From a living room somewhere in suburban Melbourne, Australia, a history scholar who wanted to reach beyond an academic audience is producing some of the world’s finest satire.
The reason I’m bringing all this up is because one of the videos deals with the horrible consequences of turning over the process of who supposedly is a welfare cheat to – an algorithm. It was similar to the latest ugly, failed fad in American child welfare – using predictive analytics to predict who supposedly is likely to abuse a child.
In a process nicknamed “robodebt” the Australian algorithm supposedly could figure out who was collecting too much. The algorithm then sent out threatening letters – backed up by, apparently, the only humans in the process: debt collectors.
But the algorithm was wrong at least 20 percent of the time, wreaking havoc in impoverished Australian families. But let the AustraliEn [sic] Government explain:
If that was a little heavy on Australian acronyms and slang, this story from New Statesman has a good explanation for an international audience.
Last year, an Australian federal court ruled the whole thing illegal. Clearly a 20 percent false positive rate was too outrageous to allow.
And surely no agency intervening in the lives of the vulnerable anywhere in the world would accept, let alone brag about, an algorithm with an error rate of up to 99.8 percent.
Or would they?