This qualitative pilot study informs my dissertation on how artificial intelligence (AI), including conversational agent (CA)-like features, can be integrated into geography-related science museum exhibits in thoughtful and just ways (Okonkwo & Ade-Ibijola, 2021; Okita & Clarke, 2021). Rather than starting from “AI solutions,” I begin with what families already do: how children and caregivers use bodies, talk, gesture, and exhibit resources to make sense of geography-rich phenomena in non-AI exhibits (e.g., gravity-themed jumping, planetary projections, and map-like spatial puzzles).

The study is grounded in sociocultural and situated perspectives on learning and treats exhibits as socio-material learning spaces where participation is organized through the entanglement of bodies, layouts, representations, and interaction rules. It draws on three strands of scholarship. First, research on family and informal science learning shows that sensemaking and identity-relevant participation emerge through intergenerational talk, gesture, and joint manipulation of museum resources (Dominke et al., 2025; Hohenstein & Tenenbaum, 2023; Nygren et al., 2024). Second, distributed scaffolding frames support as spread across tools, people, and activities—not delivered by a single “expert” (Puntambekar, 2022). Third, justice-oriented perspectives in informal STEM education argue that equity cannot be reduced to access or inclusion alone; researchers must examine how authority, recognition, and opportunities to contribute are distributed in learning spaces (Greenberg et al., 2025).

In parallel, the proposal is informed by my justice-oriented conceptual reanalysis of CA-supported collaborative science learning. From 882 initial records, PRISMA screening yielded 27 empirical studies included in the synthesis (Moher et al., 2009). I developed a three-dimensional justice lens (Fricker, 2007; Learning Towards Justice Team, 2025) focused on: (1) distribution of scaffolds (e.g., who gets explanations/translation/prompts; timing/frequency; whether support follows child vs caregiver), (2) participation structures (e.g., interface control; turn-taking; who initiates/finishes; whose ideas get ratified), and (3) epistemic agency (e.g., whether children’s hypotheses are solicited, extended, or overridden; whether uncertainty is treated as productive). I adapt this lens to museums to treat “AI readiness” as an empirical question about existing participation infrastructures, rather than a technological claim. Specifically, I treat geography-rich exhibits as participation infrastructures: their layouts, interaction rules, and representational forms quietly organize who gets to act, speak, be recognized, and taken up during family sensemaking. Here, AI is not an “add-on,” but a potential reconfiguration of these existing infrastructures.

European and international relevance is central because science museums are public educational spaces serving multilingual and mobile families, where language access, disability access, and public accountability shape what “help” can legitimately look like. In European contexts, AI-mediated visitor experiences are increasingly evaluated against expectations of data protection and transparency (e.g., GDPR principles such as transparency and data minimization) and emerging governance for AI systems in public life (e.g., the EU AI Act’s phased implementation). These conditions make museums a useful site for a European dialogue on how AI reorganizes voice, recognition, and opportunity-to-learn in public learning spaces. 

Research Questions:

1. How do children and caregivers interact with and make sense of selected geography-related exhibits in a science museum that rely on physical or rule-based interaction rather than AI/CA mediation? 

2. Where do visitors and educators perceive challenges and possibilities, and how do they imagine AI or CA-like prompts/adaptive feedback might be helpful or problematic for equity, voice, and participation structures? 

3. How do children, caregivers, and museum educators imagine connections between AI-enhanced exhibits and classroom science/geography learning for linguistically and culturally diverse learners, and what follow-up resources do they envision?

This study will be conducted in one large science museum with exhibits related to space, Earth systems, and geography. I will purposively select 2–3 focal exhibits that (a) foreground spatial/geographic sensemaking, (b) invite sustained embodied interaction (e.g., jumping, aligning, manipulating), and (c) rely on rule-based or pre-programmed interaction rather than AI mediation. To improve comparability, observations will be distributed across different days/times (e.g., weekend and weekday sessions). Participants and sampling. I will use naturalistic observation of approximately 20–30 family groups at the focal exhibits. Following the interaction, I will invite a subset of families (approximately 10–15 groups) to participate in brief semi-structured interviews (5–10 minutes). I will also interview museum educators (approximately 4–6) and, if feasible, 1–2 teachers who organize museum visits. Sampling will be purposive to include variation in child age, caregiver roles, and linguistic/cultural backgrounds when self-reported. Data sources. (1) Field notes capturing interaction “episodes” (from approach to exit), including sequences of action, gesture, talk, role shifts, interface/control patterns, and references to signage or facilitation. When feasible, I will add quick spatial sketches to document layout, positioning, and movement pathways. (2) Post-interaction interviews eliciting participants’ interpretations, moments of interest/confusion, and experiences of inclusion/exclusion, followed by 1–2 short AI design scenarios to surface perceived benefits, risks, and fairness concerns. Interviews will be audio recorded with consent; otherwise, expanded notes will be written immediately after. (3) Educator/teacher interviews focused on exhibit goals, typical visitor patterns, and anticipated shifts in learning and inclusion under AI-enhanced designs. Analysis. Analysis will be iterative. I will begin with open coding across notes and transcripts, then develop focused codes aligned with the three justice dimensions (distribution of scaffolds, participation structures, epistemic agency). I will build cross-case matrices (exhibits × family groups) to compare patterns and generate grounded design conjectures. Trustworthiness will be supported through reflexive memoing, triangulation across data sources, and examination of discrepant cases. Feasibility and timeline: Data collection is planned to begin in April 2026, pending IRB approval and museum access confirmation. Initial coding and cross-case matrices are anticipated in May–June 2026, enabling preliminary patterns and design conjectures to be shared at ERC in August 2026.

As a pilot-in-progress, the study will not claim definitive findings about AI impacts. Instead, it will produce three outcomes that strengthen a dissertation trajectory and contribute to international discussions of educational space, justice, and AI in informal learning. First, it will produce a grounded description of how families currently engage geography-rich embodied exhibits—documenting recurring participation patterns (who leads, who watches, how roles rotate), key moments of sensemaking, and where exhibit layouts and interaction rules appear to open or constrain opportunities for different family members. Second, it will surface visitor- and educator-articulated opportunities and tensions for AI integration. Anticipated themes include desired supports (e.g., translation, prompts that amplify quieter voices, explanations responsive to children’s embodied actions) and perceived risks (e.g., surveillance concerns, language-based exclusion, narrowing what counts as a “good” response). These insights will be organized using the justice lens of scaffold distribution, participation structures, and epistemic agency, so that future AI designs can be assessed not only for perceived benefits, but for how they redistribute opportunity-to-learn. Third, it will map how participants imagine connections between museum experiences and classroom science/geography learning—especially for linguistically and culturally diverse learners—and what follow-up resources (take-home prompts, classroom extensions, family conversation supports) they value. Together, these outcomes will inform design requirements for AI-enhanced geography exhibits that remain responsive to real intergenerational practices and justice-centered aims. Deliverables will include (a) a cross-case participation map (exhibits × family groups) capturing control, talk, gesture, and recognition patterns; (b) a brief justice-oriented coding memo specifying observable indicators for each lens dimension; and (c) a set of grounded design conjectures for what “just AI help” would mean in embodied, intergenerational museum settings.

Fricker, M. (2007). Epistemic injustice: Power and the ethics of knowing. Oxford University Press. Learning Towards Justice Team. (2025). Avoiding equity detours on the way to a more justice-oriented learning sciences. Journal of the Learning Sciences, 34(3), 368–402. Greenberg, D., Kim, W. J., Brien, S., Barton, A. C., Balzer, M., & Archer, L. (2025). Designing and leading justice centered informal STEM education: A framework for core equitable practices. Science Education, 109(1), 27–58. https://doi.org/10.1002/sce.21903 Nygren, M. O., Price, S., & Thomas Jha, R. (2024). The role of embodied scaffolding in revealing enactive potentialities in intergenerational science exploration. Science Education, 108(2), 495–523. https://doi.org/10.1002/sce.21845 Hohenstein, J., & Tenenbaum, H. R. (2023). Family conversations about species change as support for children’s developing understandings of evolution. Science Education, 107(3), 810–834. https://doi.org/10.1002/sce.21783 Okita, S. Y., & Clarke, S. N. (2021). Robots and agents to support collaborative learning. In International handbook of computer-supported collaborative learning (pp. 407–424). Springer. https://doi.org/10.1007/978-3-030-65291-3_22 Okonkwo, C. W., & Ade-Ibijola, A. (2021). Chatbots applications in education: A systematic review. Computers and Education: Artificial Intelligence, 2, 100033. https://doi.org/10.1016/j.caeai.2021.100033 Puntambekar, S. (2022). Distributed scaffolding: Scaffolding students in classroom environments. Educational Psychology Review, 34, 451–472. https://doi.org/10.1007/s10648-021-09636-3 Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G., & PRISMA Group. (2009). Preferred reporting items for systematic reviews and meta-analyses. Annals of Internal Medicine, 151(4), 264–269. Tang, K. (2024). Informing research on generative artificial intelligence from a language and literacy perspective: A meta synthesis of studies in science education. Science Education, 108(5), 1329–1355. https://doi.org/10.1002/sce.21875 Lee, H., Kang, D. Y., Kim, M. J., & Martin, S. N. (2023). Navigating into the future of science museum education: Focus on educators’ adaptation during COVID 19. Cultural Studies of Science Education, 18(3), 647–667. https://doi.org/10.1007/s11422-022-10142-3 Dominke, H., Barenthien, J., Oppermann, E., Burghardt, L., & Steffensky, M. (2025). The quality of interactions in the home science environment and associations with children’s science learning. Science Education, 109(5), 1365–1383. https://doi.org/10.1002/sce.21964