(Machine) Learning What Policies Value

  title={(Machine) Learning What Policies Value},
  author={Daniel Bj{\"o}rkegren and Joshua Evan Blumenstock and Samsun Knight},
When a policy prioritizes one person over another, is it because they benefit more, or because they are preferred? This paper develops a method to uncover the values consistent with observed allocation decisions. We use machine learning methods to estimate how much each individual benefits from an intervention, and then reconcile its allocation with (i) the welfare weights assigned to different people; (ii) heterogeneous treatment effects of the intervention; and (iii) weights on different outcomes… 

Figures and Tables from this paper



The Contribution of Increased Equity to the Estimated Social Benefits from a Transfer Program: An Illustration from Progresa/Oportunidades

Most impact evaluations of Conditional Cash Transfers (CCTs) and Unconditional Cash Transfers (UCTs) focus on the returns to increased human capital investments that will be reaped largely or

Proceedings of the 3rd Innovations in Theoretical Computer Science Conference

The papers in this volume were presented at the 3rd Innovations in Theoretical Computer Science (ITCS 2012) conference, sponsored by the ACM Special Interest Group on Algorithms and Computation

Targeting the Poor: Evidence from a Field Experiment in Indonesia

An experiment in 640 Indonesian villages on three approaches to target the poor: proxy-means tests (PMT), where assets are used to predict consumption; community targeting; and a hybrid, which performs somewhat worse in identifying the poor than PMT.

Targeting Impact versus Deprivation

Using Machine Learning and Qualitative Interviews to Design a Five-Question Women's Agency Index

A new method to design a short survey measure of a complex concept such as women's agency, which combines mixed-methods data collection and machine learning and applies statistical algorithms that build on LASSO and random forest but constrain how many variables are selected for the model.

Policy Learning With Observational Data

Given a doubly robust estimator of the causal effect of assigning everyone to treatment, an algorithm for choosing whom to treat is developed, and strong guarantees for the asymptotic utilitarian regret of the resulting policy are established.

Fairness, Equality, and Power in Algorithmic Decision-Making

This work argues that leading notions of fairness suffer from three key limitations: they legitimize inequalities justified by "merit;" they are narrowly bracketed, considering only differences of treatment within the algorithm; and they consider between-group and not within-group differences.

The Optimal Allocation of Resources among Heterogeneous Individuals

An aid agency has CES preference aggregation over individual outcomes and allocates W units of discrete or bounded-continuous resources among N candidate recipients and solutions are characterized by resource-invariant optimal allocation queues that sequence the order in which individuals begin and stop to receive allocations.

Balancing Competing Objectives with Noisy Data: Score-Based Classifiers for Welfare-Aware Machine Learning

The theoretical results characterize the optimal strategies in this class of policies, bound the Pareto errors due to inaccuracies in the scores, and show an equivalence between optimal strategies and a rich class of fairness-constrained profit-maximizing policies.