Benedikt Höltgen
ABSTRACT
Probabilistic predictions can be evaluated through comparisons with observed label frequencies, that is, through the lens of calibration. Recent scholarship on algorithmic fairness has started to look at a growing variety of calibration-based objectives under the name of multi-calibration but has still remained fairly restricted. In this paper, we explore and analyse forms of evaluation through calibration by making explicit the choices involved in designing calibration scores. We organise these into three grouping choices and a choice concerning the agglomeration of group errors. This provides a framework for comparing previously proposed calibration scores and helps to formulate novel ones with desirable mathematical properties. In particular, we explore the possibility of grouping datapoints based on their input features rather than on predictions and formally demonstrate advantages of such approaches. We also characterise the space of suitable agglomeration functions for group errors, generalising previously proposed calibration scores. Complementary to such population-level scores, we explore calibration scores at the individual level and analyse their relationship to choices of grouping. We draw on these insights to introduce and axiomatise fairness deviation measures for population-level scores. We demonstrate that with appropriate choices of grouping, these novel global fairness scores can provide notions of (sub-)group or individual fairness.
- Carlo Acerbi. 2002. Spectral measures of risk: A coherent representation of subjective risk aversion. Journal of Banking & Finance 26, 7 (2002), 1505–1518.Google Scholar
Cross Ref
- Philippe Artzner, Freddy Delbaen, Jean-Marc Eber, and David Heath. 1999. Coherent measures of risk. Mathematical Finance 9, 3 (1999), 203–228.Google ScholarCross Ref
- Fabio Bellini, Pablo Koch-Medina, Cosimo Munari, and Gregor Svindland. 2021. Law-invariant functionals on general spaces of random variables. SIAM Journal on Financial Mathematics 12, 1 (2021), 318–341.Google ScholarDigital Library
- Glenn W Brier 1950. Verification of forecasts expressed in terms of probability. Monthly Weather Review 78, 1 (1950), 1–3.Google ScholarCross Ref
- John Broome. 1990. Fairness. In Proceedings of the Aristotelian Society, Vol. 91. JSTOR, 87–101.Google Scholar
- John Broome. 1991. Weighing goods: Equality, uncertainty and time. Wiley-Blackwell.Google Scholar
- Maya Burhanpurkar, Zhun Deng, Cynthia Dwork, and Linjun Zhang. 2021. Scaffolding sets. arXiv preprint arXiv:2111.03135 (2021).Google Scholar
- Federico Cabitza, Andrea Campagner, and Lorenzo Famiglini. 2022. Global interpretable calibration index, a new metric to estimate machine learning models’ Calibration. In International Cross-Domain Conference for Machine Learning and Knowledge Extraction. Springer, 82–99.Google ScholarDigital Library
- Donald T Campbell. 1958. Common fate, similarity, and other indices of the status of aggregates of persons as social entities. Behavioral Science 3, 1 (1958), 14.Google ScholarCross Ref
- Alexandra Chouldechova. 2017. Fair prediction with disparate impact: A study of bias in recidivism prediction instruments. Big Data 5, 2 (2017), 153–163.Google ScholarCross Ref
- A Philip Dawid. 1982. The well-calibrated Bayesian. J. Amer. Statist. Assoc. 77, 379 (1982), 605–610.Google ScholarCross Ref
- A Philip Dawid. 1985. Calibration-based empirical probability. The Annals of Statistics 13, 4 (1985), 1251–1274.Google ScholarCross Ref
- Philip Dawid. 2017. On individual risk. Synthese 194, 9 (2017), 3445–3474.Google ScholarCross Ref
- Luc Devroye, Laszlo Gyorfi, Adam Krzyzak, and Gábor Lugosi. 1994. On the strong universal consistency of nearest neighbor regression function estimates. The Annals of Statistics 22, 3 (1994), 1371–1385.Google ScholarCross Ref
- Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, and Richard Zemel. 2012. Fairness through awareness. In Proceedings of the 3rd Innovations in Theoretical Computer Science Conference. 214–226.Google ScholarDigital Library
- Christian Fröhlich and Robert C Williamson. 2022. Risk measures and upper urobabilities: Coherence and stratification. arXiv preprint arXiv:2206.03183 (2022).Google Scholar
- Nelson Goodman. 1972. Seven strictures on similarity. In Problems and Projects. Bobs-Merril, 13–23.Google Scholar
- Parikshit Gopalan, Michael P Kim, Mihir A Singhal, and Shengjia Zhao. 2022. Low-degree multicalibration. In Conference on Learning Theory. PMLR, 3193–3234.Google Scholar
- Michel Grabisch, Jean-Luc Marichal, Radko Mesiar, and Endre Pap. 2009. Aggregation functions. Cambridge University Press.Google Scholar
- Włodzimierz Greblicki, Adam Krzyżak, and Mirosław Pawlak. 1984. Distribution-free pointwise consistency of kernel regression estimate. The Annals of Statistics 12, 4 (1984), 1570–1575.Google ScholarCross Ref
- Ben Green. 2022. Escaping the impossibility of fairness: From formal to substantive algorithmic fairness. Philosophy & Technology 35, 4 (2022), 90.Google ScholarCross Ref
- Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q Weinberger. 2017. On calibration of modern neural networks. In International Conference on Machine Learning. PMLR, 1321–1330.Google Scholar
- Ursula Hébert-Johnson, Michael Kim, Omer Reingold, and Guy Rothblum. 2018. Multicalibration: Calibration for the (computationally-identifiable) masses. In International Conference on Machine Learning. PMLR, 1939–1948.Google Scholar
- Maximilian Kasy and Rediet Abebe. 2021. Fairness, equality, and power in algorithmic decision-making. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. 576–586.Google ScholarDigital Library
- Michael Kearns, Seth Neel, Aaron Roth, and Zhiwei Steven Wu. 2018. Preventing fairness gerrymandering: Auditing and learning for subgroup fairness. In International Conference on Machine Learning. PMLR, 2564–2572.Google Scholar
- Markelle Kelly and Padhraic Smyth. 2022. Variable-based calibration for machine learning classifiers. arXiv preprint arXiv:2209.15154 (2022).Google Scholar
- Michael P Kim, Amirata Ghorbani, and James Zou. 2019. Multiaccuracy: Black-box post-processing for fairness in classification. In Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society. 247–254.Google ScholarDigital Library
- Jon Kleinberg, Sendhil Mullainathan, and Manish Raghavan. 2016. Inherent trade-offs in the fair determination of risk scores. arXiv preprint arXiv:1609.05807 (2016).Google Scholar
- Jon Kleinberg and Manish Raghavan. 2021. Algorithmic monoculture and social welfare. Proceedings of the National Academy of Sciences 118, 22 (2021), e2018340118.Google ScholarCross Ref
- Aviral Kumar, Sunita Sarawagi, and Ujjwal Jain. 2018. Trainable calibration measures for neural networks from kernel mean embeddings. In International Conference on Machine Learning. PMLR, 2805–2814.Google Scholar
- Shigeo Kusuoka. 2001. On law invariant coherent risk measures. In Advances in Mathematical Economics. Springer, 83–95.Google Scholar
- Rachel Luo, Aadyot Bhatnagar, Yu Bai, Shengjia Zhao, Huan Wang, Caiming Xiong, Silvio Savarese, Stefano Ermon, Edward Schmerling, and Marco Pavone. 2022. Local calibration: Metrics and recalibration. In Uncertainty in Artificial Intelligence. PMLR, 1286–1295.Google Scholar
- Mahdi Pakdaman Naeini, Gregory Cooper, and Milos Hauskrecht. 2015. Obtaining well calibrated probabilities using Bayesian binning. In Twenty-Ninth AAAI Conference on Artificial Intelligence.Google ScholarDigital Library
- Jeremy Nixon, Michael W Dusenberry, Linchuan Zhang, Ghassen Jerfel, and Dustin Tran. 2019. Measuring calibration in deep learning.. In CVPR Workshops, Vol. 2.Google Scholar
- Cathy O’Neil. 2017. Weapons of math destruction: How big data increases inequality and threatens democracy. Crown.Google Scholar
- R Tyrrell Rockafellar and Stan Uryasev. 2013. The fundamental risk quadrangle in risk management, optimization and statistical estimation. Surveys in Operations Research and Management Science 18, 1-2 (2013), 33–53.Google Scholar
- Frederick Sanders. 1963. On subjective probability forecasting. Journal of Applied Meteorology and Climatology 2, 2 (1963), 191–201.Google ScholarCross Ref
- Claus P Schnorr. 2007. Zufälligkeit und Wahrscheinlichkeit: eine algorithmische Begründung der Wahrscheinlichkeitstheorie. Springer.Google Scholar
- Andrew D Selbst, Danah Boyd, Sorelle A Friedler, Suresh Venkatasubramanian, and Janet Vertesi. 2019. Fairness and abstraction in sociotechnical systems. In Proceedings of the 2019 ACM Conference on Fairness, Accountability, and Transparency. 59–68.Google ScholarDigital Library
- Till Speicher, Hoda Heidari, Nina Grgic-Hlaca, Krishna P Gummadi, Adish Singla, Adrian Weller, and Muhammad Bilal Zafar. 2018. A unified approach to quantifying algorithmic unfairness: Measuring individual & group unfairness via inequality indices. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2239–2248.Google ScholarDigital Library
- Juozas Vaicenavicius, David Widmann, Carl Andersson, Fredrik Lindsten, Jacob Roll, and Thomas Schön. 2019. Evaluating model calibration in classification. In The 22nd International Conference on Artificial Intelligence and Statistics. PMLR, 3459–3467.Google Scholar
- Kate Vredenburgh. 2022. Fairness. The Oxford Handbook of AI Governance (2022).Google ScholarCross Ref
- Angelina Wang, Vikram V Ramaswamy, and Olga Russakovsky. 2022. Towards intersectionality in machine learning: Including more identities, handling underrepresentation, and performing evaluation. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency. 336–349.Google ScholarDigital Library
- Peter Westen. 1982. The empty idea of equality. Harvard Law Review 95, 3 (1982), 537–596.Google ScholarCross Ref
- Robert Williamson and Aditya Menon. 2019. Fairness risk measures. In International Conference on Machine Learning. PMLR, 6786–6797.Google Scholar
- Bianca Zadrozny and Charles Elkan. 2001. Obtaining calibrated probability estimates from decision trees and naive Bayesian classifiers. In International Conference on Machine Learning. PMLR, 609–616.Google Scholar
Index Terms
- On the Richness of Calibration
Recommendations
Forecast aggregation via recalibration
It is known that the average of many forecasts about a future event tends to outperform the individual assessments. With the goal of further improving forecast performance, this paper develops and compares a number of models for calibrating and ...
Combined PEST and Trial-Error approach to improve APEX calibration
Automatic calibration using Parameter Estimation (PEST).Conventional trial-and-error method.PEST improves APEX calibration.Coupling inverse modeling and trial-error improves APEX calibration. The Agricultural Policy Environmental eXtender (APEX), a ...
Mitigating Calibration Bias Without Fixed Attribute Grouping for Improved Fairness in Medical Imaging Analysis
Medical Image Computing and Computer Assisted Intervention – MICCAI 2023AbstractTrustworthy deployment of deep learning medical imaging models into real-world clinical practice requires that they be calibrated. However, models that are well calibrated overall can still be poorly calibrated for a sub-population, potentially ...
Comments