LLM Agents Are Hypersensitive to Nudges

2025-05-16

AI Summary

This paper investigates the sensitivity of large language model (LLM) agents to behavioral nudges—such as default options and information highlighting—in multi-attribute tabular decision-making tasks. Methodologically, it integrates multi-round prompt engineering (zero-/few-shot chain-of-thought), human behavioral experiments, data distillation, and resource-rational modeling. The study makes three key contributions: first, it systematically demonstrates that LLM agents exhibit “hyper-sensitivity” to nudges—evidenced by distorted choice distributions, maladaptive information acquisition (either excessive or zero querying), and significantly suboptimal performance relative to human rationality benchmarks; second, while human-inspired optimal nudges partially improve agent performance, chain-of-thought prompting and human demonstrations only mitigate—not eliminate—this sensitivity; third, it proposes the novel paradigm of “resource-rational alignment,” advocating pre-deployment behavioral robustness testing to ensure LLM agents generalize reliably across diverse nudge contexts.

Technology Category

• Multiagent Systems: Mechanism Design
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Learning Human Values and Preferences

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

LLMs are being set loose in complex, real-world environments involving sequential decision-making and tool use. Often, this involves making choices on behalf of human users. However, not much is known about the distribution of such choices, and how susceptible they are to different choice architectures. We perform a case study with a few such LLM models on a multi-attribute tabular decision-making problem, under canonical nudges such as the default option, suggestions, and information highlighting, as well as additional prompting strategies. We show that, despite superficial similarities to human choice distributions, such models differ in subtle but important ways. First, they show much higher susceptibility to the nudges. Second, they diverge in points earned, being affected by factors like the idiosyncrasy of available prizes. Third, they diverge in information acquisition strategies: e.g. incurring substantial cost to reveal too much information, or selecting without revealing any. Moreover, we show that simple prompt strategies like zero-shot chain of thought (CoT) can shift the choice distribution, and few-shot prompting with human data can induce greater alignment. Yet, none of these methods resolve the sensitivity of these models to nudges. Finally, we show how optimal nudges optimized with a human resource-rational model can similarly increase LLM performance for some models. All these findings suggest that behavioral tests are needed before deploying models as agents or assistants acting on behalf of users in complex environments.

Problem

• LLM choices are highly sensitive to nudges
• LLMs diverge from humans in decision strategies
• Prompting methods fail to reduce nudge sensitivity

Innovation

• Studied LLM sensitivity to nudges in decisions
• Used prompting strategies like CoT alignment
• Optimized nudges with human-rational model