[SEMINAR] Advancing Automated Ethical Profiling in SE: A Zero-Shot Evaluation of LLM Reasoning

Speaker: Mashal Afzal Memon

When: Tuesday, 29h October, 14:00-15:00

Where: Alan Turing Seminar Room

Abstract: Large Language Models (LLMs) are increasingly used across diverse domains, yet their ethical reasoning capabilities remain unclear. This talk investigates whether LLMs demonstrate ethical reasoning capabilities and explores the possibility of automating ethical profile generation through the observation of user behavior in real-life contexts. To address these questions, the talk presents results from an evaluation of 16 LLMs, each prompted with 30 real-world moral dilemma scenarios in a zero-shot setting. The findings offer insights into how ethical reasoning mechanisms can be identified, assessed, and potentially integrated into software engineering (SE) pipelines.