The limits of "banning discrimination through Artificial Intelligence"


In April 2018, the “Finnish Non-Discrimination and Equality Tribunal” prohibited “discriminatory use” of artificial intelligence. I am not sure they did the right thing.

The limits of "banning discrimination through Artificial Intelligence" /img/ai-discrimination.jpg

This is a summary of what happened, as reported by Daily Finland

  • The Finnish Non-Discrimination and Equality Tribunal has prohibited “discriminatory use” of artificial intelligence in decisions about financial credit
  • because “credit had been denied on statistical grounds based on the applicant’s gender, language, age and residential area”
  • so the the tribunal deemed that the financial institution had done “discriminating statistical profiling” and ordered not to use statistical methods “in a discriminating way”

Why ban discrimination “THROUGH Artificial Intelligence”???

Discrimination is wrong. But laws and courts trying to catch up with specific technologies, instead of caring about up to date formulations of general principles, may be equally wrong, or as a minimum really inefficient and counterproductive. In other words:

Why should any court forbid “discrimination through Artificial Intelligence”, instead of discrimination, period?

This is the core issue. Either that article is really bad reporting (at least the title and first bullet above), or that Finnish decision seems just another case of legal and regulatory institutions trying to catch up with technology fads, as it is happening with, e.g. blockchain and smart contracts.

Of course, fighting discrimination as the same cost of fighting junk food, or many other bad things: “In its response, the credit institution noted that a review of the personal situation of the applicant would have increased the cost of the loan process.”

Stop at Zona-M   Never miss a story: follow me on Twitter (@mfioretti_en), or via RSS