Which type of AI shapes our online experiences the most? Probably recommendation algorithms, which are used to generate our social media feeds by predicting what we’re most likely to engage with. This logic helps us find relevant information, but has also been blamed for stoking division and making us addicted to our phones.

We haven’t talked about recommendation algorithms in this newsletter, but they are the focus of my project this year at the Knight First Amendment Institute at Columbia University, where I’m a visiting senior research scientist. Recommendation algorithms inevitably amplify some speech and suppress others, and we urgently need to understand the consequences so that we can improve our relationship to social media and algorithm designers can shape the tech for the better.

To this end, I’m co-organizing a symposium on April 28/29 (in NYC and online): Optimizing for What? Algorithmic Amplification and Society. We’ve been fortunate to be able to assemble a lineup of about 20 speakers who are leading experts on amplification, and I believe this is the first symposium on the topic. The talks and panels will be based on the papers listed here, and the papers will be published online after the symposium. You can register for the event here.

For many years I’ve felt there needs to be an accessible primer on social media recommendation algorithms, so I used my time at the Knight Institute to write one. Contrary to the myth, recommendation algorithms are fairly straightforward to understand, and I hope this essay will help. Normative and policy debates about social media should be grounded in the technical details.

Here’s the essay: Understanding Social Media Recommendation Algorithms.