Beyond Isolation: Towards an Interactionist Perspective on Human Cognitive Bias and AI Bias

TL;DR

This paper argues that human cognitive bias and AI bias have been studied in isolation, missing critical interaction effects that can amplify or mitigate biases in AI-assisted decision-making. It advocates an interactionist framework to map compound biases to targeted mitigation strategies, including extending the cognitive bias codex to AI contexts and conducting empirical evaluations. The work outlines concrete steps for theory development, technique design, and framework integration, aiming to produce design guidelines that align AI outputs with human cognitive processes and reduce bias amplification. By addressing the reciprocal dynamics of human and AI biases, the paper seeks to safeguard human cognition and enhance decision quality in real-world AI interactions.

Abstract

Isolated perspectives have often paved the way for great scientific discoveries. However, many breakthroughs only emerged when moving away from singular views towards interactions. Discussions on Artificial Intelligence (AI) typically treat human and AI bias as distinct challenges, leaving their dynamic interplay and compounding potential largely unexplored. Recent research suggests that biased AI can amplify human cognitive biases, while well-calibrated systems might help mitigate them. In this position paper, I advocate for transcending beyond separate treatment of human and AI biases and instead focus on their interaction effects. I argue that a comprehensive framework, one that maps (compound human-AI) biases to mitigation strategies, is essential for understanding and protecting human cognition, and I outline concrete steps for its development.