Ask HN: Has anyone tried making a virtual person to be on social networks? | Hacker News Reader

Ask HN: Has anyone tried making a virtual person to be on social networks?

8 pointsthedangler13 comments

Has anyone written a bot that pretends to be a real person on facebook? Adding friends adding comments? I think this would be an interesting project.

View on HN

Very early on in the days when ICQ still had an unencrypted client that allowed you to connect to a random ICQ user that was online I played around with a very simple human backed bot that would open a bunch of connections and if it ran out of stuff to say would IM the operator with a bit of context. The reply was then saved so that the next time around when it would get in to that situation it would have a reply waiting.

It was lots of fun doing that, and I can imagine doing this on facebook would be just as much fun.

I made a couple on twitter... based on a hacked up AIML engine. This one is the best of the bunch though. http://twitter.com/coolestalison

@coolestalison

It was especially entertaining to see people trying to carry conversations with her, some even flirting. It wasn't perfect, but managed to fool quite a bit of people. I hadn't been running her for a while, maybe I should bring her back to life again.

Robin Sage at the 2010 Blackhat conference: http://www.washingtontimes.com/news/2010/jul/18/fictitious-f...

I saw this presentation, and despite the minor mistakes made, it was an incredibly successful project. Good to note that none of this was automated, and latency was one of the mistakes.

I was part of a team for Yahoo's Hackday in London a few years back.

We worked on a project that took as inspiration a photo montage research thing that stated that the average of a stack of photos is the most attractive: http://www.uni-regensburg.de/Fakultaeten/phil_Fak_II/Psychol...

For the hack day project we attempted to take a target user of a social network, and then to look at the friends to derive properties that we would then include (diluted) in our fake user.

The aim of the hack day experiment was to see whether the fake profile would both be convincing enough and attractive enough to then be befriended by the network of people that the target was a member of.

The experiment failed in the time given as hackday didn't last long enough and our target didn't log on. But I ran the code again later and it worked well (being befriended a few times before I figured that it was beyond the experiment and should be killed).

* (being befriended a few times before I figured that it was beyond the experiment and should be killed).*

These guys started hitting on you didn't they?

Yes. I know a few people who have done it with so-call "social media experts" using a Markov bot that's trained on a handful of known SMDs. Scary part, most of them garnered over a thousand followers. :-/

but how many of those are other bots?

thousands of Markov bot followers ;)

Seems like spammers have done this (convincingly enough for at least some users). I don't have some foolproof method from telling a spambot from a spammer using copy/paste everywhere, but I certainly wouldn't be surprised.

Been there, done that. Although, not on facebook, but a similar social network site, a good few years ago.

It was, indeed, an interesting experiment.

Not on facebook, but there are tons on Twitter.

Any ideas why this would be interesting (or useful)?

Reply on news.ycombinator.com