As technology advances and computers become increasingly capable, the line between human and bot activity on social media platforms like Lemmy is becoming blurred.
What are your thoughts on this matter? How do you think social media platforms, particularly Lemmy, should handle advanced bots in the future?
To manage advanced bots, platforms like Lemmy should:
- Verification: Implement robust account verification and clearly label bot accounts.
- Behavioral Analysis: Use algorithms to identify bot-like behavior.
- User Reporting: Enable easy reporting of suspected bots by users.
- Rate Limiting: Limit posting frequency to reduce spam.
- Content Moderation: Enhance tools to detect and manage bot-generated content.
- User Education: Provide resources to help users recognize bots.
- Adaptive Policies: Regularly update policies to counter evolving bot tactics.
These strategies can help maintain a healthier online community.
Verification: Implement robust account verification and clearly label bot accounts.
☑ Clear label for bot accounts
☑ 3 different levels of captcha verification (I use the intermediary level in my instance and rarely deal with any bot)Behavioral Analysis: Use algorithms to identify bot-like behavior.
Profiling algorithms seems like something people are running away when they choose fediverse platforms, this kind of solution have to be very well thought and communicated.
User Reporting: Enable easy reporting of suspected bots by users.
☑ Reporting in lemmy is just as easy as anywhere else.
Rate Limiting: Limit posting frequency to reduce spam.
☑ Like this?
image
Content Moderation: Enhance tools to detect and manage bot-generated content.
What do you suggest other than profiling accounts?
User Education: Provide resources to help users recognize bots.
This is not up to Lemmy development team.
Adaptive Policies: Regularly update policies to counter evolving bot tactics.
Idem.
We’re not handling the LLM generative bullshit bots now, anywhere. There’s a thing called the dead Internet theory. Essentially most of the traffic on the Internet now is bots.
It’s not just the internet. For example, students are handing in essays straight from ChatGPT. Uni scanners flag it and the students may fail. But there is no good evidence either side, the uni side detection is unreliable (and unlikely to improve on false positives, or negatives for that matter) and it’s hard for the student to prove they did not use an LLM. Job seekers send in LLM generated letters. Consultants probably give LLM based reports to clients. We’re doomed.
Hardly. Just do away with coursework and stick to in-person exams and orals.
Not even the biggest tech companies have an answer sadly… There are bots everywhere and social media is failing to stop them. The only reason there aren’t more bots in the Fediverse is because we’re not a big enough target for them to care (though we do have occasional bot spam).
I guess the plan is to wait until there’s an actual way to detect bots and deal with them.
Not even the biggest tech companies have an answer sadly…
They do have an answer: add friction. Add paywalls, require proof of identity, start using client-signed certificates which needs to be validated by a trusted party, etc.
Their problem is that these answers affect their bottom line.
I think (hope?) we actually get to the point where bots become so ubiquitous that the whole internet will become some type of Dark Forest and people will be forced to learn how to deal with technology properly.
Their problem is that these answers affect their bottom line.
It’s more complicated than that. Adding friction and paywalls will quickly kill their userbase, requiring a proof of identity or tracking users is a privacy disaster and I’m sure many people (especially here) would outright refuse to give IDs to companies.
They’re more like a compromise than a real solution. Even then, they’re probably not foolproof and bots will still manage.
requiring a proof of identity or tracking users is a privacy disaster and I’m sure many people (especially here) would outright refuse to give IDs to companies.
The Blockchain/web3/Cypherpunk crowd already developed solutions for that. ZK-proofs allow you to confirm one’s identity without having to reveal it to public and make it impossible to correlate with other proofs.
Add other things like reputation-based systems based on Web-Of-Trust, and we can go a long way to get rid of bots, or at least make them as harmless as email spam is nowadays.
I saw a comment the other day saying that “the line between the most advanced bot and the least talkative human is getting more and more thinner”
Which made me think: what if bots are setup to pretend to be actual users? With a fake life that they could talk about, fake anecdotes, fake hobbies, fake jokes but everything would seem legit and consistent. That would be pretty weird, but probably impossible to detect.
And then when that roleplaying bot once in a while recommends a product, you would probably trust them, after all they gave you advice for your cat last week.
Not sure what to do in that scenario, really
I’ve just accepted that if a bot interaction has the same impact on me as someone who is making up a fictional backstory, I’m not really worried wheter it is a bot or not. A bot shilling for Musk or a person shilling for Musk because they bought the hype are basically the same thing.
In my opinion the main problem with bots is not individual acccounts pretending to be people, but the damage they can do en masse through a firehose of spam posts, comments, and manipulating engagement mechanics like up/down votes. At that point there is no need for an individual account to be convincing because it is lost in the sea of trash.
Even more problematic are entire communities made out of astroturfing bots. This kind of stuff is increasingly easy and cheap to set up and will fool most people looking for advise online.
I think smarter people than me will have to figure it out and even then it’s going to be a war of escalation. Ban the bots, build better bots, back and forth back and forth.
Some news sites had an interesting take on comments sections. Before you could comment on an article, you had to correctly answer a 5 question quiz proving you actually read it.
But AI can do that now too.
Some news sites had an interesting take on comments sections. Before you could comment on an article, you had to correctly answer a 5 question quiz proving you actually read it.
It would be interesting to try that on Lemmy for a day. People would probably not be happy.
