Back to feed

The magic in the black box: Geek wizardry and the inherent unpredictability of artificial intelligence (AI)

Jul 2026 · Anthropological Theory · Vol 26, pp. 342 - 362 · 0 citations · 30 references

Abstract

Research on artificial intelligence (AI) frequently highlights the technology's inherent unpredictability – often captured through the metaphor of the ’black box’ – which makes it difficult, and perhaps even impossible, to determine how particular results are produced. These discussions typically culminate in debates about ethics, reliability and responsibility, yet they say relatively little about how those working with and developing AI themselves relate to this mystification. Drawing on extensive ethnographic fieldwork among two distinct groups in the South Indian city of Bengaluru – artists experimenting creatively with AI and data scientists developing AI-powered applications – this article examines how practitioners engage with the technology's perceived opacity. Many interlocutors acknowledged that beyond a certain threshold, they could no longer fully explain how ‘their’ AI systems arrived at particular decisions or outputs. Instead, they frequently invoked expressions such as ‘that's the magic of it’ or described their work as a form of ‘geek wizardry’. Taking these formulations seriously, the article explores how the seemingly ‘unexplainable’ aspects of AI come to be experienced as magical. Rather than dismissing such language as metaphor, it approaches these references as part of the ways practitioners make sense of technologies that appear to exceed their own knowledge and explanatory frameworks. Drawing on Ernesto De Martino's theorization of magic – particularly his emphasis on context, history and meaning – the article argues that references to ‘AI as magic’ help situate rapidly evolving technological systems within everyday lifeworlds. In a field where developments occur at extraordinary speed and even specialists struggle to remain fully up to date, the language of magic provides a way of articulating encounters with systems whose operations remain only partially intelligible. By placing practitioners’ own descriptions of AI within broader anthropological discussions of magic, the article offers a new perspective on how people live with and interpret the opacity of contemporary algorithmic systems.

View source