Human-Centered Explainable AI (HCXAI) Enhanced by Information-Seeking Psychology: A New Theoretical Framework
Summary:This position paper introduces a novel framework for Human-Centered Explainable AI (HCXAI) that incorporates insights from the psychology of information seeking. It argues that users' engagement with AI explanations is driven by three types of expected utility: instrumental, hedonic, and cognitive, all of which are influenced by well-documented cognitive biases such as the illusion of control and automation bias. The study advocates for a shift in AI explanation design—from merely making explana
Background and Motivation
In recent years, Explainable AI (XAI) has emerged as a critical area of research in AI. However, most existing XAI methods focus on technical implementation while neglecting the psychological factors and actual needs of users regarding explanations. This study draws on the psychology of information seeking to propose a new HCXAI framework that addresses this gap.
Key Concepts
-
Three Types of Expected Utility
- Instrumental Utility: Whether the explanation helps users make better decisions.
- Hedonic Utility: Whether the explanation makes users feel more comfortable or satisfied.
- Cognitive Utility: Whether the explanation enhances users' understanding.
-
Impact of Cognitive Biases
- Illusion of Control: Users tend to overestimate their control over AI systems.
- Automation Bias: Users overly rely on AI systems and neglect their own judgment.
- Unrealistic Optimism: Users have overly optimistic expectations of AI system performance.
- Confirmation Bias: Users tend to seek explanations that support their existing beliefs.
-
Two Failure Modes
- Excessive Information-Seeking: Users become distracted by excessive details, affecting decision efficiency.
- Insufficient Information-Seeking: Users fail to fully understand AI system behavior, leading to critical risks being overlooked.
Application of the Framework
The framework emphasizes that AI systems should be designed to encourage users to actively seek explanations rather than passively providing them. Particularly in complex tasks, AI explanations need to support users in comprehensively understanding system behavior, including risk assessment, action anticipation, and decision-making on intervention.
Recommendations for AI Developers
- Design Explanation Mechanisms with a User-Centric Approach: Adjust the content and form of explanations based on user needs and psychological factors.
- Introduce Mitigation Strategies for Cognitive Biases: For example, provide diverse explanation perspectives to reduce the impact of confirmation bias.
- Evaluate the Effectiveness of Explanations: Continuously optimize the explanation system through user feedback and experimental data.
Industry Impact
This research provides new theoretical guidance for the design of AI explanation systems, helping to enhance user trust and the efficiency of AI use. It is particularly significant in fields such as healthcare and finance, where the reliability of AI is crucial.
— END —Source: ArXiv AI (cs.AI) (2026-09-01)
Tags: #HCXAI #Information-Seeking Psychology #Cognitive Biases #AI Explanation #Human-Computer Interaction
Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.
Community Comments