Foveate, Attribute, and Rationalize: Towards Physically Safe and Trustworthy AI

TL;DR

This paper introduces FARM, a framework that uses external knowledge to generate trustworthy rationales for safety classification in AI, improving detection of unsafe text and enhancing interpretability for safer AI deployment.

Contribution

FARM is a novel knowledge-based approach that foveates missing information, retrieves trustworthy sources, and generates human-interpretable rationales for safety assessment in AI.

Findings

Achieves 5.9% higher safety classification accuracy on SafeText dataset.

Provides interpretable rationales to identify unsafe content.

Enhances trustworthiness and safety management in AI systems.

Abstract

Users' physical safety is an increasing concern as the market for intelligent systems continues to grow, where unconstrained systems may recommend users dangerous actions that can lead to serious injury. Covertly unsafe text is an area of particular interest, as such text may arise from everyday scenarios and are challenging to detect as harmful. We propose FARM, a novel framework leveraging external knowledge for trustworthy rationale generation in the context of safety. In particular, FARM foveates on missing knowledge to qualify the information required to reason in specific scenarios and retrieves this information with attribution to trustworthy sources. This knowledge is used to both classify the safety of the original text and generate human-interpretable rationales, shedding light on the risk of systems to specific user groups and helping both stakeholders manage the risks of their systems and policymakers to provide concrete safeguards for consumer safety. Our experiments show that FARM obtains state-of-the-art results on the SafeText dataset, showing absolute improvement in safety classification accuracy by 5.9%.