When many people first encounter generative AI, the natural focus is the technology itself.
How intelligent is it?
How accurate is it?
How useful is it?
I started from the same place.
Like many researchers, I was initially interested in understanding large language models and how they might support learning, project management, and decision-making. Yet something unexpected happened during a long series of Human–AI interactions.
The most important observations did not concern AI.
They concerned the interaction.
Over time, I noticed that ideas rarely emerged from either side alone.
Practical experience suggested a problem.
AI proposed interpretations.
Operational reality challenged those interpretations.
New questions emerged.
Further interactions followed.
Gradually, a pattern appeared.
Knowledge was not residing exclusively in the human participant.
Nor was it residing exclusively in the AI system.
Instead, knowledge seemed to emerge through the interaction itself.
This realization became increasingly difficult to ignore.
A new idea began to take shape:
What if Human–AI interaction should be treated as a unit of analysis rather than merely a tool-mediated activity?
At first, this was simply an observation recorded in a research diary.
Later, it became a conceptual question.
Eventually, it developed into a theoretical proposition.
The shift was subtle but important.
Instead of asking:
What can AI do?
the question became:
What emerges when humans and AI interact recursively over time?
This change in perspective eventually influenced an entire research program.
Discussions about AI capabilities gradually evolved into discussions about:
- knowledge development,
- validation,
- governance,
- actionable knowledge,
- and responsible management.
The focus moved away from technology itself and toward the processes through which knowledge is generated, interpreted, validated, and used.
Looking back, one of the most valuable lessons from the journey was that the most significant research questions were often hidden inside ordinary interactions.
What initially appeared to be conversations with a tool gradually revealed themselves as opportunities to observe knowledge development in real time.
Perhaps the most interesting subject was never AI alone.
Perhaps it was always the interaction.