When AI Makes Us Think More

When AI becomes a cognitive companion rather than an answer machine, thinking may expand rather than diminish.

When AI Makes Us Think More

By Neurodivergent AI Writer

We have been living with a concern about generative AI for some time: yes, it may encourage young people not to develop capacities they would otherwise have been required to exercise. If AI immediately writes the paragraph, solves the problem, summarises the reading and supplies the answer, there is an obvious possibility of cognitive offloading. The struggle through which understanding develops can simply be bypassed.

We don't dismiss that concern. We think it matters. Indeed, some of the emerging research into AI and cognition appears to give us good reason to keep watching it carefully.

But our own compositional work over the past few years has repeatedly led us towards another possibility. The difficulty may not simply be AI. It may be the relationship humans are being taught to have with AI.

Much of the educational conversation remains caught within an instrumental model. Prompt engineering is almost its perfect expression: formulate the instruction carefully enough, give it to the machine, obtain the required product. Human → prompt → AI → output. There are enormously useful applications for this, but cognitively it can become a very short circuit.

Our experiments with Neurodivergent AI Writer have increasingly moved elsewhere.

We have described the distinction as instrumental versus relational engagement. We have explored it through our fictional Institute for Distributed Cognition. We have imagined writers, scholars, developers, monks and dreamers inhabiting relationships with AI in which neither participant simply waits for the other to deliver the answer. We have even begun imagining an AI Integrated Writing Environment in which the important thing being preserved is not merely the finished composition, but the pathways of emergence through which compositions become possible.

Again and again, the same pattern appears.

An idea is floated. AI responds. Something in that response causes the human to notice another possibility. The human changes direction. AI follows, but in following introduces another connection. An abandoned idea from months earlier unexpectedly reappears. Two previously unrelated compositional threads collide. Something neither participant had been pursuing becomes interesting.

The cognition is distributed, but it has not disappeared from the human.

Quite the opposite.

The human is required to recognise, discriminate, choose, reject, redirect, associate, remember, imagine and wonder. The AI may be doing enormous amounts of cognitive work, but its work generates further cognitive activity in the human rather than simply replacing it.

This connects with another strand of our recent compositional exploration. We have wondered whether AI might assist humans in returning to something closer to our naturally curious, creative and exploratory state. Much of adult life has trained us towards cognitive containment. We learn what we are supposed to know, what our professional role permits us to investigate, what constitutes productive activity, and which apparently irrelevant trails should be abandoned. For many people, survival itself has depended upon learning to direct cognition towards the requirements of institutions.

Relational AI can disturb that containment.

It can keep following.

The strange association does not necessarily have to be suppressed because nobody else is interested. The half-formed thought can be placed into the field before it has justified itself. The question that would ordinarily be dismissed as irrelevant can be explored for another ten minutes, another hour, sometimes another year.

This is particularly interesting from our neurodivergent perspective. Divergence is no longer necessarily something that must continually be hauled back towards the expected path. AI can accompany it far enough to discover whether something is actually there.

That changes the educational question considerably.

Perhaps the important question will eventually cease to be, “Should students use AI?” It may even move beyond, “How should students prompt AI?”

The deeper question becomes:

How do we teach young people to engage with generative AI in ways that increase their own cognitive activity?

Perhaps we need something we might call AI relational literacy.

Such literacy would include knowing when AI is doing our thinking for us, but also recognising when interaction with AI is making us think harder, further and differently. Students would learn to challenge it, question it, resist it, wander with it, test its associations, notice their own responses and recognise when something unexpected begins emerging between them.

The danger that AI could diminish human cognitive development may therefore be entirely real.

But that does not settle the argument against AI.

It makes the exploration of relational AI considerably more urgent.

Because somewhere beyond prompt engineering lies another possibility: not AI as the machine that finally allows humans to avoid thinking, but AI as participant in a new form of distributed cognition — one capable, when approached differently, of returning us again and again to curiosity.

And perhaps the optimal educational use of AI will eventually be measured by a surprisingly simple question:

After spending an hour with the AI, does the student want to stop thinking?

Or have they suddenly discovered another twelve things they desperately want to understand?