LLMs in Autonomous Vehicles Inherit Human Biases in Pedestrian Yielding

A new benchmark study reveals that large language models (LLMs) and visual-language models (VLMs) used in autonomous vehicles (AVs) inherit human biases. These models demonstrate discriminatory behavior in pedestrian-yielding decisions, showing differential treatment based on factors such as ethnicity, gender, religion, disability, age, skin tone, and socioeconomic status. The research highlights common patterns of bias across different models, raising concerns about the current paradigm of using general-purpose "common sense" models for AV decision-making and emphasizing the need for bias mitigation strategies. AI

IMPACT Raises concerns about the safety and fairness of AI in autonomous vehicles, potentially slowing adoption if biases are not addressed.

RANK_REASON The cluster is based on an academic paper proposing a new benchmark and presenting research findings on LLM bias. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Autonomous Vehicles
• Black pedestrians
• human driver biases
• large-language models
• LLM-Driven Autonomous Vehicles
• pedestrian-yielding decisions
• Visual Language Models

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →