Effects of Generative AI Errors on User Reliance Across Task Difficulty

2026-04-05

AI Summary

This study investigates how users’ trust and reliance on generative AI shift when the system makes errors across tasks of varying difficulty, with a focus on the implications of AI’s “jagged” capability boundaries. Through a preregistered 3×2 between-subjects experiment in a chart-generation context, the research systematically manipulates task difficulty and AI error rates, complemented by an incentive-compatible design and quantitative behavioral analysis. Results indicate that although higher error rates generally reduce user reliance, errors in simple tasks do not significantly erode trust more than those in difficult tasks. This finding reveals that users exhibit greater tolerance for AI’s inconsistent performance than traditionally assumed, thereby challenging conventional assumptions about error sensitivity in human–AI interaction.

Abstract

The capabilities of artificial intelligence (AI) lie along a jagged frontier, where AI systems surprisingly fail on tasks that humans find easy and succeed on tasks that humans find hard. To investigate user reactions to this phenomenon, we developed an incentive-compatible experimental methodology based on diagram generation tasks, in which we induce errors in generative AI output and test effects on user reliance. We demonstrate the interface in a preregistered 3x2 experiment (N = 577) with error rates of 10%, 30%, or 50% on easier or harder diagram generation tasks. We confirmed that observing more errors reduces use, but we unexpectedly found that easy-task errors did not significantly reduce use more than hard-task errors, suggesting that people are not averse to jaggedness in this experimental setting. We encourage future work that varies task difficulty at the same time as other features of AI errors, such as whether the jagged error patterns are easily learned.

Problem

• Generative AI
• User Reliance
• Task Difficulty
• AI Errors
• Jagged Frontier

Innovation

• generative AI
• jagged frontier
• user reliance
• error rate
• task difficulty