I used to work at Facebook, YouTube, and Twitter. One of the big questions I focused on: what was the right objective function to align our AI systems towards?<p>When we started optimizing for watch time at YouTube, for example, our algorithms started suggesting longer videos for the sake of longer videos, and videos with racy thumbnails.<p>Similarly, optimizing for engagement at Facebook led to low-quality clickbait, and Hooter's appearing as the top search result when you searched for restaurants in Houston.<p>Experiments that increased favorites and replies at Twitter invariably increased toxic content as well.<p>So while watch time, engagement, and replies would always go up -- were these really the products we wanted to build? What happened to Facebook's original mission of connecting users with their friends and family? What did "favorites" have to do with being <i>the</i> platform for public conversation at Twitter? A lot of work at these companies is spent measuring active users, but where were the dashboards measuring progress to these broader goals? It's easy to become blinded by standard metrics and lose sight of the original product principles that made us stand out -- and I say this as a data scientist at heart!<p>So could we figure out a metric that was better tuned to human values and the product mission we cared about, but also fast, rigorous, and easily measurable? After all, we still need our A/B tests, ML objective functions, and OKRs! This question is particularly important today, with all the troubles that social media platforms face, and I wrote up an approach that I've often worked on.