10 comments

[ 2.9 ms ] story [ 30.7 ms ] thread
>"If these issues are not resolved quickly they risk fuelling conspiracy theories and damaging Meta's reputation."

It's now a "conspiracy theory" to believe your eyes. Lord end me now.

Examples of pairs of keywords and their effects:

- #dnc (no results) vs #rnc (normal results)

- #voteblue (no results) vs #votered (normal results)

- #fuckbiden (normal results) vs #fucktrump (no results)

Most likely explanation is whatever algorithm change they pushed on 20/Jan to boost Trump-aligned posts and bury Trump unaligned ones was accidentally tuned too aggressively and became too obvious. Please accept our apologies, we will be rectifying the issue and fixing the Algorithm so the manipulation of public opinion is properly hidden, as intended.

(comment deleted)
> Most likely explanation is whatever algorithm change they pushed on 20/Jan to boost Trump-aligned posts and bury Trump unaligned ones was accidentally tuned too aggressively and became too obvious.

I wonder if this was an intentional act by a disgruntled employee to make the scheme public. "Oops - I accidentally set the 'punish political enemies' value to 100" etc.

Yes, malicious compliance seems a very likely explanation.
I guess malicious compliance also comes with less of a risk of sudden death than whistle-blowing.
This is such a naked attempt at manufacturing consent. Really bizarre, 1984-level remove-dissent-from-the-vocabulary kind of play.
(comment deleted)
The sad thing is, I feel like Meta has used up so much trust that I wouldn't even entirely dismiss the possibility that this was a deliberate thing to show a reduction in certain metrics on certain topics to curry favor.

Betting odds are this is just an embarrassing bug. But maybe there's a lesson here about the value of earning a reputation of trustworthiness.