Although I don't believe computers are smart enough to detect human pedophiles, I bet it wouldn't be too hard to develop a system that could reliably discern between chatbots and humans. (Cameron Marlow suggested to me in an email that it would be valuable for CS students to consider how they're able to tell the difference themselves -- what cues they notice and how they do their classification.)
With the current state of the natural language processing art, I think the classification problem is pretty trivial. A computer system could be fed samples of transcripts with human and robot participants, and little more than a statistical analysis would be necessary to categorize the differences. After all, computers can already tell the difference between male and female authors with 80% accuracy.
Just for fun, go try the gender genie with some text written by you (or your favorite author). Pasting in the front page of my blog reveals that I'm male!
Words: 8388(NOTE: The genie works best on texts of more than 500 words.)
Female Score: 10987
Male Score: 17405The Gender Genie thinks the author of this passage is: male!
Whew!
That gender genie keeps telling me that I blog like a man. But when I run my fiction through it, it says it's written by a woman. I don't know if that means I am myself when I'm making stuff up or when I'm writing about reality.
The chatbot/human classification problem is an interesting one, and one that as far as I'm aware has not received any attention from the NLP/ML crowd.
In terms of discriminative features, I think you've already mentioned sentence length in a comment about the NannieBots. More telling, I think, would be the percentage of sentences for which a parser can produce complete parses for. I would expect this number to be very high (approaching 100%) for chatbots, since their sentences are (usually) constructed using a grammar, and to be markedly lower for humans (especially in an informal situation, where parse accuracy is traditionally very low).
Many of Christ's passages (albeit they are translations) come out female, particularly when he is reasoning with the disciples or speaking of spiritual matters. Other passages come out distinctly male. I wonder if anyone has run all of Christ's words through.
JT: I'm not sure how much it matters, considering that JC wasn't a native English speaker! Who knows how Hebrew men and women 2000 years ago spoke? I doubt we have much female writing to compare it to.
Also, the statistical analysis is based on writing, not speaking. Speech patterns are quire different from writing patterns.