Unseen, unsaid, unchallenged: Gender bias in generative AI science outputs

This session presents my original research into the emergent bias evident in both text and image outputs from generative AI tools used in educational settings, particularly when asked to give science explanations or generate science-related images. Based on my own data and analysis, it highlights how these tools subtly - and sometimes not so subtly - reinforce stereotypes. It explores the risks this bias poses for inclusion, aspiration, representation, and social equity. As students increasingly rely on AI for homework support, revision tasks, and independent learning, and as teachers use AI tools to plan lessons and create resources, the potential impact of these biases becomes more significant. Reinforced stereotypes in AI outputs may influence students' perceptions of who belongs in science, shape their science identities, and affect their confidence, aspiration, and progression into science careers. The session is designed for all teachers and teacher educators engaged with teaching science. I will share examples from my study, including real prompt-response output pairs and AI-generated images. Together, we will examine how these outputs shape narratives about who is seen to ‘belong’ in science and consider the broader consequences for uptake, retention, and diversity in the subject. You will leave with a deeper understanding of how these biases could be connected to your GenAI requests. You will take away practical strategies to mitigate, challenge, and reduce bias when using AI tools with pupils, trainees, colleagues, or in your own lesson and curriculum preparation and professional development. Intended Outcomes: By the end of the session, participants will have developed a stronger awareness of how gender bias is embedded in both text and image outputs from generative AI tools and will understand how this bias is often formed through the language used in prompts. They will reflect on how these outputs can influence pupil identity, aspiration, and representation in science, particularly in ways that may impact progression into science careers. Participants will be equipped with inclusive, bias-aware prompting strategies that they can apply in their own teaching or teacher education contexts, supporting more critical, equitable, and aspirational use of generative AI.