Figure 2
The diagram presents the hypothesised structural model and estimated relationships. Interaction quality, human agency, perceived humanness, and voice are modelled as antecedents of fairness, accountability, and transparency. Fairness, accountability, and transparency are then positively related to trust in AI-supported performance evaluation. Solid paths indicate statistically significant relationships, while dashed paths indicate non-significant relationships. Interaction quality, human agency, and perceived humanness show significant positive relationships with fairness, accountability, and transparency. Voice is significantly related to transparency but not to fairness or accountability. Task complexity and perceived uncanniness are modelled as moderators of the relationships between fairness, accountability, transparency, and trust; the moderation effects shown in the figure are non-significant. The model explains 50.1% of the variance in fairness, 53.7% in accountability, 61.0% in transparency, and 61.1% in trust in AI-supported performance evaluation.Estimated model. Note: Solid paths represent significant relationships and dashed paths represent non-significant relationships. R2 values shown for the endogenous constructs correspond to the main-effects model. The dashed H6 and H7 paths report moderation estimates from the extended moderated model, in which the explained variance in trust increased to R2 = 0.630. Source: Authors' own work