The latest issue of First Monday carries an article by Daniel J. Cohen and Roy Rosenzweig which shows that on at least basic historical facts, the internet can provide a surprisingly accurate indication of the historical consensus...
the H-Bot does remarkably well. When tested on the 33 multiple-choice questions from the fourth-grade NAEP American history exam, it got 27 answers right, a respectable 82 per cent. As the authors note, "the average fourth grader does far worse". And "...its successors are likely to reach 95 or even 99 percent."
...It surely won't be long before an economics professor or grad student with a software bent develops the Econ-bot and takes economic teaching down the same path
I wonder how these bots deal with the issue Bryan recently raised of expert opinion vs. naive opinion. Perhaps there is selection bias in that non-experts will tend not to post their opinions on the web.
But my guess is we could see web crawlers with statistical algorithms try to solve problems like predicting economic indicators.