The Selective Labels Problem: Evaluating Algorithmic Predictions in the Presence of Unobservables

By Himabindu Lakkaraju, Jon M. Kleinberg, Jure Leskovec, Jens Ludwig, Sendhil Mullainathan. 2017

180 citations recorded by the source corpus.

Evaluating whether machines improve on human performance is one of the central questions of machine learning. However, there are many domains where the data is selectively labeled, in the sense that the observed outcomes are themselves a consequence of the existing choices of the human decision-makers. For instance, in the context of judicial bail decisions, we observe the outcome of whether a defendant fails to return for their court appearance only if the human judge decides to release the defendant on bail. This selective labeling makes it harder to evaluate predictive models as the instances for which outcomes are observed do not represent a random sample of the population. Here we propose a novel framework for evaluating the performance of predictive models on selectively labeled data. We develop an approach called contraction which allows us to compare the performance of predictive models and human decision-makers without resorting to counterfactual inference. Our methodology harnesses the heterogeneity of human decision-makers and facilitates effective evaluation of predictive models even in the presence of unmeasured confounders (unobservables) which influence both human decisions and the resulting outcomes. Experimental results on real world datasets spanning diverse domains such as health care, insurance, and criminal justice demonstrate the utility of our evaluation metric in comparing human decisions and machine predictions.

Subfields: Artificial Intelligence, Statistics and Probability, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

In Evaluation & Benchmarks

Citations in the corpus

Characterizing Fairness Over the Set of Good Models Under Selective Labels (2021, background): “[15] Himabindu Lakkaraju, Jon Kleinberg, Jure Leskovec, Jens Ludwig, and Sendhil Mullainathan.”
Residual Unfairness in Fair Machine Learning from Prejudiced Data (2018, background): “Lakkaraju et al. (2017) identify a similar structural setting with “selective labels” where they learn a decision rule for pre-trial risk assessment from the decisions made from judges (which affect whether or not the outcome of interest will be observed).”
Fairness through Causal Awareness: Learning Causal Latent-Variable Models for Biased Data (2018, background): “Some conceptually similar work has looked at the “selective labels” problem [9, 29], where only a biased selection of the data has labels available.”
Fair Decisions Despite Imperfect Predictions (2019, background): “This problem has been referred to by Lakkaraju et al. (2017) as selective labels .”
Algorithmic Fairness: Choices, Assumptions, and Definitions (2021, background): “This problem is sometimes called selective labels (Lakkaraju et al.2017).”
Learning under selective labels in the presence of expert consistency (2018, background): “ndence to: Maria De-Arteaga <mdeartea@andrew.cmu.edu>. FAT/ML ’18, Stockholm, Sweden, July 2018. This problem has been recently formalized as the selective labels problem (Kleinberg et al.,2017;Lakkaraju et al., 2017). Under this setting, if machine learning algorithms are trained using the ob”