HumanAgencyBench: Scalable Evaluation of Human Agency Support in AI Assistants
read the original abstract
As humans delegate more tasks and decisions to artificial intelligence (AI), we risk losing control of our individual and collective futures. Relatively simple algorithmic systems already steer human decision-making, such as social media feed algorithms that lead people to unintentionally and absent-mindedly scroll through engagement-optimized content. In this paper, we develop the idea of human agency by integrating philosophical and scientific theories of agency with AI-assisted evaluation methods: using large language models (LLMs) to simulate and validate user queries and to evaluate AI responses. We develop HumanAgencyBench (HAB), a scalable and adaptive benchmark with six dimensions of human agency based on typical AI use cases. HAB measures the tendency of an AI assistant or agent to Ask Clarifying Questions, Avoid Value Manipulation, Correct Misinformation, Defer Important Decisions, Encourage Learning, and Maintain Social Boundaries. We find low-to-moderate agency support in contemporary LLM-based assistants and substantial variation across system developers and dimensions. For example, while Anthropic LLMs most support human agency overall, they are the least supportive LLMs in terms of Avoid Value Manipulation. Agency support does not appear to consistently result from increasing LLM capabilities or instruction-following behavior (e.g., RLHF), and we encourage a shift towards more robust safety and alignment targets.
This paper has not been read by Pith yet.
Forward citations
Cited by 6 Pith papers
-
Cognitive offloading and the speedup illusion in human-AI interaction
Preregistered behavioral study identifies a speedup illusion where users overestimate time savings from AI assistance on cognitive tasks despite no actual difference in completion times.
-
The efficiency-gain illusion: People underestimate the rate of AI use and overestimate its benefits on simple tasks
Three pre-registered studies with 2691 participants show people underestimate their AI usage rate and overestimate efficiency gains on simple tasks, with prior use entrenching further adoption.
-
What Did They Mean? How LLMs Resolve Ambiguous Social Situations across Perspectives and Roles
LLMs produce interpretive closure in 87.5% of ambiguous social scenarios through narrative alignment, reversal, or normative advice, with first-person perspectives increasing alignment tendencies.
-
Effects of Generative AI Errors on User Reliance Across Task Difficulty
Higher generative AI error rates reduce user reliance, but task difficulty does not significantly moderate this effect.
-
AI usage patterns are shaped by perceived gains in human agency
Ethnographic study of 51 AI chatbot users finds that perceived gains in individual agency shape sustained usage patterns more than accuracy or reliability concerns.
-
Althea: Human-AI Collaboration for Fact-Checking and Critical Reasoning
Althea integrates retrieval-augmented reasoning with varying levels of user scaffolding to improve fact-checking accuracy and foster persistent improvements in critical thinking.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.