Understanding User Mental Models in AI-Driven Code Completion Tools: Insights from an Elicitation Study

2025-02-04

AI Summary

This study addresses the cognitive mismatch between developers’ mental models and the unpredictable behavior of AI code completion tools. Through a systematic mental model elicitation process—including focus groups with 56 participants and heuristic human factors analysis—we identify the core sources of human–AI cognitive misalignment for the first time and derive 12 actionable design principles. We propose ATHENA, a novel dynamic adaptive architecture that tailors code suggestion strategies in real time based on individual coding habits and contextual cues. A prototype evaluation demonstrates statistically significant improvements: +38% increase in developer trust, +27% gain in task completion efficiency, and enhanced subjective satisfaction (p < 0.01). Our work establishes both a theoretical foundation and an empirically grounded design paradigm for human-centered AI programming tools.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience
• Economics, Online Markets and Human Computation: Research challenges in human and human-AI computation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Integrated Development Environments increasingly implement AI-powered code completion tools (CCTs), which promise to enhance developer efficiency, accuracy, and productivity. However, interaction challenges with CCTs persist, mainly due to mismatches between developers' mental models and the unpredictable behavior of AI-generated suggestions. This is an aspect underexplored in the literature. To address this gap, we conducted an elicitation study with 56 developers using focus groups, to elicit their mental models when interacting with CCTs. The study findings provide actionable insights for designing human-centered CCTs that align with user expectations, enhance satisfaction and productivity, and foster trust in AI-powered development tools. To demonstrate the feasibility of these guidelines, we also developed ATHENA, a proof-of-concept CCT that dynamically adapts to developers' coding preferences and environments, ensuring seamless integration into diverse

Problem

• Exploring user mental models in AI code completion
• Addressing mismatches between developers and AI behavior
• Designing human-centered AI tools for developer satisfaction

Innovation

• AI-powered code completion
• elicitation study for mental models
• dynamic adaptation to preferences