3 ms·

Meanwhile, a friend was posting suicide notes on Facebook, and Facebook did nothing about it. She was already in a hospital, but for some reason she had access

by mgbmtl 4y ago

Meanwhile, a friend was posting suicide notes on Facebook, and Facebook did nothing about it. She was already in a hospital, but for some reason she had access to her phone for a while. She also posted a fake death announcement by her mom. Then attempted suicide, and was in ICU for a few days. People were engaging a lot, freaking out. So yeah, Facebook fails miserably at everything.

ev1 4y ago
Does Facebook really consider this a failure if it increased clicks and engagement? Something to think about.
- exikyut 4y ago
  Hmm. Ostensibly Facebook has a party line and attendant menagerie of moderation and AI tooling as a broad response to self-harm, but at the end of the day all of that has its box to work within, and thinking and breathing this one subject does not define Facebook's identity and core DNA. And that may be the core of the problem. I wonder if the auto-assessment ML model passed right on by the described post *precisely because* the "clicks and engagement" ranked it as "good", where "good", intentionally or otherwise, has wound up becoming the TensorFlow distillation of "not our problem"... because *that* is where the focal point, the emergent foundation everything else finds itself orbiting around, at the end of the day. From the angle I'm squinting at this from, this situation seems unsustainable intractable. FB is attempting to "benignly police" N billion people at once in an effort to monetize a private (non-government, publicly-traded), commercially-delivered digital town square experience. IMHO there is no global point of cohesion that can encapsulate this effort, which is trying to unify multiple fundamentally incompatible dimensions at once. It doesn't work.
  - ev1 4y ago
    A real death notice removed as targeted harassment would also probably result in some form of public relations nightmare. It's not like the ML model will know about the severe and difficult to reverse reputational damage of a fake harassment death notice.

by mgbmtl 4y ago

ev1 4y ago

Does Facebook really consider this a failure if it increased clicks and engagement? Something to think about.

exikyut 4y ago
Hmm. Ostensibly Facebook has a party line and attendant menagerie of moderation and AI tooling as a broad response to self-harm, but at the end of the day all of that has its box to work within, and thinking and breathing this one subject does not define Facebook's identity and core DNA. And that may be the core of the problem. I wonder if the auto-assessment ML model passed right on by the described post *precisely because* the "clicks and engagement" ranked it as "good", where "good", intentionally or otherwise, has wound up becoming the TensorFlow distillation of "not our problem"... because *that* is where the focal point, the emergent foundation everything else finds itself orbiting around, at the end of the day. From the angle I'm squinting at this from, this situation seems unsustainable intractable. FB is attempting to "benignly police" N billion people at once in an effort to monetize a private (non-government, publicly-traded), commercially-delivered digital town square experience. IMHO there is no global point of cohesion that can encapsulate this effort, which is trying to unify multiple fundamentally incompatible dimensions at once. It doesn't work.
- ev1 4y ago
  A real death notice removed as targeted harassment would also probably result in some form of public relations nightmare. It's not like the ML model will know about the severe and difficult to reverse reputational damage of a fake harassment death notice.