LLM Targeted Underperformance
Reduced LLM Reliability for Global, Less-Educated, and ESL Users [Archived]
Research
While state-of-the-art Large Language Models (LLMs) have shown impressive performance on many tasks, systematically evaluating undesirable behaviors of generative models remains critical. In this work, we visit the notion of ethics and bias in terms of how model behavior changes depending on three user traits: English proficiency, education level, and country of origin. We evaluate how fairly LLMs respond to different users in terms of information accuracy, truthfulness, and refusals. We present extensive experimentation on three state-of-the-art LLMs and two different datasets targeting truthfulness and factuality. Our findings suggest that undesirable behaviors occur disproportionately more for users with lower English proficiency, of lower education status, and originating from outside the US, rendering these models unreliable sources of information towards their most vulnerable users.
Collaborators
More Research
New Social Forms
Exploring new rituals, formats, and structures for coming together
Pilots & Programs
Storyline
An auditable AI framework for tracing competing narratives across podcasts, conversations, and news
Research
AI-Assisted Conversation Highlighting
Discovering semantically or emotionally salient moments in spoken discourse using LLMs
Research
realtalk@Boston
A new civic infrastructure in Boston grounded in dialogue as a way to building “civic muscle” of democracy
Pilots & Programs
Anthology
An AI interface that turns raw conversation audio into interactive maps
Research
Agora
A training ground to practice consensus-finding with real human perspectives
Research
