Last week I came across a fascinating article in The Economist exploring the value systems embedded in today's AI models. Frankly, what captured my attention wasn't the findings themselves. I was not surprised by the value maps of the main AI models, especially given their origins.
What I did find intriguing was something much more fundamental—the fact that the World Values Survey, a framework designed to study human values, can now be applied to a non-human conversational partner.
That realization immediately made me think about the behavioral traps we have been exploring throughout our AI × HI journey, and about the fact that the truly important question here is not whether AI has values, but whether we have realized what that means for us.
Behavioral science has spent decades studying how our thinking is shaped by the people around us.
Parents.
Teachers.
Friends.
Colleagues.
Communities.
The media.
Our social environment has always been understood as something fundamentally human—a context that, in turn, shapes us.
Today, for the first time, something else is joining that environment.
Millions of people now turn to AI not simply to write emails or summarize documents, but to think through difficult decisions, navigate relationships, challenge assumptions, seek reassurance, and ask for advice.
The interaction is more than technological. It is conversational.
AI is becoming socially present, and it is here for the long term. This is where I believe a fascinating asymmetry emerges.
Our social habits are rooted in our familiarity with human-to-human conversations. When another person speaks, we hear their words, but we also hear the individual behind those words—their experiences, personality, values, and worldview.
We instinctively understand that no advice is ever completely value-free, and we naturally interpret it through everything we know about the person offering it.
We might not reject their advice, but neither do we simply accept it.
We filter it.
We calibrate it.
We adjust it.
As part of our social intelligence, we employ psychological safeguards that humans have developed through thousands of years of interacting with one another.
With AI, however, something subtly changes.
While in human conversations we never hear words without imagining the person behind them, with AI there is no person behind the conversation.
The person disappears. The worldview still enters the conversation.
To me, this is another defining asymmetry in the AI × HI relationship. Because there is no visible person, we often fail to activate the psychological safeguards we naturally use when another person's values might influence us.
Instead, we experience the conversation as neutral.
Objective.
Grounded purely in data and reason.
This matters because influence rarely arrives all at once. In fact, it often feels invisible precisely because it is gradual. Very few of our beliefs are shaped by a single conversation or social episode. Instead, they emerge through thousands of small interactions, tiny suggestions, subtle reframings, and micro-moments where one perspective feels just a little more natural than another.
This is why behavioral science treats the social environment as one of the primary forces shaping what we perceive as normal, reasonable, and worth carrying forward.
If AI becomes one of our most frequent, socially present conversational partners, then it inevitably becomes part of that environment. And the greatest influence AI will have on us won't come from having values.
It will come from the fact that, during our interactions, we often don't consciously consider that it has them.
And because of the profound influence of social context on human behavior, one more detail from The Economist article feels worth mentioning. It brings up how governments, AI labs, and technology companies are now actively considering which values are—and should be—embedded in AI models.
There is something remarkable about this.
For centuries, societies debated what should be taught in schools, written in books, published in newspapers, or broadcast on television.
Today, we are debating the worldview of something that millions of people converse with every day, and that is rapidly becoming part of the social environment shaping our behaviors.
This makes me think that the next frontier of AI literacy, beyond learning to write better prompts, will be learning to become more intentional about the AI × HI relationship itself.
We will need to recognize that every conversation has a voice carrying a worldview—even when no person stands behind it.
Understanding AI may ultimately become less about understanding the machine and more about understanding how humans behave when faced with something we have never encountered before: a worldview without a person.
References:
The Economist. (2026, June 25). AI models' values are very different from most people's.
World Values Survey. World Values Survey project. (The framework discussed in the article.)
🗣 ME (90%): The ideas and the answers to the many questions my AI collaborator raised, the behavioral perspective, and the pivots in the main perspectives.
🤖 AI (10%): Brainstorming questions, amplification with some flattering comments, stress-testing & editorial refinement.