Cheating Algorithms

No Child Left Behind
A plan created to make schools more accountable to parent, NCLB has faced criticism as more of a folly of rewarding A while hoping for B. The plan seeks to reward schools who show progress and penalize schools who fail. But progress is measured in numbers, and these numbers measure other numbers which measure how well school children are learning their core curriculum. See any weaknesses in that chain?

Critics call for the repeal of NCLB, and we can argue until we're blue in the face about the validity of standardized testing, but an article I found shows that we can strengthen the plan by using Decision Modeling in the area of cheating. The article is called "To Catch a Cheat," and it proposes that cheaters, whether they be teachers or students, have a hard time faking out statistics. Questions on a test are deemed easier or harder, and harder questions can be correlated with other harder questions via a correct answer coefficient. This means a student who answers a hard question correctly is less likely to answer an easier question incorrectly, and on average, this trend should be dominant. But a company called Caveon Test Security can recognize when things ain't right, especially when a teacher is leading a classroom of cheaters.

Ronald Reagan's signature phrase "Trust but Verify" would come in handy here, but as time passes, the trust fraction could fizzle away to nothing as the verify fraction becomes more and more capable.
-----