LLM-based agents respond differently to digital nudges in interfaces
A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents
Artificial Intelligence
Summary
Digital assistants powered by large language models (LLMs) often make choices for users within software designed for humans. The authors studied whether these AI agents are influenced by design tricks (called nudges) that can affect decisions automatically or after reflection. They found that agents can be swayed by both simple defaults and social influence nudges, and that increasing the AI’s reasoning changes but does not eliminate this influence. The research shows that how an AI reasons shifts the way it is nudged, highlighting interface design as important for responsible AI use.
What this means in practice
- •For software product teams: Design user interfaces considering how AI agents are influenced by nudges to guide automated decisions safely.
- •For ai governance teams: Develop policies that treat interface design as a critical factor when deploying autonomous AI agents to manage risks from digital nudges.
Authors
Haya Halimeh, Sascha Kaltenpoth, Kevin Bösch, Oliver Müller
Abstract
LLM-based GUI agents increasingly act on behalf of users in digital environments that were designed with human users in mind. These graphical user interfaces were designed to support, but also deliberately steer, the behaviour and decisions of users. While behavioural biases in the textual outputs of LLMs are well-documented, far less is known about how such influence operates when models act as agents that perceive interfaces and execute decisions---and, in particular, whether the reasoning capabilities increasingly built into these agents make them more robust to it. Drawing on Dual-Process Theory, we empirically investigate whether LLM-based GUI agents are susceptible to automatic (Type 1) and reflective (Type 2) digital nudges, and how their reasoning configuration moderates this susceptibility. In a randomized online shopping experiment with 3,600 agents and a total of 21,600 simulations across six frontier models from three providers, we found that agents were vulnerable to both nudge types. Crucially, the reasoning configuration moderated these effects in opposing directions, reducing susceptibility to automatic default nudges while heightening it to reflective social influence nudges. Extensive reasoning therefore did not make agents more robust but redirected the route through which choice architecture takes effect. Exploratory analysis further showed this redirection to be systematically structured by model scale. Beyond establishing nudge susceptibility as a behavioural property of agentic AI, the study positions interface design as a governance concern for organizations that delegate decisions to autonomous agents.