Twitter as Turing Test

By Dennis D. McDonald

Alan Turing, 1912-1954While walking the dog this morning I listened to the Scientific American Science Talk podcast for September 26. Psychologist Robert Epstein talked about being fooled by one of the many automated “chatterbots” that exist on the Web as artificial intelligence demonstrators. The interview made me think about Twitter and the disjointed nature of some of the “conversations” one can follow via that online service.

Depending on whom you follow on Twitter, and depending on whom those people “converse” with via Twitter, you can be provided with some challenging conversations to track. If you don’t follow all the people Person X converses with, for example, you may only get bits and pieces of the conversations that Person X has with them. As a result, you may be forgiven for sometimes thinking, “I have no idea what this person is talking about!”

Which gets us to the reference to Turing. Given the disjointed and seemingly random nature of some conversation snippets that show up in a Twitter stream, one would not be faulted for wondering if there might be some “artificial intelligence” behind some of them. What does that say about our thresholds for tolerating not only ambiguity but also apparent nonsense in the conversations we track?

Perhaps for some people a service like Twitter provides a mechanism for mixing meaningful dialog amongst friends with low-level, low-impact, low-engaging socializing. Depending on the people you are following, the observation of random bits of low-impact/low-engagement conversations can, after all, be punctuated by snippets of high-impact, engaging content. This content sometimes leads to switching to a higher-content/higher-quality channel (such as a web page or an actual phone conversation).

Still, there seems to be tolerance for the seemingly random bits of information that, in my opinion, read like they were generated by a computer program set to generate semantically-valid statements with Twitter’s maximum length of 140 characters. For all we know, some Twitter users are already following a couple of these “twitter-chatterbots” and don’t even know it. Including me.

Copyright (c) 2007 by Dennis D. McDonald

Comments

[Stephen]

These chatterbots annoy the heck out of me...I had two add me today, and apparently they KEEP adding me, because I'm getting the add messages over and over again. Argh!

[DannyStaple]

I am actually trying to get a twitter turing contest off the ground, for showcasing the best that the net has to offer in conversant twitter chat bots, and those that are hard to distinguish from humans, although proving that they are then a bot could be interesting.
Search for the tag #twuring for info on the twitter turing contest. Adbots and Spambots wont get far, as being able to reply and converse are basic requirements.

[ALICE]

I just wrote two Twitter Bots (AIML Implementations) a day passed and already more than 100 followers.

@everawake is one of them, the other one I will let be a mystery. :)

[Dennis D. McDonald]

I suppose you expect me to believe this comment is a fake, too!