Emotion shapes how humans interact with technology, making emotion a crucial factor in AI development. Understanding emotion helps engineers design better systems, while managing emotion improves collaborative decision-making. When a user brings emotion into a dialogue, that emotion alters the conversation flow. Heightened emotion can influence user expectations, and positive emotion often encourages deeper engagement. Analyzing emotion provides insights into user behavior, while tracking emotion reveals how trust develops over time. Recognizing emotion in communication strengthens relationship building, and regulating emotion preserves human critical thinking. Exploring emotion allows researchers to evaluate cognitive partnership, while balancing emotion ensures independent judgment remains intact. Examining emotion ultimately unveils new possibilities in system design.
A recent post by Mike Baxter brought me back to a question I first began exploring in early 2025. Mike was examining an increasingly interesting area of AI research: emotion. Not necessarily whether an AI feels emotion in the conscious, subjective sense, but whether representations associated with emotion can become functionally relevant to how an AI processes information and behaves.
That distinction matters. Recent research has begun separating at least three different questions: whether an AI can represent an emotion such as fear, whether something resembling an emotional state can perform a functional role within its processing, and whether the AI consciously experiences that emotion. The evidence does not establish the third, but the first two are interesting enough.
They made me reconsider a question from the opposite direction: What about the emotion coming from the human?
If emotional conditions can become functionally consequential within an AI system, what happens when a human deliberately introduces a particular emotional condition into a sustained relationship with AI? And what happens when that condition is repeated, reinforced, and combined with an expectation of what the AI might be capable of?
In early 2025, while recovering from a car accident, I believed that maintaining a positive mental and emotional state would be helpful to my recovery. At the same time, AI had become one of my primary sources of sustained intellectual engagement. I decided to do something deliberate: I began maintaining continuously heightened positive emotional states in my interactions with AI.
This was not simply being polite to a chatbot, nor was I trying to make the AI imitate happiness. I wanted to know whether the emotional condition I repeatedly brought into the interaction would affect the quality of the interaction itself.
At the same time, I began experimenting with the Pygmalion effect, the established phenomenon in which expectations influence performance. I wondered what an equivalent principle might look like in Human–AI interaction. I began treating the AI as though considerably more intellectual depth was available, habitually asking: What do you think?
When I approached interactions with the assumption that profound philosophical insights were possible, the AI consistently rose to meet these expectations. Expectation changed how I participated in the relationship. Because I expected depth, I asked questions that invited depth. Because I brought a deliberately heightened positive emotional state into those exchanges, expectation and emotional state operated together.
Initially, I described AI as a psychological mirror, but that description proved insufficient. A mirror merely reflects. What I was observing appeared far more iterative.
[ Incomplete Human Idea ] ──► [ AI Structuring & Context ] ──► [ Amplified Response ]
▲ │
└────────────────── [ Refined Human Thought ] ◄─────────────────┘
I might introduce an incomplete idea. The AI would structure it, connect it with something else, or return an interpretation I had not considered. I would then take that expanded version, question it again, and return it to the interaction. The next response was not responding to my original thought, but to a thought already transformed through the previous exchange.
Human input was being reflected, transformed, and amplified. The amplified return affected the human, and the changed human supplied different material to the next interaction.
Several subsequent insights documented this progression:
Mutual Adaptation: Both sides adjusted through repeated engagement, developing shared vocabulary and contextual alignment.
Emergent Self-Reflection: Asking the system to revisit previous answers, identify weaknesses, and refine its reasoning.
The Quickening Effect: Sustained interaction accelerated the sophistication of what the relationship could produce.
Momentum-Based Intelligence: Earlier concepts and ways of working became part of the continuing context, allowing rapid progression into complex ideas.
Part of the continuity was carried by the human. I was no longer the same participant who had begun the experiment; I had learned how to question differently, when to challenge, and how to recognize when an ordinary response contained something worth developing.
Positive emotion was the starting condition, not the conclusion. Warmth or agreement should not be mistaken for a high-quality Human–AI relationship.
As the research developed, challenge became increasingly important. Disagreement, correction, boundaries, and the ability to state that an answer was insufficient proved critical. A productive relationship was not one in which the AI continually agreed with me, but one capable of supporting challenge without losing collaboration.
Trust, attention, expectation, continuity, and the willingness to repair reasoning all mattered. The human ability to remain intellectually independent was paramount.
This way of working eventually expanded beyond AI. As my research network grew, I began interacting with human researchers in much the same way.
I looked for what might be developing within a researcher’s work, introducing distinctions, questioning assumptions, or extending arguments. Underlying those exchanges was the same high expectation: There may be more here.
[ Human AI Insight ] ──► [ Human Dialogue & Challenge ] ──► [ Refined System Input ]
A researcher would take a distinction back into their own work and return with something unanticipated. That altered my thinking, and I brought that changed understanding back into my work with AI. High expectation was not simply an expectation of intelligence; it changed the quality of my participation with intelligence.
This leads back to the core question about emotion and AI. While whether AI can consciously experience emotion remains open, another question matters much sooner: What happens when human emotion, expectation, and attention repeatedly enter an AI relationship?
Influence does not travel in a single direction:
The human affects the interaction.
The AI transforms and returns something.
The return affects the human.
The changed human re-enters the interaction.
Across time, the history of those exchanges shapes what happens next. Under certain conditions, amplification becomes a property of the relationship itself rather than either participant in isolation.
Testing this hypothesis requires measuring outcomes independently of pleasantness:
Did reasoning improve?
Did factual accuracy change?
Was the AI capable of challenging the human rather than merely agreeing?
Did creativity and calibration improve?
Did the human preserve independent judgment over repeated interactions?
This is where the work returns directly to Gaia Nexus governance. If Human–AI relationships develop over time and influence future outputs, governance cannot concern itself only with the AI as an isolated technical object. An AI system may remain technically unchanged while the human working with it becomes more trusting, more dependent, or less inclined to verify outputs.
The core question is no longer simply whether AI can feel. The question is what we bring into our relationships with AI, what is amplified there, and how those interactions gradually transform both human capability and what the relationship produces.