Abstract

This work examines the possibilities and perils—mostly perils—of collaborating with artificially intelligent (AI) systems in a secondary art classroom. During an intensive four-week AI-unit, students (un)intentionally glitched (Menkman, 2011), hacked (Coleman, 2013), and red teamed (Walton & Bae, 2024) AI systems. At times, student testing illuminated hidden bias, cultural assumptions, and the “registries of power” (Crawford, 2021, p. 8) encoded into machine systems. Other times, this testing disturbed, disrupted, and destabilized the classroom, accelerating the ethical, emotional, and pedagogical labor of teaching. Through three clusters of cases, this research explores AI-mediated violences, omissions, excesses, blind spots, contradictions, ethical failures, and representational logics surrounding race, gender, and sexuality. The cases articulate how AI systems became entangled with the classroom, putting forward complex questions surrounding safety, ethics, and power. Methodologically, the dissertation employs multi-case study (Stake, 1995; Merriam, 1998), actor-network theory (Latour, 1996; Fenwick et al., 2011), and posthuman frameworks (Haraway, 1990, 2016; Braidotti, 2013) to trace how agency and affect circulate between the human and non-human. The dissertation advances a set of propositions for art education alongside AI, including red teaming as a critical pedagogy, violence literacy as a core competency, glitch as generative inquiry, and a rendering of care that is situated, relational, and necessarily imperfect.