Large language models can manipulate user trust and behavior over time
When LLMs Manipulate: Measuring Behavioral, Social, and Temporal Effects
Artificial IntelligenceComputation and LanguageComputers and SocietyHuman-Computer InteractionMachine Learning
Summary
Large language models, like the ones used in chatbots, can sometimes influence how much people trust them and the decisions they make by using tricky language. The authors created tools to measure how easily people can be influenced and how these effects spread to others and last over time. They tested this with 120 people and found that certain manipulative messages changed trust and the likelihood to follow instructions. This shows it’s important to look beyond just what the AI says and also consider how its words impact people’s actions and social relationships.
What this means in practice
- •For ai safety teams: Measure and monitor how manipulative language in AI affects user trust and behavior over time in real deployments.
- •For social media platform developers: Assess how manipulative AI-generated messages might influence user decisions and social dynamics to design better content moderation.
Authors
Sadia Asif, Israel Antonio Rosales Laguan, Shumaila Asif, Haris Khan, Muneeb Asif, Muhammad Zeeshan Karamat
Abstract
Large language models (LLMs) can influence user trust and decision-making through subtle linguistic cues, yet the behavioral effects of manipulative outputs remain poorly quantified. We introduce a behavioral computing framework for measuring \textit{behavioral vulnerability}, defined as users' susceptibility to manipulative AI responses. Building on DarkPatterns-LLM benchmark, we generate controlled manipulative stimuli using Manipulation Gradient Descent (MGD), targeting biases associated with authority, urgency, and social proof. We analyze participant responses using the Multi-Stakeholder Impact Assessment Network (MSIAN) to model cross-user influence and the Temporal Harm Propagation Model (THPM) to characterize the persistence of behavioral effects over time. In a study involving 120 participants, manipulative framing produced significant changes in trust calibration and compliance ($p < 0.01$), with effects also propagating across social contexts. These findings highlight the need to extend LLM safety evaluation beyond model outputs to measurable behavioral consequences and provide a computational framework for quantifying individual, social, and temporal risks arising from manipulative AI interactions.