From Present Tense to Future (Im)perfect

This chapter explores how anthropologists can meaningfully contribute to the design of generative AI systems by bringing theory, method, and disciplinary insight into the structure of algorithms themselves. Framed through a linguistic-temporal metaphor—Present Tense, Past Simple, and Future (Im)perfect—the chapter traces the current challenges posed by GenAI, reflects on anthropology’s emergence as a methodological response to earlier moments of systemic harm, and proposes three concrete interventions: improving training data, fine-tuning models, and designing new model architectures based on anthropological theory. Drawing on examples from public sector algorithms, image generation tools, and language models, the chapter shows how biases in data and model behavior can be both reproduced and resisted. Special attention is given to the potential of theory-informed model design, including examples of architectures inspired by kinship systems, resistance strategies, and non-ergodic systems. Rather than positioning anthropology solely as a critical outsider, this chapter argues for anthropologists to participate directly in model development—translating theoretical insights into practical frameworks that make generative AI more equitable, contextual, and accountable.