Position: AI Evaluation Should Work With Humans

Jan Kulveit, Gavin Leech, Tomáš Gavenčiak, Raymond Douglas
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:171140-171149, 2026.

Abstract

This position paper argues that the dominant paradigm of AI evaluation (which focuses on superhuman autonomous performance and so implicitly targets the goal of replacing humans) is guiding AI development in the wrong direction. Instead, the AI community should pivot to evaluating the performance of human–AI teams. We specify roughly what this would mean and argue that this collaborative shift will foster AI systems that act as true complements to human capabilities and therefore lead to far better societal outcomes than will the current process.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-kulveit26a, title = {Position: {AI} Evaluation Should Work With Humans}, author = {Kulveit, Jan and Leech, Gavin and Gaven\v{c}iak, Tom\'{a}\v{s} and Douglas, Raymond}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {171140--171149}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/kulveit26a/kulveit26a.pdf}, url = {https://proceedings.mlr.press/v306/kulveit26a.html}, abstract = {This position paper argues that the dominant paradigm of AI evaluation (which focuses on superhuman autonomous performance and so implicitly targets the goal of replacing humans) is guiding AI development in the wrong direction. Instead, the AI community should pivot to evaluating the performance of human–AI teams. We specify roughly what this would mean and argue that this collaborative shift will foster AI systems that act as true complements to human capabilities and therefore lead to far better societal outcomes than will the current process.} }
Endnote
%0 Conference Paper %T Position: AI Evaluation Should Work With Humans %A Jan Kulveit %A Gavin Leech %A Tomáš Gavenčiak %A Raymond Douglas %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-kulveit26a %I PMLR %P 171140--171149 %U https://proceedings.mlr.press/v306/kulveit26a.html %V 306 %X This position paper argues that the dominant paradigm of AI evaluation (which focuses on superhuman autonomous performance and so implicitly targets the goal of replacing humans) is guiding AI development in the wrong direction. Instead, the AI community should pivot to evaluating the performance of human–AI teams. We specify roughly what this would mean and argue that this collaborative shift will foster AI systems that act as true complements to human capabilities and therefore lead to far better societal outcomes than will the current process.
APA
Kulveit, J., Leech, G., Gavenčiak, T. & Douglas, R.. (2026). Position: AI Evaluation Should Work With Humans. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:171140-171149 Available from https://proceedings.mlr.press/v306/kulveit26a.html.

Related Material