/i/2004735554.png?fit=%2C&ssl=1)
Nevertheless, reconstruction reveals that the algorithm was not created by statistical evaluation in any respect, however on account of a committee that began to theorize a bit about what “alerts” of fraud might be. Comparable to “dwelling with household”, or “being an MBO”. (Two issues which are extra widespread amongst out-of-home college students with a migration background).
Subsequently, the algorithm has by no means been validated. DUO doesn’t have e.g. despatched out a number of units of inspectors, one primarily based on random sampling, one primarily based on the algorithm, after which outcomes in contrast.
The explanation it got here to mild is as a result of there was an enormous share (one thing like 25%) of accused college students who defended themselves, both straight or via the courts, once they challenged DUO. Word: that is with *reversed* burden of proof! As an accused it’s essential to show your innocence, not DUO your fraud.
(If you happen to learn how the house visits went, there may be additionally one thing to be stated in regards to the high quality of the inspectors, however that’s one other story).
Briefly, in the event that they have been certainly onerous info for this algorithm, then we might have talked about one thing. The issue right here is that, identical to with the Supplementary Affair, individuals needed to point out decisiveness after which recognized a weak group.
*10,000 allegations, of which 6,000 appealed, of which 1,200 have been upheld, and one other 1,500 lawsuits with 375x winnings. So 25% of the objectors have been proper and 16% of the completely accused. These are alarming numbers!
[Reactie gewijzigd door Keypunchie op 5 juli 2023 22:37]
