The Problem with Metrics is a Fundamental Problem for AI

TL;DR

This paper highlights the fundamental issues with overreliance on metrics in AI, demonstrating how it leads to negative consequences and proposing a comprehensive framework to mitigate these harms.

Contribution

It introduces a new framework that combines multiple metrics, qualitative insights, and stakeholder involvement to address the problems caused by metric overemphasis in AI.

Findings

Metrics can be manipulated and lead to short-term focus.

Current practices exacerbate metric-related failures.

A multi-faceted framework can reduce harms from metric overuse.

Abstract

Optimizing a given metric is a central aspect of most current AI approaches, yet overemphasizing metrics leads to manipulation, gaming, a myopic focus on short-term goals, and other unexpected negative consequences. This poses a fundamental contradiction for AI development. Through a series of real-world case studies, we look at various aspects of where metrics go wrong in practice and aspects of how our online environment and current business practices are exacerbating these failures. Finally, we propose a framework towards mitigating the harms caused by overemphasis of metrics within AI by: (1) using a slate of metrics to get a fuller and more nuanced picture, (2) combining metrics with qualitative accounts, and (3) involving a range of stakeholders, including those who will be most impacted.