Perhaps it’s even disingenuous. Roland Fryer, of Harvard University, wants to bring algorithms into college/university admissions offices under the claim that it would help reduce endless, unproductive battles over race- and class-based affirmative action.

Of course, they would do no such thing. Algorithms are nothing more than software packages written by humans as tools to speed decision-making in narrowly defined areas in accordance with equally narrowly defined criteria. Inescapably, though, those decision-making software packages and their defined criteria have embedded within them the personal biases and world views of those human programmers, software testers, and their supervisors. The databases on which those algorithms are trained also are written by humans and so themselves contain human biases and world views. On a (slightly) higher level, there’s another layer of human bias and world view: it’s humans who select the databases to be used for algorithm training.

The presence of these biases and world views is inescapable and wholly independent of the intentions of the humans involved. Even those with the purest of intentions and the strongest efforts to control the impact of their biases and world views will have those leak into their algorithms and training databases. Even efforts to balance to net zero those biases and world views by using a political and social cross-section of humans in those production efforts are subject to the same shortcomings via the selection of those humans and the definitions of what constitutes a suitable cross-section.

That all of this still could lead to strong bias in the algorithms, despite the best of intentions, is amply demonstrated by Alphabet’s overtly racist Google Gemini AI package (which Alphabet claims has been corrected, but only after the fact, not before the AI was released and in use), Robby Starbuck’s suit against Meta over that company’s AI smearing him as involved with the J6 riot and an extremist organization, and Mark Walter’s suit against OpenAI over its ChatGPT smear accusing him of embezzlement.

Colleges and universities, despite Supreme Court rulings banning consideration of race in their admissions decisions, still do that under other names and rationalizations. Using these human-written and so too-likely biased algorithms in their admissions decisions would merely give those institutions another way to disguise their admissions misbehaviors.