Teddy Bear + Bayesian Filter = Programming Oracle

Of the bucket load of books I have to read during the holidays, I started reading Pair Programming Illuminated.

One of the Seven Synergiestic Behaviours is "Pair Debugging", that the act of explaining your code/problem to something else will cause you to think about it in a way that will reveal alternatives. I'm all of us have experienced that at some time.

At a helpdesk they even had a Teddy Bear that students had to explain their mysterious bug too before talking to a human. This probably started as the sysadmins trying to get out of some work, although the bear was probably just as productive

So what if we could use a Bayesian Filter to answer the question instead of the typical scenario of identifying some text as spam. Would we end up with a programmer version of an oracle/ask jeeves that could respond in a way that would aid the consumer?

Would this mean sayings like, "If I had a dollar for everytime someone asked me " would become obsolete?

I assume the problem is the control, how do you feedback that the advice was incorrect or inefficient? Or how do you teach it that it was inefficient and now it should respond in this way?

How can other consumers ignore other consumers evaluations/posts? Most user based forums have ratings of members, admittedly don't rate the quality of posts, but they do rank the quantity.

Could this be the future for help systems, the pre-dominance of answers matched to bayesian filtered questions based on user forum knowledge posts over pre-packaged help text?

I'll have to ask ash his opinion.