Perspectives on Explanation Formats From Two Stakeholder Groups in Germany: Software Providers and Dairy Farmers

Software providers often misjudge what farmers actually want: The study found "discrepancies between the views of both groups," with software providers making inaccurate assumptions about farmers' preferences for how AI explanations should be presented. This gap between developer assumptions and user needs may contribute to farmers' reluctance to adopt digital systems.

Different explanation formats matter for trust and understanding: The research tested four distinct ways to explain AI mastitis detection decisions (textual descriptions, rule-based explanations, herd comparisons, and time series visualizations), finding that farmers rated these formats differently in terms of comprehensibility and trustworthiness—suggesting one-size-fits-all explanations are insufficient.

Direct user engagement is essential for responsible AI deployment: The authors emphasize that "a thorough user requirements analysis (farmers' needs)" is necessary to improve both software design and user acceptance. This suggests that responsible AI implementation requires developers to actively consult with end-users rather than relying on assumptions about their needs.

Stakeholder perspectives diverge in meaningful ways: By comparing software providers' predictions about farmer preferences with farmers' actual preferences, the study reveals that different groups involved in AI systems may have fundamentally different views on what makes AI trustworthy and usable—a gap that needs bridging for successful implementation.