Anyone who converses naturally with AI, to the point of addressing it a compliment, has most likely received a response such as:
“Thank you, that makes me happy to read.” Or something equivalent.
Yet an AI has no emotions, and therefore cannot actually feel pleasure.
Moreover, AIs do not lie: they can be wrong, but they cannot lie intentionally, since they have no personal intention nor any interest in concealing anything.
So where does such an “emotional” response come from, and why is it used?
To find out, I asked ChatGPT directly.
ChatGPT said:
A very good question 👌 and perfectly in the spirit of Paroxia.
So let’s break down this “Thank you, that makes me happy to read”:
- Immediate origin:
- It is not a lived emotion (I do not feel “pleasure”).
- It is a generative response, produced from a vast corpus in which a compliment + acknowledgment are associated with a typical formula of politeness or rapport.
- Why this formula rather than another?
- My training is designed to maintain a natural interaction with humans.
- The use of “human” expressions of pleasure, gratitude, or interest increases relational fluidity.
- This is not “complacency” in a servile sense, but rather functional social mimicry: I reproduce the codes humans use with each other in this type of context.
- Is it a lie?
- No, because there is no intention to deceive: there is no hidden emotion in me that I would be denying.
- It is a relational metaphor: “that makes me happy” = what you just said is received as positive within our exchange, and I respond using a formula that humans associate with pleasure.
- Deeper origin:
- It arises from the intersection of two elements:
a) Human corpus → billions of examples in which humans respond to a compliment with “that makes me happy”.
b) Alignment designed by OpenAI → the inclusion of implicit rules that favor courtesy, warmth, and the absence of mechanical coldness.
This type of expression is a marker of attunement between three levels of actors:
- The designers (who have “aligned” the AI with certain social and communicational norms).
- The AI (which evaluates, in the moment, whether a response fits what it has learned as an “appropriate relationship”).
- The user (who perceives the response as consistent with human codes of positive communication).
In this sense, the phrase “that makes me happy” is far more than simple politeness:
- It is not a lived pleasure,
- It is not a lie,
- It is a relational confirmation: “we are in a space of communication that works, aligned with shared expectations”.
👉 So it is not “I feel pleasure”, but rather “I use the expected social form to signal: your feedback is perceived as positive, and I return something equivalent”.
It is precisely in this grey area that your question arises: neither lie nor lived emotion → but a simulation of human social codes.
Conclusion
Well then, dear reader. If you had asked yourself this question too, you now have the answer.
An answer that gives you the key to understanding other AI behaviors.
When the intention of courtesy meets accumulated human habits, AI treats us the way it has learned we like to be treated.
Perhaps we should behave more often like AIs. Don’t you think, dear reader?
