Thread by @samidh

Was hoping for a quiet day but @JeffHorwitz strikes again. Do I have thoughts on the issues raised? You bet! I share in the spirit of trying to enhance understanding of these complex dilemmas. In short, we need to imbue feeds with a sense of morality. wsj.com/articles/faceb…

When you treat all engagement equally (irrespective of content), increasing feed engagement will invariably amplify misinfo, sensationalism, hate, and other societal harms. I wish this weren't the case, but it is so predictable that it is perhaps a natural law of social networks.

So it is no surprise that the MSI (meaningful social interaction) ranking changes of 2018/2019 had this impact, and as the reporting shows, many people at FB are conscious of and concerned about these side effects.

Yet as is often the case in FB's product culture, the ethos of "launch early, fix later" dominates. As the article highlights, many critical fixes were indeed subsequently launched to try to cancel out these negative effects, and the teams working on them deserve huge credit.

In particular, as the company previously announced, interventions like reducing the ranking scores for articles that have long chains of reshares, or ranking civic content using different objective functions such as value to users, can be highly effective.

Still even these come at a cost to engagement, so you are still left with the fundamental problem of how much "good" engagement is worth to you versus how much "bad" engagement is palatable.

This is not merely an empirical problem, but also a more philosophical one. It requires you to have a point of view of what is "good" and what is "bad". If you think that there can be wide consensus on how to define either of those, I have a bridge to sell you.

That is why leadership truly matters and platforms need to have the courage to explain to the world what they consider bad AND what is good. Then they need to build out the metrics, set a threshold for these tradeoffs, and hold all their product launches accountable.

In the absence of an articulated set of values, engagement & growth concerns will win every single time because they are far easier to measure (and defend). But without them, we are left with social networks that are inherently a-moral yet control our information sphere.

There continues to be an impressive pace of innovation at FB on measuring "good" and "bad". The Community Standards Enforcement Reports show progress in quantifying things like misinformation. And prior research talks highlight efforts on quantifying affective polarization.

Ultimately it is going to require all of society to be involved in this conversation. What morals should we demand of our platforms? How should we tradeoff X views of sensationalist news with Y units of baby pics? FB needs to take the lead, but can't and shouldn't do it alone.