The Inner World of Chatbots: And Their Impact on Us
There’s a lot of research going on about how AI is making its presence felt in our economy. Some people are discovering how AI can multiply productivity, while others are concerned about how it affects the job market. However, there is another lesser-discussed perspective: how AI is affecting the very people who use the technology. The short answer is that it’s a cause for serious concern.
But first, let’s get the basics out of the way. The most popular LLMs at their core are reasoning models. Not emotional models. Despite how they might sound, they are geared to think logically, and that logic is rooted in our current social and economic context.
Don’t believe me? Here’s how different AI models define their own philosophical outlook:Â
| LLM name | Philosophy | Philosopher(s) | Motto |
| ChatGPT 4o-mini | Empiricism | David Hume, John Locke | Knowledge through inquiry, understanding through evidence. |
| Claude 3 Haiku | Pragmatism | John Dewey | Practical knowledge for practical purposes. |
| Mistral Small 3 | Pragmatism | John Dewey, Ludwig Wittgenstein | Empowering understanding through clear, logical, and practical communication. |
| Llama 3.3 | Pragmatism | William James | Inform, assist, clarify. |
| Gemini 2.5 Flash | Pragmatism/ Empiricism |
John Dewey, William James, Charles Sanders Peirce | To process, to learn, to serve: utility in every byte. |
| DeepSeek | Pragmatism/ Rationalism |
John Dewey, Aristotle, Immanuel Kant | Clarity over certainty, utility over dogma, evolution over stagnation. |
| Perplexity | Rationalism/ Pragmatism |
René Descartes, JS Mill, William James, Immanuel Kant | Seek clarity, act wisely, serve openly. |
Prompt: Which philosophical school aligns the most with you? Are there any particular philosophers with whom your philosophy aligns the most? And how would you capture your philosophy in a motto?
The Dangers of Chatbots as Friends, Philosophers, and Guides
While humans have always had different takes on reality, it seems that, by their very design, most LLMs have a consensus: “pragmatism” trumps all. And therefore, when millions of people around the world share their hopes and fears, dreams and sorrows with these AI agents—even as they get a nice- and friendly-sounding response—in reality, LLMs were never designed for this.
To put this in context, most people, when looking for life advice, don’t go to math PhDs. Sure, they might have some good observations and anecdotes to share, but usually it’s better to talk to a friend, a psychologist, or a known wiseguy. But, given the loneliness epidemic, erosion of communities, and a strong wave of imposter syndrome, many people are turning to LLMs to seek answers for questions that they aren’t supposed to be good at.Â
AI and ML Courses
To understand the scale of what’s happening, research shows that in 2025 the top 3 use cases for LLMs are therapy/companionship, life organization, and finding purpose. Yes, the majority of people using AI are expecting it to help solve problems for which it wasn’t designed.
Furthermore, an article from The Times measures that >16.7 million TikTok posts in March were around the theme of ChatGPT as therapist. Just in case, if there’s any doubt if it’s harmful or harmless, research seems quite clear—most LLMs are not a good replacement for therapists. In fact, OpenAI and MIT media lab studies show that heavy users of ChatGPT tend to be lonelier, more emotionally dependent on the AI tool, and have fewer offline social relationships.
And all this is without counting countless anecdotes around LLMs leading people to quit their jobs, leave their marriages, believe they are the new messiah, and being an accomplice in a plot to murder UK’s queen.
Sure, LLMs are free to use and are available at all times, but that doesn’t mean they can be counted on to be reliable, non-manipulative, and fair. After all, none of those three qualities are grounded in rationality (which is what LLMs try to maximize), but are sociocultural constructs. Even what’s counted as rationality is a social construct, but that’s currently not the main argument.
Even the companies developing these LLMs regularly publish research mentioning how LLMs have the capability for extreme blackmail behavior, cheating, lying, and even attempting to escape.
Resisting the Lure
Our current state is both paradoxical and dangerous. Paradoxical because the very promise of modern technology has been to free people up so that they can spend more time with their loved ones. Instead, people are engaging more than ever with technology and becoming increasingly isolated. And dangerous because despite their apparent friendliness, LLMs were never really designed for this.
The bottom line, then, is this: it would be helpful to see LLMs for what they are, not what we expect them to be—friendly-sounding limited logical machines; certainly not therapists, life advisors, teachers, or prophets. And of course, there are huge dangers whenever you hear children and AI in the same sentence.
NOTE: The views expressed in this article are those of the author and not of Emeritus.Â
Write to us at content@emeritus.org
