Evaluating language models as risk scores

AF Cruz, M Hardt, C Mendler-Dünner - arXiv preprint arXiv:2407.14614, 2024 - arxiv.org
Current question-answering benchmarks predominantly focus on accuracy in realizable
prediction tasks. Conditioned on a question and answer-key, does the most likely token …

Allocation Requires Prediction Only if Inequality Is Low

A Shirali, R Abebe, M Hardt - arXiv preprint arXiv:2406.13882, 2024 - arxiv.org
Algorithmic predictions are emerging as a promising solution concept for efficiently
allocating societal resources. Fueling their use is an underlying assumption that such …

Learning treatment effects while treating those in need

B Wilder, P Welle - arXiv preprint arXiv:2407.07596, 2024 - arxiv.org
Many social programs attempt to allocate scarce resources to people with the greatest need.
Indeed, public services increasingly use algorithmic risk assessments motivated by this goal …