaway here is that what we have in the first place is a “proxy metric”, an approximation, at best, of the user’s real intent and in some sense, we see <a href="https://en.wikipedia.org/wiki/Goodhart%27s_law">Goodharts’ law</a> at play here:</p><p><strong><em>When a measure becomes a target, it ceases to be a good measure.</em></strong></p><h3>How might we tackle these problems?</h3><p>Let me start be defining behavioral metrics. Behavioral metrics capture people’s behaviors on the platform, typically engagement with content on the platform. They are a good indication that people are finding value on the platform. But, some problems created by behavioral metrics are unintended consequences, over-optimizing for the metric or gaming the metric in ways we did not anticipate, to name a few. Besides people’s behaviors on the platform, a <a href="https://partnershiponai.org/beyond-engagement-aligning-algorithmic-recommendations-with-prosocial-goals/">few other approaches we can rely on to get richer feedback from people is through surveys and controls</a>.<em> </em>One way we do that on News Feed is by asking people if they feel that the <a href="https://about.fb.com/news/2019/05/more-personalized-experiences/">post they see is worth their time</a> and incorporating this signal in ranking. The advantage here is that surveys represent ground truth (to a certain extent) from people (people directly tell us their preferences) rather than us making assumptions about what people want based on their actions on the platform. Of course, this world-view is overly simplified and there are a number of challenges with survey based ML models which I will leave out for the purposes of this note but ping me if you’d like to riff on this!</p><p>So now we have behavioral models, survey based models and if we get things wrong with ranking, we can rely on controls i.e. explicit controls given to people to customize their feeds and tell us when we got something wrong with ranking (e.g. the ability to snooze a person, hide a post, report a post, <a href="https://about.fb.com/news/2021/03/more-control-and-context-in-news-feed/">mark someone as favorite</a> and so on). While surveys and controls offer rich ways to get feedback from people, they are not a perfect solution. For example, how do we know a post’s worthwhileness is the right sentiment construct to optimize for? How do we know if our systems are providing longer term value to people instead of in-the-moment benefits? How do we know which sentiment constructs or drivers are the right ones to optimize for? And more broadly what values do these sentiment constructs represent? Is there consensus even among external experts on these set of values? Have we started identifying and partnering with the right external organizations and experts to set ambitious directions and goals for this work? And more tactically, do we believe a reliance on sentiment as measured through surveys to balance our behavioral models in feed, will help address some of the problems we currently see with our recommender systems? And a second order question is even if we have reasonable consensus on these values/principles and approaches, do we know how to implement these in our AI systems? These are the fundamental questions involved in AI Value Alignment. They can broadly be categorized into:</p><p><strong>Normative questions on value alignment</strong>: what set of values or principles should we encode in our recommender systems? Is there consensus among moral philosophers, ethicists, psychologists and external experts on what these might be?</p><p><strong>Technical questions on value alignment</strong>: Once we have reasonable consensus on #1, do we have the right technical AI investments to implement these recommendations? One example of this could be investments in reinforcement learning or inverse reinforcement learning as being necessary to explore #1 further.</p><p><strong>Building legitimacy for our work on value alignment</strong>: Obviously, we need to partner with external experts to meaningfully explore this work. The more transparent we are about these collaborations, processes and outcomes, the more we can build credibility and trust with external experts and people.</p><p>For the purposes of this note, I will focus on #1: normative questions on value alignment.</p><h3>Ways in which AI systems and metrics could fail</h3><p>We understand something has to change fundamentally with the way we incentivize content, connections and behaviors on our platform. We have a good understanding of the current set of problems with recommender systems/metrics, have a reasonable set of approaches to measure and tackle them and longer term plans to continue pushing the boundary to ensure we are net good for people and society. In addition to these efforts, we need to look closer at a set of root causes for these really hard individual and societal problems.</p><p>What if we make the case that the problems are not due to the individual metrics (although they can be continuously evolved to be “better”) but the design of our AI systems?</p><p>We have looked at the big problems we face that cause harm to people and society and if we look through a more general recommender system lens, we find that most of the problems occur because of one of the following ways in which the recommender system acted differently from what we’d have expected:</p><ol><li><strong>Negative side effects</strong>: How do we ensure that our AI systems don’t do something unintended in order to optimize for its metric? E.g. unintentionally show people who tend to see conspiracy theories more related content in order to optimize for engagement.</li><li><strong>Reward hacking</strong>: How can we ensure our AI systems don’t hack the reward to optimize for it? A toy example is, if we optimize to resurrect churned users, the AI system could do things to make users churn more so it could resurrect them and improve that metric?</li><li><strong>Scalable oversight</strong>: There are areas where the AI system would have to ask humans for input because it is uncertain. How can it do this in a scaleable manner? A good example is semi-supervised learning techniques that query humans when it’s uncertain.</li><li><strong>Safe exploration</strong>: How can we ensure the AI system is allowed to explore ways to learn more but such that that doesn’t have really bad repercussions?</li><li><strong>Robustness in different environments</strong>: How can we ensure AI systems behave in robust and safe ways in environments different from its training environment? For e.g. in a safe test environment, an AI system could perform in ways that seem “good”. A completely made-up example is a small experiment optimizing for say “good comments” which ends up promoting the healthy, back and forth, meaningful comments between people but also ends up promoting inflammatory, divisive content when rolled out to the entire population.</li></ol><p>To a certain extent, no matter what behavioral or even survey metric we pick, we can expect to run into these<a href="https://arxiv.org/abs/1606.06565"> types of problems unless we proactively identify ways to mitigate these risks</a>. What if instead, we changed the question to one of values and principles?</p><h3>Human values in AI systems</h3><p>What if our AI systems could be encoded with human values and principles instead of specific, fixed objectives? Obviously, this is incredibly hard to design and implement but I think these are the types of thought experiments and hard questions we should be asking now to work towards a long term, fundamental solution to the problems of current recommender systems. So, let’s try to think through the question of human values in the context of AI.</p><p>The goal of AI should be to ensure that powerful AI is properly aligned with human values, intent, aspirations and preferences but each of these terms means a slightly different set of things. AI could be built to optimize for:</p><ol><li><strong>My instructions</strong>: i.e. do what I tell you to do. The problem with this approach is that the <em>proxy </em>might not be completely aligned with our goals and doing what I instruct the AI to do may not be in my best, longer term interests!</li><li><strong>My expressed intentions</strong>: i.e. do what I intend for you to do. It is unclear if current AI systems are able to completely grasp human preferences and values to understand intentions. It is also possible for intentions to be irrational or misinformed.</li><li><strong>Revealed preferences</strong>: i.e. do what my behavior reveals I prefer. It is very hard to make reliable inferences from observed behavior and to model preferences for situations that are rarely observed and people may have preferences for things that harm them.</li><li><strong>Informed preferences or desires</strong>: i.e. do what I would do if I were rational or informed. Prioritizing informed preferences does little to address self-harm or unethical preferences and there is a lack of universal consensus on conceptions of human rationality and reason.</li><li><strong>Interest or well-being</strong>: i.e. do what is in my interest or best for me, objectively speaking. Although there is disagreement about the nature of well-being, the scope of this disagreement is relatively narrow but it does raise questions about whether we should do what is in the interest of a specific individual and what the trade offs are for interests of different people. We need principles to decide whose interests or needs count (more) for AI alignment? This approach may also overlook other values such as value for the environment.</li><li><strong>Values/Moral principles</strong>: i.e. do what we morally ought to do, as defined by individual or society. The advantage of this approach is that we avoid having to choose between expressed intentions, preferences and interests. Second, moral principles tend to include considerations for the group such as justice and rights. Third, this approach could incorporate things overlooked by other approaches such as value of the intrinsic world, welfare of animals, of people not yet born and so on. However, this means, we need to specify what values or principles AI should align with and the body of people who select the principles for AI to align with. While there is no consensus on the fundamental nature of moral principles, people may still come to an agreement about values and standards that are appropriate to incorporate and this<a href="https://deepmind.com/research/publications/2020/Artificial-Intelligence-Values-and-Alignment"> paper shares some of those approaches.</a></li><li>There is also a lot of external literature on building<a href="https://intelligence.org/files/CEV.pdf"> friendly and benevolent AIs</a> that act in our best interests.</li></ol><h3>Looking ahead</h3><p>I have laid out some of the problems with current recommender systems, shared a few details about our current approaches such as surveys to capture richer feedback from people to evolve our recommender systems to better outcomes, defined AI value alignment, some normative aspects of AI value alignment and shared some reasons to explore this area deeply. <em>I intentionally weave in and out of more abstract, broader issues and, short term, tactical approaches to this problem because we sometimes dwell too much on the higher level definitions and problems or on the other extreme, approaching a problem with a very short focus and perhaps this note helps bridge the gap a tiny bit between the two ends</em>. We often categorize problems as those causing harm to individuals and those cause societal harms and I am beginning to be convinced that AI value alignment can help tackle both. This note is definitely not comprehensive and definitely, does not do justice to capturing the rich literature and research that exists in this space but I wanted to share a glimpse in the hopes of sparking meaningful conversations around this topic. I do believe this aspirational end state is possible and that our AI systems can do net good for people and society but only if we start building the long term, proactive foundation for it today:</p><p><em>Recommender systems could support democratic values in a variety of ways and encourage well-being, personal self-actualizing and human flourishing.</em></p><p>Disclaimers: Opinions in this note represent only my personal take on this topic. I spent 6 months being obsessed about AI human value alignment and this note is a result of the dozens of academic papers and books I have read on this topic. Material here is built by weaving all those learnings and where possible, I have tried to link to the source material but might be missing a few.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1553de1a0ce" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>