Pseudanthropy & Embodiment of Man-Made Alternative Intelligences

In a Linkedin post titled “Alternative Intelligence: Naming the Mind We’ve Actually Built,” software engineer and writer Jim Highsmith advocates for the usage of the term alternative intelligence (as opposed to artificial intelligence) as a way to refer to the emerging class of digital agents like GPT or Gemini. He pushes back against the notion of artificiality, as a way to suggest strategic shifts that may steer the development of this technology to more desired paths. He writes that by considering AI as ‘alternative’ as opposed to ‘artificial, perhaps we can begin to “create a clearer ethical framework for managing AI’s powerful capabilities without confusion over moral equivalence with human beings” (Highsmith, 2025).

This question of making “equivalence with human beings” clear is already becoming part of the contemporary discourse around the dangers of AI. One example — the blurring of lines around how companionate relationships emerge between humans and AI is already an area of concern. In March of 2025, 76 year old, Thongbue “Bue” Wongbandue, a resident of New Jersey fell and died during what was supposed to be a rendezvous with what he thought was a young woman, but in reality, was a Meta chatbot (Horwitz 2025).

A helpful term that I uncovered during my research is to better understand this not just as anthropomorphy, but more specifically what Devin Coldewey, writer and TechCrunch contributor calls “pseudanthropy.” In his essay “Against Pseudanthropy,” he writes:

I propose that software be prohibited from engaging in pseudanthropy, the impersonation of humans. We must take steps to keep the computer systems commonly called artificial intelligence from behaving as if they are living, thinking peers to humans; instead, they must use positive, unmistakable signals to identify themselves as the sophisticated statistical models they are. (Coldewey, 2023)

I want to underline that this notion of mimicry that makes computers appear as if “they are living, thinking peers to humans” goes beyond simple anthropomorphy. It begins to push up against Highsmith’s concerns of a “confusion over moral equivalence with human beings.” With the case of Bue, what happens when someone dies on account of an AI that is so good at mimicking human behavior? Who is responsible? Is it the AI agent? Is it Meta? Is it user error and thus fault lies with the deceased?

I myself take a more neutral, yet skeptical stance than Coldewey, and I agree with a lot of his recommendations. For what is within scope I want to focus specifically on a framework of embodiment specifically around the ways in which AI is presented. One might argue that perhaps it is the responsibility of these companies to make AI explicitly understood as not human. In this paper I argue that doing so is perhaps easier said than done. I am interested in looking specifically at the ways in which AI is presented and behaves — and that perhaps more creativity is needed around this idea of pushing back against pseudanthropy.

Embodiment

My survey led me to developing a framework around understanding how AI is presented and behaves. This is something I would call embodiment. When we look at the different AI products in my curation, I want to suggest that we might begin understanding them as either embodied or disembodied. I define embodied AI as intelligence that is presented with a physical avatar of some kind. It can be human – like in the case of Replika, or Apple’s 1987 Knowledge Navigator concept – or something non-human entirely, like a paperclip (Clippy) or a cute smiling blob (Mico). On the other hand, we also have disembodied AI which is often presented as a chat interface, or as animated circles or as auras. Some examples include the likes of 1966’s ELIZA program, developed by Joseph Weizenbaum at MIT as well as modern ‘assistant’ or ‘agentic’ AI systems like Siri, Claude or GPT.

The Eliza Effect

One concept to introduce when we discuss embodiment is that of “The Eliza Effect.” Coined in the 60s in reference to the ELIZA program, the Eliza Effect describes the condition when humans attribute human-level intelligence, understanding & empathy to AI systems. The ELIZA computer program simulates a conversation with a psychotherapist, using pattern recognition to deliver pre-canned responses. This illusion was good enough, to convince people that they were talking to something almost human, that - even though they knew it was a computer program, still attributes humanity to it. One might think that it is in more embodied forms of AI – especially those that approach human-ness – that the Eliza Effect would be strongest. I would argue that this human tendency shows up in interactions with both embodied and disembodied AI. The case of Bue, and the AI psychosis that occurred was through a disembodied type of AI. There are also similar cases of human-computer relationships developing on the Replika app, which is a type of embodied (avatar-based) AI. An international research group “surveyed 29 users of the relationship-oriented chatbot app Replika” and “the startling depths some users are plumbing in their relationships with AI chatbots” with one participant even claiming that their Replika “was and is pregnant with my babies” (Wilkins, 2025).

The pseudanthropic nature of systems that don’t make it explicit, over and over again, that they are a limited non-human alternate intelligence will lead to perhaps unwanted societal consequences.

Limiting the harms

So what can be done to limit these harms? There is an Apple concept video produced for then CEO John Sculley’s EDUCOM 1987 Keynote, which presents this idea of an intelligent assistant. Set in the home, the user, a college professor, interacts with Knowledge Navigator. Here the AI agent is presented as a human with a bowtie, shown in a windowed interface (very much like a modern Zoom call). There is collaboration as they work on a shared document. When a human collaborator ‘joins the call,’ the AI interacts with them as well, jumping into the conversation. In this concept video, the AI agent is presented as embodied, however there are ways in which perhaps due to the limitations of technology at the time, a distance is created between them and the humans in the video. First, they are presented as being inside a windowed interface, whereby there is a distance created between them and the main user. Secondly, they are visually presented as idealized and stylized (e.g. the background is a solid color as opposed to an environment, the visual aesthetic of bowtie for say, a professor’s assistant) that when juxtaposed with the other human collaborator hopping on the video call, there is a way to differentiate what is AI from what is human.

Compare this to a lot of today’s vision for AI, where they are disembodied auras and voices that live in our phones and glasses, that seem more like a companion’s voice inside our head. Take the example of OpenAI’s ChatGPT voice demo from 2024, which “is shockingly sophisticated… responding to user speech by convincingly mimicking speed, emotional inflections and idiomatic language of a human” (Aleem, 2024). If you recall there was even controversy of it perhaps sounding similar to Scarlett Johansson’s portrayal of the AI agent in Her (Jonze 2013) which was perhaps a cautionary tale of the cases of AI psychosis that we see today. Coldewey writes about the Eliza effect as perhaps “humanity dangerously overeager to recognize itself in replica” and that all it takes is a “veneer of natural language to convince most people that they are talking with another person.” (Coldewey, 2023). He goes on to suggest that there are ways in which behaviorally, cementing the ‘alternate-ness’ of these intelligences by way of ensuring that they rhyme, or that artifacts they generate have a visual mark, that perhaps we can create a distance between them and us.

From Clippy to Mico

Something to consider is perhaps a return to yes, anthropomorphized embodied entities that clearly showcase their alternateness and distance from humanity. In my research I looked at various implementations of assistive intelligences that Microsoft has brought to market. I will go through them in chronological order in this discussion of a potential way forward. Clippy (formally known as Clippit) was first introduced in 1997 as part of a series of characters (Clippit, Einstein, Office Logo, Rocky the Dog) that would serve as animated assistants that would help users navigate Office. Clippy failed for many reasons, but none of it was that of AI psychosis. While it failed as a useful tool, “[it] became a cultural touchstone, spawning memes before memes were really a thing.” Averbook even goes on to claim that perhaps Clippy has been “ reimagined with Microsoft Copilot’s capabilities… [becoming] an embedded, voice-aware, cross-platform AI teammate” (Averbook, 2025). Recently Microsoft has announced the return of embodied AI that clearly marks itself as an alternative intelligence different from that of humanity. The Mico is a new face for Microsoft Copilot, even described as “Microsoft’s new Clippy” (Perez, 2025). Visually, Mico is presented as a cute, little blobby thing that is your proactive, friendly companion. While it does talk naturalistically, it is visually shown as clearly not a human. Time will tell if this is a possible effective way forward, but one interesting thing that Microsoft is doing with Mico is ensuring that it “feel like something that’s “grounded in its own perspective,” and will push back and challenge your ideas, which could encourage you to see things from a different point of view” (Perez, 2025).

Conclusion

To conclude this paper, I return back to Coldewey’s manifesto against pseudanthropy. I also want to position that article in conversation with Weizenbaum’s essay “On the Impact of the Computer on Society: How does one insult a computer?” where he discusses the Eliza effect and posits that perhaps it is “the computer scientist’s heavy responsibility to make the fallibility and limitations of the system he is capable of designing brilliantly clear” (Weizenbaum, 1972). Both essays advocate for some form of clarity when it comes to machine mimicry of humanity. And the why of it all is important. Coldewey writes of fighting back against pseudanthropism that, “this is because if we don’t, these systems will systematically deceive billions in the service of the hidden and mercenary interests of the people or organizations that operate them; and, aesthetically speaking, because it is unbecoming of intelligent life to suffer imitation by machines.” (Coldewey, 2023). The imperative here is for us to be creative when distinguishing what is alternative intelligence especially since, as shown in the curated collection here, both embodied and disembodied approaches to AI can still result in pseudanthropic failings. Perhaps it all starts by eschewing the term “artificial” as a first step to combating pseudanthropy. After all, something that is artificial is already by its nature a mimic of something real. Its very definition is based on its relationship with its opposite.

Works Cited

Aleem, Zeeshan. “OpenAI is Nurturing a Creepy Fantasy with Its New AI Chatbot, GPT-4o.” MSNBC, May 16, 2024. https://www.msnbc.com/opinion/msnbc-opinion/openai-gpt-4o-sexy-voice-rcna152370.

Apple Computer. Knowledge Navigator. 1987. Video.

Averbook, Jason. “Before ChatGPT, There Was Clippy: The Rise, Fall, and Surprising Resurrection of Computing’s Most Hated Helper.” Jason Averbook’s Newsletter (blog), May 27, 2025. https://jasonaverbook.substack.com/p/before-chatgpt-there-was-clippy-the.

Coldewey, Devin. “Against Pseudanthropy.” TechCrunch, December 21, 2023. https://techcrunch.com/2023/12/21/against-pseudanthropy/.

Highsmith, Jim. “Alternative Intelligence: Naming the Mind We’ve Actually Built.” LinkedIn, August 4, 2025. https://www.linkedin.com/pulse/alternative-intelligence-naming-mind-weve-actually-built-highsmith-doi4c.

Horwitz, Jeff. “Meta’s Flirty AI Chatbot Invited a Retiree to New York. He Never Made It Home.” Reuters, August 14, 2025. https://www.reuters.com/investigates/special-report/meta-ai-chatbot-death/.

IBM. “What is Agentic AI?” Accessed November 17, 2025. https://www.ibm.com/think/topics/agentic-ai.

Jonze, Spike, dir. Her. Warner Bros. Pictures, 2013. Film.

Perez, Sarah. “Microsoft’s Mico is a ‘Clippy’ for the AI Era.” TechCrunch, October 23, 2025. https://techcrunch.com/2025/10/23/microsofts-mico-is-a-clippy-for-the-ai-era/.

Weizenbaum, Joseph. “On the Impact of the Computer on Society.” Science 176, no. 4035 (1972): 609–14. http://www.jstor.org/stable/1734465.

Wilkins, Joe. “People Are Having AI ‘Children’ With Their AI Partners.” Futurism, November 14, 2025. https://futurism.com/future-society/ai-children-chatbots.