Arvind Narayanan

For years I’ve felt there needs to be an accessible primer on social media recommendation algorithms. So I used my sabbatical to write one! We can level up normative/policy debates about social media if the tech is better understood.

In the essay I also discuss the flaws of engagement optimization and the optimization mindset in general. I have strong opinions on this and I don’t mince words. I’ve clearly separated the normative analysis from the expository parts.
https://knightcolumbia.org/content/understanding-social-media-recommendation-algorithms

Understanding Social Media Recommendation Algorithms | Knight First Amendment Institute

Mar 9, 2023 at 8:22pmWeb
Show threadArvind NarayananMar 9, 2023
I introduce three stylized models of information propagation (subscription, network, algorithmic) and discuss how virality operates. Turning to algorithmic recommendations, I give a bit of history, an overview of the main modern ideas, and a case study of Facebook's Meaningful Social Interaction metric. But the major platforms’ algorithms are far more similar than they are different — they’re all flavors of engagement optimization — a point I previously made about TikTok. https://knightcolumbia.org/blog/tiktoks-secret-sauce
TikTok’s Secret Sauce

Show threadArvind NarayananMar 9, 2023
I plan to publish two follow-up essays in a few months. I’m super grateful to
@knightcolumbia for the opportunity — this type of writing is hard to do because the traditional paper publication route isn’t available. I enjoyed writing this and I hope you enjoy reading it!
Show threadVirginia MurrMar 9, 2023

@randomwalker

Kudos for having the patience to write it (well done).

Also, this is perfectly stated:
"As a result, I argue that social media platforms are weakening institutions by undermining their quality standards and making them less trustworthy. While this has been widely observed in the case of news, my claim is that every other institution is being affected, even if not to the same degree."

Show thread𝙺𝚊𝚛𝚕 DEFEND DEMOCRACY 𝙱𝚞𝚗𝚍𝚊𝚢 🗽Mar 9, 2023
@randomwalker Thank you. This is extremely important work, especially as new networks develop while Twitter is under stress.
Show threadRichard ReismanMar 9, 2023
@randomwalker
A very important contribution to understanding -- explaining these are complex systems but the principles are not that complex.
Show threadTim KelloggMar 11, 2023
@randomwalker this is awesome. i tried writing a similar essay several times, only to realize how complex it is, and how far i am from understanding it. looking forward to reading this tonight
Show threadDavid SlifkaMar 11, 2023

@randomwalker Thank you for creating this great resource.

Are you aware of anyone who is (or might want to be) working on implementing new algorithms for Mastodon?

Show threadMike MacgirvinMar 11, 2023
We built a number of algorithms for streams and gave them all to our members. The only thing the admin needs to do is moderate the public stream (but we also let you disable that completely since it's usually just a cesspool of spam and bullshit); and most of our sites are small sites for family and friends who don't want the legal liability of unwashed public content that can land them in jail.
Show threadNathan CarpenterMar 13, 2023
@randomwalker excellent work! Thank you for this- I wish it had come out sooner so I could assigned it as a reading in my class this semester.
Show threadTulio TecJul 10, 2023
@randomwalker, I'm reading your paper and I became curious about the fact that you don't mention the use of biometric data by recommendation systems. Did you find anything that could prove that big tech companies are using eye tracking and other biometric data to optimize for engagement?
Show threadArvind NarayananJul 10, 2023

@tuliotec With modern mobile OSes, surreptitious eye tracking is not technically possible. It's also a legal risk that IMO outweighs the benefits.

But face analysis has been used for recommendations https://twitter.com/MarcFaddoul/status/1232014908536938498

Other biometrics like gait and activity recognition are also used, but not for recommendations AFAIK.

Use of location data is ubiquitous, of course.

Hope that helps!

Marc Faddoul on Twitter

“A TikTok novelty: FACE-BASED FITLER BUBBLES The AI-bias techlash seems to have had no impact on newer platforms. Follow a random profile, and TikTok will only recommend people who look almost the same. Let’s do the experiment from a fresh account: 1/6”

Twitter