Okay here’s an opinion. Markov bots should have a CW on every toot they make, that reads “output of markov bot” or similar. Because when they get boosted onto my timeline with no CW I’m left stressing out trying to work out why I don’t comprehend what this person is saying, until eventually I work out it was a bot and I wasted that energy because it was meaningless. Can y’all see why an autistic person would find text that looks like real language but doesn’t make sense a trigger and a drain of energy? And that it’s a valid position to not want to waste my time reading words recycled by an unthinking computer script even if other people find them hilarious?
Even putting “bot” or “ebooks” in the display name isn’t good enough because people do that too as a joke! But it’s more ableist than it is funny.
In addition, you can’t completely prevent a bot from saying something triggering, bigoted, or otherwise deserving its own CW because contextual meaning is more complex than a simple statistical analysis can process. Even on a G-rated corpus there will be potential outputs that are racist, antisemitic, transphobic, violent, horrible, etc etc. I don’t want to see those!
@s0 I agree with this. When I'm in an episode it's hard enough nailing my train of thought to reality. Every time I have one I have to just leave social media in general for a while or shit goes sideways immediately. For the record it's not autism for me.
@s0 I tend to be pretty paranoid, and when I see one of these, I wonder if I'm having a stroke or hallucinating or something
@MadestMadness yeah!! Should have said in the original, it’s not just autistic folks! There’s lots of reasons to get stressed/triggered by it!
@s0 once upon a time, basically all the markov bots on fedi did follow this rule, surprisingly
a few influxes ago though someone who doesn't care about this joined and started flooding fedi with bots that don't cw :/
@s0 Tusky allows you to display a robot icon on bots avatars ; it might be a feature that should be implemented on Mastodon web client, and all other clients ?
@gabriel it was merged in and then reverted 3 days later in 2018 on mainline mastodon. Fuck knows eugen’s reasoning. I opened a feature request on Toot! today for a similar feature not knowing tusky did that.
@s0 Fully agree, but hard to enforce. I recognize the sense of wasted effort you describe to find the person behind random words. I also believe some people may have a talent for spotting robots, although I'm not sure this talent would correlate with autism positively. The advancement has been quick and gpt-3 is now simply too good for this claim. I don't believe either of us could reliably spot its output today.
@s0 I have literally blocked the work e b o o k s because these bots annoy me so much. I literally wish I could filter every user with that in their username, but I haven't found a way to do that yet. I don't want to stop other people from seeing them since clearly some people like them, but yeah: really hard for me to process, and very rarely funny to me when I can. Just... weird for weird sake? Occasionally accidentally apt? Meh.
@s0 ah! i have felt this way too, but couldn't articulate it!! they're sometimes so confusing and upsetting, it's so exhausting
(though i'm not autistic, i have adhd)
@s0 I feel that. I'd be OK with a tag to the same effect if it were at the start of the toot, too, but SOMETHING to alert people that the following toot is random generated, yes, please
Seconded. Markov bots in particular are not a huge issue for me (and frankly, there are some human-made toots which I find hard to believe are not from Markov bots...), but even for mildly-affected or neurotypical people, it's important to know whether you're wasting time talking to a bot (Markov or otherwise) or an actual human trying to communicate, who may be either too upset to type straight or just having a hard time with a foreign language.
...but how could such a rule be enforced?