Dale Joseph

· Dale Joseph

The Human Resource We Waste Most

Stranded cognition is capacity that exists but cannot effectively enter shared use — and we waste it at a scale we rarely measure. The question is whether AI recovers it, or quietly replaces the exercise that keeps it alive.

Most of what a mind can do never makes it into the world.

The judgment a machinist holds but has never written down. The language a nurse speaks fluently in a country that will only let her work as a technician. The sentence a non-speaking child can form but cannot say. The invention a bright student never makes, because she was never shown anyone like her making one.

Call it stranded cognition: capacity that exists but cannot effectively enter shared use.

Notice what the phrase does not mean. It is not a mind destroyed by disease — that is a different and harder thing. It is not unemployment, and not a productivity gap. It is something narrower and stranger: the cognition is there, intact, and the world cannot receive it.

This is not a new problem, and it does not wait for artificial intelligence to be real.

The Channel, the Institution, and the Tool

Three ways a mind can be stranded, and they are not the same problem.

The first is a blocked channel. Locked-in syndrome is the extreme case: cognition fully intact, every motor path out of it closed. For decades, assistive technology has been the recovery here — speech-generating devices and eye-tracking have turned latent speech into shared speech for people whose minds work while their voices do not. The principle was proven before AI existed. And the old fear, that giving someone a communication device would keep them from learning to speak, turned out to be false.

The second is a blocked institution. Here the channel is fine and the world simply declines to use it. Economists have a name for one version: brain waste — college-educated immigrants working far below their training. Another version is the estimated twenty-seven million people in the United States screened out of hiring by filters that reject qualified applicants for not matching a job description exactly. Another is the inventor who is never made, because invention runs in families and neighborhoods, and exposure — not ability — decides who becomes one. None of this is about what people can do. It is about what institutions recognize.

The third is different again, and it is the one worth watching. A tool can extend a mind, or it can quietly take the mind’s place.

The Recovery Question

The tempting move is to say AI solves the first two and to stop there. It is a real temptation, and there is real evidence behind it. A tutoring program in Nigeria raised measured learning by about a third of a standard deviation in six weeks. Assistants measurably lift the work of novices, sometimes to the level of people with months more experience. The gains are real, and in some important settings they are largest among less-experienced users.

But the pattern is not a law. The same evidence shows it reverses with the task and the person: sometimes AI lifts the weakest performer most; sometimes it amplifies the strongest. There is no rule that says AI helps novices most. There is only a question.

That question is the one worth keeping, because it is the difference between recovery and something that only looks like it. In one randomized study, students given an AI tutor that answered directly did better while using it, and worse afterward than students who had never used one — immediate performance up, durable capacity down. Students given the same tutor, instructed to offer hints instead of answers, kept the gains and the backslide did not appear.

That is the whole argument in miniature. Scaffolding versus substitution is not a property of AI itself. It is a property of the design, the task, and the learner.

And the two are rarely a clean either-or. A writing aid that lifts every author’s work can flatten the variety of what they all produce. Recovery and stranding usually arrive together. The useful question is which one is winning.

What to Ask of Any System

Four questions separate recovery from replacement.

Does the system actually work at this task, and how does its competence compare with the human’s? A tool whose limits are misunderstood can become a liability rather than a scaffold.

Does it force the human to think — to retrieve, reason, critique, compare, practice — or does it hand over the answer? The cognitive exercise is the point. Skill is built by the doing, not by the result.

Does the person become more capable after the tool is removed, merely perform better while it is present, or become less capable without it? This is the test that separates everything else.

And is the goal to raise human capability, or to raise throughput and cut headcount? The same technology serves both. Only one of them is recovery.

None of this is free. The systems that might recover cognition carry their own bill — the energy and water behind them, the concentration of power in a few hands, the surveillance that trains on the people using them, and access that is anything but equal. And the deeper cost is the one this essay is about: a tool that remembers for you, decides for you, and imagines for you can, over time, strand more cognition than it ever frees. Skill decays when it stops being used. A machine that removes the need to use it accelerates the decay.

The point is not that AI is good or bad. It is that the outcome is not decided by the technology. It is decided by the question we keep asking — or stop asking.

Stranded cognition is capacity that exists but cannot effectively enter shared use. It was being wasted long before AI arrived. The first time we have a tool that might recover it is also the first time we have a tool that might quietly replace the exercise that sustains it. The difference is not in the machine.

It is in the four questions. And in whether anyone still asks them.

Dale Joseph is the author of Thought Partners: Preserving Cognitive Sovereignty in the Age of AI and founder of the Emergence Institute. He worked for years as a consultant helping install hospital networks before turning to writing and systems thinking. He lives in Boynton Beach, Florida.