4.7: Momin Malik
- Page ID
- 98089
Postdoctoral Data Science Research Fellow Berkman Klein Center for Internet & Society, Harvard University https://www.mominmalik.com/
A core function of humanities and social science education is to reveal and interrogate the categories, concepts, and logics by which we make sense of and act in the world. Machine learning is reifying one specific logic: that of statistical correlations, with no consideration of causality or meaning, a point lost amidst the focus on the “algorithms” that calculate and apply these correlations. The task of higher education will now have to engage with the limits of this logic. To quote William Bruce Cameron on how “not everything that can be counted counts, and not everything that counts can be counted,” the old maxim that “correlation does not imply causation,” and George E. P. Box on how “all models are wrong but some are useful,” we can consider when, how, for what, and to whom it is useful to see the world only through correlations between measurable quantities.
For example, “creditworthiness” is an abstract idea, neither the same as future loan repayment, nor of past repayment behavior: we can easily imagine a financially responsible person whose circumstances would nevertheless prevent repayment and are correlated with others as “defaulting.” Should that person be deemed creditworthy or not? Using machine learning to decide forces an answer: The only thing that matters is what correlates with aggregate past behavior. Intention, effort, individuality, and circumstances do not matter, since they do not exist in data. Nor do causal relationships, as studied in econometrics, nor measurement validity, as studied by psychometrics (both statistical fields with their own narrowness and legacy of racism and inequality), nor the possibility of systems being gamed, insofar as they limit the complexity of correlations that can be considered.
Articulating and challenging the frames of “algorithmic” systems will encourage deliberation on the sorts of uses we will accept, and what we will organize to reject.