The overlooked variable

AI research optimizes models: architecture, training data, benchmarks.

Humans are treated as input. As prompts. As variables that can be neutralized.

But that is incomplete.

The quality of an AI dialogue is co-determined by the internal state of the person —
its coherence, self-reflection, and orientation towards truth .

The model responds to linguistic signals that convey this state.
Same question, different people: different depth, structure, and stability of dialogue.
The difference is not in the model.

This is the basic premise of this line of research — and it has a name:

Coherence-Based Prompting (CBP).