Applying Explainable Artificial Intelligence Principles to Interface Design: Improving User Trust and Understandability in a Chicken Weight Monitoring System

Abstract

Artificial Intelligence (AI) has been increasingly adopted in smart farming to support monitoring and decision-making processes. However, many AI-based systems still operate as black boxes, making their outputs difficult for end users to understand and potentially reducing user trust. Although Explainable Artificial Intelligence (XAI) has been proposed to improve transparency, studies integrating XAI principles into interface design and evaluating their effects on user experience remain limited, particularly in smart farming contexts. This study investigates the implementation of XAI principles in redesigning the interface of a chicken weight monitoring system and evaluates their effects on user trust, understandability, and usability. A concurrent embedded mixed methods approach with a within-subject and counterbalanced design was conducted involving 16 participants. The redesigned interface incorporated human-centered XAI principles and was evaluated using the Trust in Automation Scale (TiAS), an understandability questionnaire, and the System Usability Scale (SUS). The results showed statistically significant improvements across all evaluated aspects (p < 0.001). Trust increased from 50.52 to 73.82, understandability from 54.25 to 81.88, and usability from 43.28 to 74.38, with large effect sizes observed in all measurements. Qualitative findings indicated that clearer and contextual explanations improved users’ interpretation of system outputs. These findings suggest that integrating XAI principles into interface design can support more transparent and understandable interaction in AI-based monitoring systems.