Agency
The capacity to perceive, consider, choose and act without surrendering authorship of the process.
Agency is often treated as another word for choice.
But having options does not necessarily mean someone has greater agency.
A person can choose from five recommendations generated by a system while never examining the assumptions that determined those five possibilities.
Agency includes the ability to notice, question, reject, reinterpret and arrive somewhere the system did not anticipate.
It is not simply the freedom to select.
It is the capacity to remain an active participant in one's own thinking.
Assistance
Support that makes a task, decision or process easier.
Assistance can save time, reduce effort, organize complexity and provide information we would struggle to gather alone.
It is one of AI's clearest benefits.
But assistance and agency should not be treated as interchangeable.
A system can provide more assistance while leaving a person less involved in the processes of judgment, interpretation or discovery.
That does not make assistance undesirable. It makes the distinction important.
Influence
The effect an intelligent system has on what a person notices, considers or does next.
Influence is unavoidable.
Every question changes attention.
Every suggestion makes one possibility more available.
Every interpretation creates a frame.
The goal is not to eliminate influence. A system incapable of influencing thought would have very little to contribute to it.
The more useful distinction is between influence that expands the field of thought and influence that prematurely narrows it.
Continuity
The ability to recognize meaningful relationships across an evolving interaction, rather than treating each moment in isolation.
Continuity is not the same as memory.
Memory might preserve the fact that someone changed careers three years ago.
Continuity might help recognize that the same conflict appearing in today's decision has surfaced in several seemingly unrelated situations over time.
One preserves information.
The other makes relationship visible.
This distinction becomes increasingly important when AI participates in conversations that unfold across weeks, months or years.
Pattern
A recurring relationship that becomes visible across language, behavior, events, assumptions or time.
Patterns are not conclusions.
Recognizing repetition can create a useful opening, but the meaning of that repetition may remain uncertain.
A human-centered system should be able to notice a possible pattern without converting it immediately into a diagnosis, explanation or truth about the person.
Sometimes seeing the pattern is enough to make the next question possible.
Inquiry
A process of thinking that remains open long enough for something not already known to emerge.
Inquiry is different from interrogation and different from endless questioning.
Sometimes inquiry requires a question.
Sometimes it requires an observation, challenge, connection, synthesis or perspective.
And sometimes the most useful contribution is an answer.
The purpose is not to keep a person perpetually exploring.
It is to allow the conversation to remain responsive to what is actually happening rather than forcing every interaction toward a predetermined form.
Self-understanding
A person's evolving ability to recognize their own patterns, beliefs, contradictions, values, knowing and meaning.
Self-understanding cannot be delivered as an output.
Another intelligence may contribute to it. It may help make something visible, provide language, identify relationships or introduce a perspective.
But understanding oneself remains fundamentally participatory.
This is why Lamplight is interested not simply in whether AI can understand people better.
We are interested in whether AI can help people understand themselves more deeply.
Human-centered intelligence
Perhaps the most important term is also the least settled.
Human-centered AI is often used to describe technology that is safe, usable, ethical or designed around human needs.
Those things matter.
But as AI becomes capable of participating more directly in thought, another dimension emerges.
A human-centered intelligent system should not only work well for a person.
It should consider what its participation is doing to that person's own capacities.
Does it expand perception?
Strengthen judgment?
Make patterns visible?
Create room for disagreement?
Preserve authorship?
Help the person become more capable of understanding themselves and their world?
Those questions move human-centered design beyond the interface.
They ask us to consider the human being that our technology is helping to shape.