3 ms·

Nice response. Unfortunately, the original response is dismissive even though it’s about a technical topic, but it wouldn’t be if discussing just about any othe

Nice response. Unfortunately, the original response is dismissive even though it’s about a technical topic, but it wouldn’t be if discussing just about any other content category and you’re right that those kinds of mechanisms are worthy of analysis. I haven’t worked at FB but don’t metadata-adding contractors training ML models serve the function of invisible, more complete likes? My understanding is all these large social networks employ or employed human-created content analysis. And then the right amount of noise and divergent suggestions can be added to serve the goals of the network.

bertil 5y ago
The issue rarely is whether the content can be interpreted properly, or at least hadn’t in the last few years. There are contractors flagging edge-cases, issues around nuances like photo-realistic first-person shooter video game footage (common) vs. footages from actual fire-fights (banned since the Christchurch attack) so it _is_ and interesting technical question but that’s not what is challenging to improve the experience. Some people find others talking about their success inspiring, others find the bargain obnoxious, others still feel demoralised by it. Those can be the same viewer at different time. Without a better way to let people express their reaction than Likes, Comments and dwell time, you won’t be able to tell what’s the long-term impact. “Blocking Facebook” is the only external solution, and getting no-signal isn’t going to help Facebook improve. Every idea under the sun was suggested internally, but whether they are given a chance is a delicate question.