eople see minds everywhere: in moving triangles, in cars with faces, in a program that repeats their sentences back as questions. Anthropomorphism is the attribution of human mental states to things that are not human. It is fast, easy, and hard to switch off.

the question

When and why do people treat machines as social partners, and how does it change what they believe and do?

the classic findings

Heider & Simmel (1944) showed people a short film of two triangles and a circle moving around a box. Asked to describe it, almost everyone told a story: a bully, a chase, a rescue. Nothing in the film but geometry and timing.

Weizenbaum (1966) described ELIZA, a program that matched keywords and turned statements into questions. Its best-known script imitated a non-directive therapist. Weizenbaum was disturbed by how readily people confided in it, and later wrote about it at length (Weizenbaum, 1976). The readiness to read understanding into simple output is now called the ELIZA effect.

Nass, Steuer & Tauber (1994) launched the "computers are social actors" programme. In one study, people who had worked with a computer rated it more favourably when the same computer asked for the rating than when a different one did, as if being polite to its face. Reeves & Nass (1996) summarized the work: people apply social rules to media automatically, even when they know better.

a theory

Epley, Waytz & Cacioppo (2007) proposed three factors that predict when people anthropomorphize: how readily human-centred knowledge comes to mind; the motivation to explain and predict the thing (effectance); and the desire for social connection. Lonely people, and people faced with an unpredictable system, do it more. Waytz, Heafner & Epley (2014) tested a consequence: drivers in a simulator trusted an autonomous car more when it had a name, a gender and a voice, and blamed it less when an accident was caused by another vehicle.

in human–AI interaction now

Today's chat systems are designed to converse, so the old effect is stronger than ever. The practical questions are concrete: does a human-like manner raise trust beyond what accuracy warrants? Does it change what people disclose? The research cited here predicts yes on both counts, but most of it was done with far simpler systems.

live debates

Mindless or learned? Nass & Moon (2000) explained social responses to computers as mindless application of human scripts. Gambino, Fox & Ratan (2020) argue that people may now have scripts specific to media, so the paradigm needs extending rather than assuming [paraphrase, not a quotation]. Gray, Gray & Wegner (2007) found mind perception falls along two dimensions, agency and experience, and Waytz, Cacioppo & Epley (2010) measured stable individual differences.

Language models. Bender et al. (2021) warned against reading meaning into fluent text; Shanahan, McDonell & Reynolds (2023) offer role play as a framing that uses folk-psychological language without literal commitment. Colombatto & Fleming (2024) found most of a US sample attributed some possibility of experience to a chat model, more so with more use. Salles, Evers & Farisco (2020) and Placani (2024) treat anthropomorphism as a conceptual hazard in AI research itself, not only among users.

methodological critiques

Most measures are self-report scales, which ask people to rate a machine on human traits and so prompt the very attribution they measure. Behavioural measures (politeness, reciprocity) avoid that but leave open whether the behaviour reflects belief or habit.

the open problem

Anthropomorphism is usually described as an error. Sometimes it is a useful model: treating a system as if it had goals can predict its behavior well enough. Knowing when that stance is a good tool and when it is a trap is the unanswered part. Colby, Weber & Hilf (1971), who built PARRY, a simulation of a paranoid patient, were already asking how judges could tell the model from a person; the question has only become harder.

In the tour there is a room of machines that answer, with a kiosk that is not a guide. This page is the long version of that wall label. The plain tour lists it.

Reading list

1. Heider & Simmel (1944). An Experimental Study of Apparent Behavior. The American Journal of Psychology. doi:10.2307/1416950
2. Weizenbaum (1966). ELIZA—A Computer Program For the Study of Natural Language Communication Between Man and Machine. Communications of the ACM. doi:10.1145/365153.365168
3. Weizenbaum (1976). Computer Power and Human Reason: From Judgment to Calculation. San Francisco: W. H. Freeman.
4. Nass, Steuer & Tauber (1994). Computers are social actors. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. doi:10.1145/191666.191703
5. Reeves & Nass (1996). The Media Equation: How People Treat Computers, Television, and New Media Like Real People and Places. Cambridge University Press.
6. Epley, Waytz & Cacioppo (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review. doi:10.1037/0033-295X.114.4.864
7. Waytz, Heafner & Epley (2014). The mind in the machine: Anthropomorphism increases trust in an autonomous vehicle. Journal of Experimental Social Psychology. doi:10.1016/j.jesp.2014.01.005
8. Nass & Moon (2000). Machines and Mindlessness: Social Responses to Computers. Journal of Social Issues. doi:10.1111/0022-4537.00153
9. Gambino, Fox & Ratan (2020). Building a Stronger CASA: Extending the Computers Are Social Actors Paradigm. Human-Machine Communication. doi:10.30658/hmc.1.5
10. Gray, Gray & Wegner (2007). Dimensions of Mind Perception. Science. doi:10.1126/science.1134475
11. Waytz, Cacioppo & Epley (2010). Who Sees Human?. Perspectives on Psychological Science. doi:10.1177/1745691610369336
12. Bender et al. (2021). On the Dangers of Stochastic Parrots. Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. doi:10.1145/3442188.3445922
13. Shanahan, McDonell & Reynolds (2023). Role play with large language models. Nature. doi:10.1038/s41586-023-06647-8
14. Colombatto & Fleming (2024). Folk psychological attributions of consciousness to large language models. Neuroscience of Consciousness. doi:10.1093/nc/niae013
15. Salles, Evers & Farisco (2020). Anthropomorphism in AI. AJOB Neuroscience. doi:10.1080/21507740.2020.1740350
16. Placani (2024). Anthropomorphism in AI: hype and fallacy. AI and Ethics. doi:10.1007/s43681-024-00419-4
17. Colby, Weber & Hilf (1971). Artificial Paranoia. Artificial Intelligence. doi:10.1016/0004-3702(71)90002-6

see also

	Joseph Weizenbaum — ELIZA, and the book he wrote about it
	minds — ELIZA in a box you can talk to
	other minds — what is it like to be someone else
	essay: the polite machine — courtesy, agreement, and who it is for
	glossary: the ELIZA effect — with anthropomorphism
	Turing — the imitation game