4 ms·
As an ex-FB, which has a pretty similar culture, I'm apparently picking today to halfheartedly stand up to these grave injustices against company culture. My in
by CGamesPlay 5y ago
As an ex-FB, which has a pretty similar culture, I'm apparently picking today to halfheartedly stand up to these grave injustices against company culture. My input here is mostly just to humanize the people who stand on the other side of the robotic wall that is preventing our OP from posting videos of his son, but understanding the problem a little better may help resolve it in the end.
> * They're content with algorithmic approaches to spam prevention and other moderation that results in numerous false positives and don't see the problem with that.
Spot on. False positives are part of the game when you're trying to improve on $1.3MM revenue/employee.
> * They think their support channels are more than adequate.
Anybody at any internal discussion would probably disagree with this, instead saying something along the lines of "there may be deficiencies with our support process but we can always improve them later on."
> * They think their bans result solely in bad actors being harmed and don't realize that shady businesses that actually do spam have endless fake companies at their disposal to keep on trucking.
Nobody thinks that the bans solely result in bad actors being harmed and the fact that there are endless fake companies is what necessitates these robot walls that don't have human connections. False positives will happen and a goal is minimizing them... however goal #1 is growing the product (which indirectly requires keeping false positives from getting too high) and goal #2 is preventing company liability (which directly requires that false negatives stay really low). When you have strong pressure to keep false negatives low and less pressure to keep false positives low, then you end up with a bias toward positive results.
OK, well, that's why we end up here. Is this helpful to OP? Probably not. Sorry, OP!
- xg15 5y ago> Spot on. False positives are part of the game when you're trying to improve on $1.3MM revenue/employee. Algorithmic moderation will always have false positives. The question is how you deal with them: Do you have an appeals process that involves human review or do you just say "sucks to be you" and let the false positive be the user's problem? In a similar way, car accidents will always happen. However the car industry decided (not completely on their own) to invest in safety so that when accidents do happen, they are less deadly. > Anybody at any internal discussion would probably disagree with this, instead saying something along the lines of "there may be deficiencies with our support process but we can always improve them later on." The mythical time of "later on". When exactly would that be? And yes, you can always improve anything. This is a vaguely pleasant-sounding non-statement.
- ethbr0 5y agoThe false positives at scale problem isn't something that could be solved by an appeals process, because the problem is a step above that: a lack of continuous process improvement. Why was the false positive generated? Why was the user unable to use the process in place to appeal? And then improve the root process cause. This is what seems to be broken at Google and Facebook, and is the fundamental difference between a black hole support model (make problems invisible to us, because they never get to us) and one that works at scale. You're always going to have false positives, but I don't think anyone feels like the mega-corps actually have a process in place to monitor and improve the process for dealing with them.
- rixed 5y agoHow do you design the appeal process to ensure it's not going to be spammed?