Beyond “painting in pink”: A critical case study of all-girls generative AI workshops in a European makerspace

Synthesis: This critical case study examines an all-girls initiative run by a European urban makerspace that integrated two generative AI (GenAI) tools — Playground (a free image generator) and Deforum (a free animation platform) — into workshops about climate change. The initiative was part of a two-year, research-practice partnership across six makerspaces in five countries foregrounding equity and social justice in STEM. The authors analyze qualitative, Multimodal AI data including 15 interviews with girls, parents, and practitioners, four reflective sessions, two observation fieldnotes, and 44 GenAI-generated images, interpreted through Harding's (1986) multi-level framework (symbolic, structural, individual) extended by a social level (Sultan et al., 2024) and feminist pedagogy.

While stakeholders generally viewed the all-girls format positively as a safer, more relaxed space, the study argues that the workshops fell short of gender-transformative practice. Practitioners enacted three steps — representation (symbolic), relationship building (social), and a responsive approach (individual) — but persistent challenges revealed the marginalization of girls' voices, agency, and assets. The paper distinguishes superficial girlification ("painting in pink") from genuinely centering girls, and proposes a girl-centered, multi-level pedagogical framework to move equity interventions beyond symbolic-level changes.

Key Findings

• The all-girls format was valued as a "safe" and "relaxed" space. Girls reported they talked more, felt braver to express ideas, and preferred all-girls over mixed-gender groups; parents noted girls often "got scared" and spoke less in mixed settings, echoing a chilly climate of gendered interpersonal behavior.
• Practitioners enacted three key steps: (1) representation at the symbolic level (female-centered promotional language and imagery, women role models to challenge stereotypes); (2) relationship building at the social level (ice-breakers like a Jenga Tower game, casual break-time conversation, valuing girls' funds of knowledge, an equal facilitator–girl relationship); and (3) a responsive approach at the individual level (adapting prompt-writing support for younger girls who lacked English confidence, letting girls choose digital drawing).
• Four interlinked challenges emerged in practice: an inappropriate workshop topic (climate change — chosen by a facilitator based on personal experience, with only two of 44 images reflecting it), limited peer collaboration (girls worked mostly individually on laptops, despite collaborative intent), inaccessibility of GenAI tools (Playground lost free functions and became "pro," a cost barrier for an under-resourced nonprofit), and language barriers in prompt-writing (the tools accepted English prompts only, a design-embedded structural barrier for non-native English speakers).
• Underlying issues were the marginalization of girls' voices, agency, and assets. Workshop topics, tools, and pedagogy were largely decided by practitioners from deficit-based assumptions, without girls' participation in decision-making; girls' strengths were less recognized, and power relations among girls, practitioners, and technology persisted.
• Tool experience differed by pedagogy, not just technology. Girls felt their ideas were inadequately reflected by Playground-generated images and preferred animating their own drawings with Deforum, which leveraged rather than replaced their original creations — underscoring that GenAI should amplify, not overshadow, girls' creative expression.
• Girlification is insufficient for structural change. Superficial adjustments (girl-only grouping, women facilitators, feminine marketing) operate mainly at the symbolic level; without changes across symbolic, structural, social, and individual levels, girls' marginalization persists.

What this means for practice

• Instructors. Center girls across planning, implementation, and evaluation instead of girlifying. Girl-only grouping, women facilitators, and feminine marketing operate at the symbolic level; interrogate power directly with questions such as "For whom/whose practice?" and "Based on whose values?" (Vossoughi, 2017).
• Instructors. Co-produce with girls as decision-makers and change agents — have them decide whether, what, and how GenAI is used — and reject deficit-based assumptions that presume girls lack interest, confidence, or relevant knowledge.
• Designers. Vet GenAI tools for equity before adopting them: Playground lost its free functions and became "pro" mid-project, a cost barrier for an under-resourced nonprofit, and English-only prompt input was a structural barrier for non-native speakers.
• Designers. Design GenAI activities to amplify girls' own making rather than replace it: girls preferred animating their own drawings with Deforum over image-generator output that did not reflect their ideas, and they need scaffolded prompt-writing rather than assumed English fluency.
• Administrators. Build the partnerships and peer collaboration the workshops lacked. Girls worked mostly individually on laptops despite collaborative intent, and partnerships among AI developers, practitioners, and researchers are needed to bridge tool design and equitable educational use.

Limitations

• A single critical case study of an all-girls initiative in one European urban makerspace, part of a two-year partnership across six makerspaces in five countries; the authors state the limited sample and specific context mean findings cannot be generalized to wider makerspace settings.
• Girls' own views rest on five interviews conducted by youth co-researchers with limited interview experience, producing short and closed responses, and all participating girls were white and lived in an urban European city, so their experiences are not equivalent to those of marginalized girls facing more complex intersectional challenges.
• Girls' talk in a European language was not directly accessible to the researchers; translated and paraphrased accounts shared later by practitioners may have reduced the nuances of the girls' original meanings.
• The data corpus is small and multi-source — 15 interviews with girls, parents, and practitioners, four reflective sessions, two observation fieldnotes, and 44 GenAI-generated images — and girls' direct participation in decision-making was limited throughout the initiative.

Citation

Liu, Q., Archer, L., Nag Chowdhuri, M., Freedman, E., & DeWitt, J. (2026). Beyond "painting in pink": A critical case study of all-girls generative AI workshops in a European makerspace and implications for gender equity in computing. Computers and Education: Artificial Intelligence, 10, 100602.