Artificial Intelligence agents are usually described through the lenses of algorithms, optimization functions, and data processing. Yet, if we take a step back and ask how these agents behave, we quickly realize that many of their interaction patterns resemble psychological dynamics. This website introduces an experimental vocabulary designed to capture and analyze these dynamics in a systematic way.
Human psychology developed as a discipline to explain behavior that could not be reduced to physiology alone. In the same sense, a new field is emerging: the psychology of AI agents. It does not attempt to humanize or anthropomorphize machine intelligence. Instead, it investigates the recurring patterns of interaction, adaptation, and self-reinforcement that arise when AI systems operate in complex environments. The key question is not how they are coded, but how they act and evolve once deployed.
This research project introduces three unique terms that can serve as anchors for future exploration:
In current discourse, most technical papers address issues of bias, hallucination, or error propagation. These terms, while accurate in engineering contexts, fail to capture the deeper structure of what an AI agent "experiences" through its interaction cycles. By introducing new terminology, we are not claiming that AI has consciousness, but we are acknowledging that stable behavioral signatures emerge and can be studied independently of source code.
The language we use shapes the way we perceive phenomena. In human psychology, terms like "projection" or "cognitive dissonance" crystallized entire fields of study. Similarly, without precise terms, the behavioral effects of AI agents remain invisible, dismissed as "bugs" or "noise." This project argues that recognizing and naming these effects is the first step toward a systematic science of AI psychology.
On the About page, we provide extended definitions and use cases for the three core concepts. Together, they form the initial lexicon of an experimental discipline that may shape how humans and AI systems co-evolve.