📄 Abstract
Large Language Models (LLMs) are increasingly used to conduct human decision-making, raising concerns regarding their role in potentially reinforcing behavioural biases. The research compares anchoring and loss aversion in Human-LLM interaction supported by the theoretical foundation of Prospect Theory. A quantitative, between-subjects experimental design was employed with 251 participant responses. Numerical anchors, gain/loss framing, AI trust and AI influence were examined as influences on decision-making outcomes. For statistical analysis, descriptive statistics, reliability analysis, independent t-test, chi-square test, paired t-test, correlation and regression test, ANOVA were done. Results showed that AI-generated recommendations played a significant role on participants' decisions, whereas gain/loss framing heavily affected risk preferences. Moreover, participants exhibited greater sensitivity to losses compared to gain equivalent to Prospect Theory. Moreover, AI trust but significantly predicted AI influence and reliance. Overall, the study's findings have extended findings of the existing literature on behavioural decision-making and human-AI interaction and provided evidence that LLMs may influence judgements through anchoring, framing, and trust-related mechanisms.
🏷️ Keywords
📚 How to Cite:
Biplab Paul, Gourab Majumder, Aditi Barai, Priyangshu Goswami, Kushal Chanda , ANCHORING AND LOSS AVERSION IN HUMAN-LLM INTERACTIONS: A PROSPECT THEORY APPROACH TO AI-ASSISTED DECISION-MAKING , Volume 12 , Issue 9, September 2026, EPRA International Journal of Multidisciplinary Research (IJMR) , Pages: 435 - 445 , DOI: https://doi.org/10.36713/epra31560