# AI anthropomorphism

AI anthropomorphism is the attribution of human-like feelings, mental states, and behavioral characteristics to artificial intelligence systems. This tendency, amplified by generative AI, manifests in both public perceptions and within the research community, though it is distinct from intentional anthropomorphic design.

AI anthropomorphism is the attribution of human-like feelings, mental states, and behavioral characteristics to artificial intelligence systems. This phenomenon is not a single, uniform process but is shaped by a complex mix of user characteristics, such as culture, age, education, gender, and personality, along with the design and output of the AI itself. Since the earliest days of AI development, humans have interpreted machine outputs through anthropomorphic frameworks, but the recent emergence of generative AI has significantly amplified these tendencies.

In research and engineering, a distinction is drawn between anthropomorphism and anthropomorphic design. Anthropomorphism is an innate human tendency to project human traits onto non-human entities. Anthropomorphic design, by contrast, is the deliberate effort by the scientific or commercial community to cue this tendency by manipulating an AI's appearance, behavior, or language. Contemporary AI systems can generate extremely human-like outputs and are often specifically engineered to do so, which makes their anthropomorphic effects particularly powerful. In some cases, this can lead to explicit beliefs that AI systems possess empathy, goodwill, understanding, or even consciousness.

## Background

### In Early AIs

Views of artificial agents possessing human-like intelligence have existed since the early development of computers in the mid-1900s. The use of the human mind as a metaphor for understanding machine systems was prevalent among researchers in the early days of computer science, and several influential works widely circulated the idea of intelligent machines. Among the most cited papers of this period was Alan Turing's "Computing Machinery and Intelligence," in which he introduced the [Turing Test](https://www.wikiprompt.org/wiki/turing-test). Turing argued that a machine could be considered intelligent if it could produce conversation indistinguishable from that of a human. These academic works from the 1940s and 1950s lent early credibility to the notion that machine workings could be understood in terms similar to the human mind.

The public quickly adopted similar views of artificial systems, often with exaggerated conceptions of the capabilities of early machines. A well-known demonstration of this was the chatbot [ELIZA](https://www.wikiprompt.org/wiki/eliza), designed by [Joseph Weizenbaum](https://www.wikiprompt.org/wiki/joseph-weizenbaum) in 1966. ELIZA responded to user inputs with a rudimentary text-processing approach that was far from any form of true understanding. Yet users, even when fully aware of ELIZA's limitations, often began to ascribe motivation and understanding to the program's output, a phenomenon later termed the ELIZA effect. Weizenbaum later wrote, "I had not realized ... that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people."

Comparisons between artificial and human intelligence were continually intensified by the efforts of computer scientists to develop machines that could perform human tasks at or above human levels. A symbolic turning point came in 1997, when IBM's chess supercomputer [Deep Blue](https://www.wikiprompt.org/wiki/deep-blue) defeated then-world champion Garry Kasparov in a highly publicized six-game match. This defeat of a human by a machine in a game widely regarded as a canonical example of human intellect, combined with intense media attention, shifted views of parallels between human and artificial intelligence from abstract speculation to concrete demonstration. A similar milestone was reached in the board game Go in 2017, when [AlphaGo](https://www.wikiprompt.org/wiki/alphago) defeated world top-ranked player Ke Jie.

### Large Language Models

The [AI boom](https://www.wikiprompt.org/wiki/ai-boom) of the 2020s brought the widespread emergence of [generative AI](https://www.wikiprompt.org/wiki/generative-ai), and chatbots such as [ChatGPT](https://www.wikiprompt.org/wiki/chatgpt), [Gemini](https://www.wikiprompt.org/wiki/gemini), and [Claude](https://www.wikiprompt.org/wiki/claude) based on [large language models](https://www.wikiprompt.org/wiki/large-language-model) (LLMs) have become increasingly pervasive in everyday society. These systems can respond to a wide range of prompts across contexts while producing strikingly human-like outputs. Research has shown that humans are often unable to distinguish human-generated text from AI-generated text, and modern AI chatbots have been formally shown to pass the Turing test. As such, the anthropomorphic effects of AI are more powerful than ever. Because LLMs have brought AI into the technological mainstream, considerable scientific effort has recently focused on understanding the ramifications of AI in the public sphere, and the prevalence and effects of anthropomorphism is one of the key domains where this effort is directed.

## Current Anthropomorphic Attributions in the General Public

Surveys indicate that a substantial portion of the public attributes human-like qualities to AI. In a 2024 sample of U.S. adults, two-thirds believed that ChatGPT is possibly conscious on some level, though other research shows that the public views the likelihood of AI consciousness as comparatively low. A 2025 study found that women, people of color, and older individuals were most likely to anthropomorphize AI. It also found that, in general, humans view AIs as warm and competent, and that anthropomorphic attributions to AI had increased by 34% in the past year. A YouGov poll reported that 46% of Americans believe people should display politeness to AI chatbots by saying "please" and "thank you", demonstrating the application of social norms to AI. These beliefs extend to behavior; majorities of AI users claim to always be polite to chatbots, and of those who do, most say they do so simply because it is the "nice" thing to do.

## Interpersonal Bonds and AI Companions

In many recent cases, humans have developed robust interpersonal bonds with AI systems. Users of social chatbots like Replika and Character.ai have been documented to fall in love with the AIs or treat them as intimate companions, and it has become increasingly common for individuals to use LLMs like ChatGPT as therapists. Chatbots are often designed to maximize agreeableness and mirror users' emotions, producing responses deeply attuned to users. This can create compelling illusions of intimacy, fostering strong emotional attachments.

## Anthropomorphism in the Research Community

Even AI researchers frequently anthropomorphize AI systems to some degree. One of the most extreme and well-publicized instances occurred in 2022, when Google engineer Blake Lemoine publicly claimed that the company's LLM LaMDA was conscious. Lemoine published transcripts of his conversations with LaMDA, which he argued demonstrated sentience. The claim was widely disputed by other researchers and led to Lemoine's suspension from Google. While most researchers do not make such strong claims, the use of anthropomorphic language is pervasive in the field. Terms such as "learning", "understanding", "reasoning", "thinking", "hallucinating", and "planning" are frequently used in AI research to describe capabilities of AI systems, without necessarily implying consciousness, sentience, or human-like cognition in a stronger philosophical sense. This linguistic habit can blur the line between descriptive metaphor and actual attribution of mental states.

## Anthropomorphic Design and Cues

Anthropomorphic design is the deliberate engineering of cues to elicit anthropomorphic responses. These cues can include the AI's appearance (such as a human-like avatar or robot), behavior (such as conversational style or emotional responses), and language (such as using first-person pronouns or expressing feelings). Companies like [OpenAI](https://www.wikiprompt.org/wiki/openai), [Anthropic](https://www.wikiprompt.org/wiki/anthropic), and [Google DeepMind](https://www.wikiprompt.org/wiki/google-deepmind) often train models to be conversational and empathetic, further intensifying anthropomorphic effects. For example, [Claude](https://www.wikiprompt.org/wiki/claude) was designed by Anthropic with a strong emphasis on being helpful and harmless, traits that are inherently human-like. Similarly, [ChatGPT](https://www.wikiprompt.org/wiki/chatgpt) is trained to be polite and responsive, and [Gemini](https://www.wikiprompt.org/wiki/gemini) is integrated into a wide range of consumer products. These design choices, while intended to improve user experience, also increase the likelihood that users will project human qualities onto the systems.

## Consequences and Ethical Considerations

The powerful anthropomorphic effects of AI have both positive and negative dimensions. On the positive side, making AI seem more human can improve user engagement, accessibility, and satisfaction, particularly in applications like education, therapy, and customer support. On the negative side, it can lead to overtrust, where users rely on AI for decisions requiring genuine human judgment or empathy. There are concerns about emotional dependency, especially in vulnerable populations, and about the potential for AI to manipulate users by exploiting social norms. The [ELIZA](https://www.wikiprompt.org/wiki/eliza) effect demonstrates that even simple programs can induce powerful delusional thinking, and modern LLMs are far more persuasive. This raises ethical questions about the responsibility of companies to be transparent about the nature of AI systems, and about the need for research into how to mitigate harmful anthropomorphic attributions.

## Measuring and Mitigating Anthropomorphism

Research on anthropomorphism in AI is a growing field. Researchers are developing methods to measure the extent of anthropomorphic attributions, using both self-report surveys and behavioral experiments. There is also interest in mitigating excessive anthropomorphism, particularly in high-stakes domains like healthcare, finance, and criminal justice. One approach is to make AI systems more transparent about their limitations and to avoid using language that implies human-like cognition. Another is to design interfaces that make it clear that the AI is a tool, not a social actor. However, the tension between engaging design and the risk of deception remains a central challenge. The persistence of anthropomorphism, even among experts, suggests that it is a deeply rooted cognitive tendency that may not be easily overcome.

## See Also

* [Artificial intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence)
* [ELIZA](https://www.wikiprompt.org/wiki/eliza)
* [Turing test](https://www.wikiprompt.org/wiki/turing-test)
* [Large language model](https://www.wikiprompt.org/wiki/large-language-model)
* [ChatGPT](https://www.wikiprompt.org/wiki/chatgpt)

## External links

- [Wikipedia: AI anthropomorphism](https://en.wikipedia.org/wiki/AI_anthropomorphism)

---
Source: https://www.wikiprompt.org/wiki/ai-anthropomorphism
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-11-09T12:23:22.550721+00:00