E62 Teachers' Awareness of Biases in AI (Short paper) Introduction AI-based systems—including generative AI, recommender systems, and automated assessment—are rapidly entering schools. While these tools promise efficiency and personalization, they also raise equity risks when outputs are treated as neutral or objective. Cultural bias may emerge when training data underrepresent groups, when labels reflect dominant norms, or when model performance differs across populations, leading to stereotypical content, unequal recommendations, and misclassification (Buolamwini & Gebru, 2018; Caliskan et al., 2017; Mehrabi et al., 2021; Noble, 2018). In education, such bias can shape what students see, how they are evaluated, and which opportunities they are encouraged to pursue. Teachers are key mediators: they choose tools, interpret outputs, and decide whether to trust, question, or override algorithmic suggestions. Prior work indicates that technology-related factors (digital use and anxiety) influence teachers' engagement with educational technologies (Parasuraman & Colby, 2015; Scherer et al., 2019; Teo, 2011), whereas multicultural competence may increase sensitivity to exclusion and inequity (Gay, 2018; Siwatu, 2007). Schools also rely on ICT coordinators as local technology leaders; however, the contribution of this formal role to teachers' awareness of AI bias remains underexplored. Theoretical model and hypotheses Figure 1 presents the theoretical model. We conceptualized awareness of cultural biases in AI as teachers' tendency to recognize biased patterns in AI systems (e.g., stereotypical representations, unequal recommendations) and to critically evaluate outputs. We hypothesized that awareness would increase with (H1) higher intercultural competence and (H2) higher social proximity. We further hypothesized that awareness would be positively related to (H3) digital use and negatively related to (H4) technological anxiety. Finally, we expected role-based technological experience—(H5a) higher technological training and (H5b) serving as an ICT coordinator—to be associated with higher awareness and to explain incremental variance beyond cultural and digital factors. The model was tested using group comparisons and hierarchical regression. Figure 1. Conceptual model of associations with AI Bias Awareness. Method Participants were 120 Israeli K–12 teachers (81.7% women), aged 22–69 (M = 40.16, SD = 10.04). Participants completed an online survey distributed through teacher professional networks; participation was voluntary and anonymous.
RkJQdWJsaXNoZXIy Mjk0MjAwOQ==