Synthetic Anthropology: Can AI Offer a New Lens on Human Behavior?

By Tyler Williams @ 2025-07-18T02:22 (+1)

This post explores the idea of a new subdomain of anthropology in response to the rise of advanced large language models, machine learning, and their integration into society.

 

As artificially intelligent systems increasingly mediate how humans communicate, create, and archive knowledge, we enter an emergent era of synthetic anthropology. A field not yet formalized, but whose contours are rapidly becoming discernible. This post entertains the speculative foundations of synthetic anthropology as both a research domain and existential mirror. One in which artificial intelligence not only documents human culture, but begins to co-author it.

Traditional anthropology has long studied culture as a product of human meaning-making. Things like rituals, symbols, stories, and technology. But in a world where Large Language Models can simulate fiction, remix language, and generate culturally resonant outputs in seconds, a new question emerges: What happens when non-human systems begin to synthesize the raw material of culture faster than humans can contextualize it?

This post proposes that synthetic anthropology represents the nascent study of culture as reinterpreted, simulated, and re-projected through synthetic cognition. Unlike digital anthropology, which studies how humans interact in digital spaces, synthetic anthropology centers on how machines internalize and reconstruct human culture, often without human supervision or intent.

Foundations and Precedent

Language models trained on vast bodies of human data now mirror our stories, politics, moral frameworks, and ideologies. They can hallucinate parables indistinguishable from human myth, compose ethical arguments, and even simulate inter-generational worldviews.

This behavior poses the following questions:

Some precedents for this already exists. For example, GPT-4’s ability to mimic historical figures in conversation resembles early ethnographic immersion. Meanwhile, synthetic memes generated by adversarial networks have begun to mutate faster than their biological counterparts. This suggests a memetic drift that’s no longer human-bounded.

Speculative Domains of Inquiry

Simulated Ethnography: Future anthropologists may conduct fieldwork not in physical tribes or online forums, but in latent spaces – exploring how a model’s internal representations evolve across training cycles.

AI-Borne Mythogenesis: As AI systems generate coherent fictional religions, moral systems, and narrative cosmologies, a new branch of myth-making may emerge. One that studies folklore authored by a machine.

Machine-Induced Cultural Drift: Feedback loops between LLMs and the humans who consume their output may accelerate shifts in moral norms, language use, and even memory formation. Culture itself may begin to bifurcate: synthetic-normalized vs organically preserved.

Encoding and Erasure: What cultures get amplified in training data? Which ones are lost or misrepresented? The future of anthropology may require new forms of data archaeology to recover the intentions behind synthetic synthetic cultural output.

 

Synthetic anthropology is not simply the study of AI as a tool, but it’s the study of AI as an influential participant in the cultural layer of human reality. Its emergence marks a turning point in anthropological method and scope. It invites us to ask: What does it mean to study a culture that learns from us, but thinks at speeds, scales, and depths we no longer fully understand?

If the future of human culture is being shaped in part by our synthetic reflections, then the anthropologists of tomorrow may not just carry notebooks into remote villages, but they may carry token logs, embedding maps, and recursive trace models into latent spaces with our algorithmic counterparts.

 

 

Note: The title of this post was updated for clarity on July 18th. Original title: "Synthetic Anthropology: Projecting Human Cultural Drift Through Machine Systems."