Language, Epistemology, Metaphysics, and Mind Research Interest Group (Brian Hedden, MIT)

The Language, Epistemology, Metaphysics, and Mind Research Interest Group is delighted to welcome as guest speaker Brian Hedden, a professor in the Department of Linguistics and Philosophy at MIT. He works in epistemology and decision theory, as well as related areas of ethics and political philosophy.

Talk Title

Algorithmic Monoculture

Talk Abstract

Algorithmic decision-making is replacing idiosyncratic human judgment in domains such as hiring, lending, and college admissions. But what happens if all firms use the same algorithm to make their hiring decisions? Or if all banks use the same algorithm to assess creditworthiness? Or if all colleges use the same algorithm to decide which students to admit? Recently, several philosophers, legal scholars, and computer scientists have raised the alarm about such algorithmic monoculture, in which all decisions across a domain (all hiring in a given sector, say) are made by a single algorithm. They warn of the rise of “algorithmic leviathans” and “outcome homogenization,” which may “institutionalize systemic exclusion and reinscribe social hierarchy.” I argue that their concerns are largely misplaced. I consider four objections to monoculture, having to do with systematic exclusion, information aggregation, exploration of unfamiliar types of candidates, and Goodhart’s law, and find them unpersuasive. Monoculture needn’t be dystopian, and depending on the nature of its lone algorithm, it could even be better than polyculture, in which different decision-makers use different algorithms.