I remember the first time I saw a computer, it was a Power Macintosh 5260 (with Monkey Island on it). I was around 5 years old and I looked at it as if it belonged to another universe. It did, I was not allowed to get anywhere close to it within a 5 mile radius; it was my older brother’s! That did not stop me. I browsed it for hours. The possibilities of computers were infinite and fuelled by the inspiration of sci-fi worlds the dream of talking machines, machines that can assist humans, think themselves and even have feelings never stopped. I kept dreaming about the possibilities of the future.
One of the questions that arises when designing models that learn to solve multiple tasks simultaneously is how much of the available training budget should be devoted to each individual task. We refer to any formalized approach to addressing this problem (learned or otherwise) as a task selection policy. In this work we provide an empirical evaluation of the performance of some common task selection policies in a synthetic bandit-style setting, as well as on the GLUE benchmark for natural language understanding. We connect task selection policy learning to existing work on automated curriculum learning and off-policy evaluation, and suggest a method based on counterfactual estimation that leads to improved model performance in our experimental settings.
The use of machine learning systems to support decision making in healthcare raises questions as to what extent these systems may introduce or exacerbate disparities in care for historically underrepresented and mistreated groups, due to biases implicitly embedded in observational data in electronic health records. To address this problem in the context of clinical risk prediction models, we develop an augmented counterfactual fairness criteria to extend the group fairness criteria of equalized odds to an individual level. We do so by requiring that the same prediction be made for a patient, and a counterfactual patient resulting from changing a sensitive attribute, if the factual and counterfactual outcomes do not differ. We investigate the extent to which the augmented counterfactual fairness criteria may be applied to develop fair models for prolonged inpatient length of stay and mortality with observational electronic health records data. As the fairness criteria is ill-defined without knowledge of the data generating process, we use a variational autoencoder to perform counterfactual inference in the context of an assumed causal graph. While our technique provides a means to trade off maintenance of fairness with reduction in predictive performance in the context of a learned generative model, further work is needed to assess the generality of this approach.
Learning to Rank (LTR) from user interactions is challenging as user feedback often contains high levels of bias and noise. At the moment, two methodologies for dealing with bias prevail in the field of LTR: counterfactual methods that learn from historical data and model user behavior to deal with biases; and online methods that perform interventions to deal with bias but use no explicit user models. For practitioners the decision between either methodology is very important because of its direct impact on end users. Nevertheless, there has never been a direct comparison between these two approaches to unbiased LTR. In this study we provide the first benchmarking of both counterfactual and online LTR methods under different experimental conditions. Our results show that the choice between the methodologies is consequential and depends on the presence of selection bias, and the degree of position bias and interaction noise. In settings with little bias or noise counterfactual methods can obtain the highest ranking performance; however, in other circumstances their optimization can be detrimental to the user experience. Conversely, online methods are very robust to bias and noise but require control over the displayed rankings. Our findings confirm and contradict existing expectations on the impact of model-based and intervention-based methods in LTR, and allow practitioners to make an informed decision between the two methodologies.
We offer a graphical interpretation of unfairness in a dataset as the presence of an unfair causal path in the causal Bayesian network representing the data-generation mechanism. We use this viewpoint to revisit the recent debate surrounding the COMPAS pretrial risk assessment tool and, more generally, to point out that fairness evaluation on a model requires careful considerations on the patterns of unfairness underlying the training data. We show that causal Bayesian networks provide us with a powerful tool to measure unfairness in a dataset and to design fair models in complex unfairness scenarios.
Marginal structural models (MSM) with inverse probability weighting (IPW) are used to estimate causal effects of time-varying treatments, but can result in erratic finite-sample performance when there is low overlap in covariate distributions across different treatment patterns. Modifications to IPW which target the average treatment effect (ATE) estimand either introduce bias or rely on unverifiable parametric assumptions and extrapolation. This paper extends an alternate estimand, the average treatment effect on the overlap population (ATO) which is estimated on a sub-population with a reasonable probability of receiving alternate treatment patterns in time-varying treatment settings. To estimate the ATO within a MSM framework, this paper extends a stochastic pruning method based on the posterior predictive treatment assignment (PPTA) as well as a weighting analogue to the time-varying treatment setting. Simulations demonstrate the performance of these extensions compared against IPW and stabilized weighting with regard to bias, efficiency and coverage. Finally, an analysis using these methods is performed on Medicare beneficiaries residing across 18,480 zip codes in the U.S. to evaluate the effect of coal-fired power plant emissions exposure on ischemic heart disease hospitalization, accounting for seasonal patterns that lead to change in treatment over time.