Supervisors

Suitable for

Abstract

Agentic systems, powered by large language models (LLMs), rely heavily on APIs to perform tasks such as navigating the internet, booking hotels, scheduling flights, and managing calendars. These systems depend on API interactions to make decisions, but this reliance raises significant fairness concerns. For example, biases in API design, ranking, or pricing models can lead to unfair preferences for certain APIs over others, disproportionately benefiting some API providers—especially in usage-based compensation models. This project investigates fairness challenges in API interactions for LLM-based agents, explores how API biases emerge, and evaluates their impact on system behavior and equity. It will focus on developing fairness metrics and robust mitigation techniques to ensure unbiased decision-making in agentic systems.

This project is designed to lead to a high-quality publication, and we are looking for a highly motivated student to contribute to this critical and impactful research. 

 

We will aim to collaborate with the UK’s AI Safety Institute and Guohao Li (co-founder of Eigen AI and Camel AI).