Striving for data-model efficiency: Identifying data externalities on group performanceDownload PDF

Published: 21 Nov 2022, Last Modified: 05 May 2023TSRML2022Readers: Everyone
Keywords: data-focused learning, data externalities, evaluation
TL;DR: We study a phenomenon in which adding training data from certain data sources can decrease performance on key groups of the target population, toward understanding how algorithms might make more efficient and effective use of the data available.
Abstract: Building trustworthy, effective, and responsible machine learning systems hinges on understanding how differences in training data and modeling decisions interact to impact predictive performance. In this work, we seek to better understand how we might characterize, detect, and design for data-model synergies. We focus on a particular type of data-model inefficiency, in which adding training data from some sources can actually lower performance evaluated on key sub-groups of the population, a phenomenon we refer to as negative data externalities on group performance. Such externalities can arise in standard learning settings and can manifest differently depending on conditions between training set size and model size. Data externalities directly imply a lower bound on feasible model improvements, yet improving models efficiently requires understanding the underlying data-model tensions. From a broader perspective, our results indicate that data-efficiency is a key component of both accurate and trustworthy machine learning
3 Replies