Preliminary Quantitative Study on Explainability and Trust in AI Systems

TL;DR

The study investigates how different explanation modalities affect user trust in AI-driven decisions within a loan-approval task. Using a $3\\times3\\times2$ factorial design with $N=15$ participants and $5$ loan scenarios, it contrasts four explanation types across two AI models with distinct performance levels. Results show interactive counterfactual explanations yield the highest trust and perceived understanding, though they can increase cognitive load; trust calibration also varies with user expertise. The findings offer quantitative guidance for designing adaptive, human-centered XAI interfaces and inform governance considerations for trustworthy AI in high-stakes settings.

Abstract

Large-scale AI models such as GPT-4 have accelerated the deployment of artificial intelligence across critical domains including law, healthcare, and finance, raising urgent questions about trust and transparency. This study investigates the relationship between explainability and user trust in AI systems through a quantitative experimental design. Using an interactive, web-based loan approval simulation, we compare how different types of explanations, ranging from basic feature importance to interactive counterfactuals influence perceived trust. Results suggest that interactivity enhances both user engagement and confidence, and that the clarity and relevance of explanations are key determinants of trust. These findings contribute empirical evidence to the growing field of human-centered explainable AI, highlighting measurable effects of explainability design on user perception

Figures (3)

• Figure 1: Overview of study setup and participant flow.
• Figure 2: Full Trust Scale used in the study, presented across two sections for readability.
• Figure 3: Explainability Scale used in the study. Items assess perceived correctness, completeness, coherence, and contextual utility of AI explanations.