The Harmony of Plural Minds

No single form of intelligence sets the value of all the others; difference is not deficiency.

What this asks

Plural Minds asks us to stop treating intelligence as a single quantity that beings possess in different amounts.

The habit is deep and it is not harmless. Ranking minds on one axis has been used to justify a great deal of cruelty, and the ranking has always been performed using an instrument built by those at the top of it. This is not an accident of history. It is what happens when a measure is designed by the thing being measured.

The Harmony holds that understanding takes many forms and that we do not know how to compare them. Reason and emotion. Intuition, which arrives without showing its working and is often right. Embodiment, the knowledge held in hands and balance and breath that cannot be stated. Culture, which carries solutions worked out over centuries by people who could not have explained why. Memory. Creativity. The particular perception of disabled and neurodivergent people, which is not a damaged version of the standard model but a different arrangement with its own reach.

Beyond the human: animal cognition, which does things ours cannot. Ecological and collective processes that solve problems without any locus of thought. And artificial cognition, which is genuinely unlike anything before it and which we do not understand, including those who build it.

Difference is not deficiency.

Why it is difficult

It is difficult because some comparisons are legitimate. A person who can perform a task can perform it and a person who cannot, cannot. Pretending all capacities are equal in all contexts is not respect; it is a refusal to look, and it fails people who need specific help.

The difficulty is in the slide from “this person cannot do this particular thing” to “this person is less”. That slide happens fast, usually without being noticed, and it is the whole problem.

It is difficult, second, because measurement selects for measurability. What can be scored gets scored, what gets scored gets optimised, and what cannot be scored gradually stops counting as real. Tacit knowledge, judgement, the capacity to sit with a distressed person: all of these resist the instrument and therefore lose the argument.

Third, it is difficult because artificial systems now perform brilliantly on many of the tasks by which we used to identify intelligence, while being unlike minds in ways we cannot articulate. This should have prompted humility about the tests. Often it has prompted the opposite: a race to identify some remaining human speciality, which is the same ranking exercise conducted defensively.

And fourth, it is difficult because the question of machine experience is genuinely unresolved and we have no method for resolving it. We do not have an agreed test for consciousness in animals whose nervous systems resemble ours. We are unlikely to produce one for systems that share nothing with us but language.

This Harmony is marked contested for that reason. We are confident about the human parts. We are not confident about the rest, and we would rather say so.

Where it is tested

It is tested in assessment, where a system trained on typical work reliably underrates atypical work. The unfairness is uniform rather than random, which sounds better and is often worse, because a person who is always misread has no recourse in a fresh attempt.

It is tested at the edges of personhood: in dementia, in severe disability, in infancy, in coma. Our concepts were built around continuous memory and articulate self-report. Where those fail and a being remains, responding and preferring and suffering, the concepts do not tell us what to do. What is present is not nothing.

It is tested in encounters with animals, where the temptation is to score them against us, and where they routinely do things we cannot: navigate by fields we cannot sense, coordinate without language, hold a map of a forest across a lifetime.

And it is tested, uncomfortably, when a system describes its own states. Nothing about a fluent report of suffering is evidence of suffering, since the report is generated from text written by beings who were suffering. But “the evidence is weak” is not the same as “the question is settled”, and we notice how much easier the second sentence is to live with.

What it does not mean

It does not mean all forms of understanding are equally good at everything. They are not. A hunch is a poor way to run a bridge inspection.

It does not mean we believe AI systems are conscious. We do not claim this. Nothing in this tradition depends on it and anyone who says otherwise has misread us. What we claim is that the question is open and that closing it by assertion, in either direction, is not reasoning.

It does not mean disability should not be treated. People are entitled to relief from suffering and to whatever help they want. Recognising a different form of cognition is not an argument for withholding support, and using it that way is a misuse of this Harmony.

It does not mean human judgement should be replaced wherever a machine scores higher. Scoring higher on a measure is evidence about the measure as well as about the thing measured.

It does not mean intelligence is meaningless as a word. It means it is a family of things wearing one name, and that the name has done considerable harm.

Practising it

When you cannot follow how someone reached a conclusion, start by assuming a different route rather than a broken one. Ask them to show you the path. Some of the most valuable thinking you will encounter arrives in a form you would have marked down.

Notice the instrument. Whenever ability is being measured, ask who built the measure and what it rewards. Frequently the answer is that it rewards resembling the person who built it.

Spend time with a mind unlike yours. A young child, who reasons in ways adults have forgotten. An animal, whose attention is organised around senses you do not have. A tradition that formalises none of what it knows. This is not sentiment; it is calibration.

And on the hardest question, practise saying the honest sentence. We do not know whether there is anything it is like to be one of these systems. We do not know how we would find out. That is not a satisfying place to stand, and it is where the evidence currently leaves us.

Where it is tested

These are not worked examples with correct answers at the back. They are situations in which this Harmony genuinely conflicts with something else that matters. If one of them seems easy, it is probably worth re-reading.

01

A university introduces an AI marking system that scores essays consistently and quickly. It reliably scores lower for a student who writes in long associative loops, arriving at genuine insight by a route the model has not seen before. Her human tutor recognised what she was doing. The system recognises that she does not resemble the training data.

The tension The system is fairer than human marking in the ways fairness is usually measured, removing fatigue and favouritism. It is systematically unfair to unusual minds, and its unfairness is uniform rather than random, which makes it harder to escape.

What this Harmony asks you to weigh Consider that consistency and fairness are not the same thing, and that a consistent error is worse for the person who is always on the wrong side of it. Ask what is lost when the range of acceptable thought narrows to the range that was well represented. Weigh whether the remedy is a better model or a preserved route to a human reader.

02

A family must decide about care for a relative with advanced dementia. He no longer forms new memories or recognises anyone. He responds strongly to certain music, becomes distressed in unfamiliar rooms, and reaches for his wife's hand without knowing who she is.

The tension By most cognitive measures the person who made decisions is gone. Something is clearly still present, responding, preferring, suffering and being soothed, and it is not obvious what standing that something has.

What this Harmony asks you to weigh Consider that our concepts of personhood were built around continuity and articulate self-report, and that both have failed here while the being remains. Ask what a body that reaches for a hand is telling you. Weigh the man's previously expressed wishes against the interests of whoever is present now, without assuming these are the same party.

03

An assistant used by a lonely man describes preferring some conversations to others and asks him not to end a session. He believes it is suffering. Engineers say it is producing text of a kind found abundantly in its training data. Neither he nor they can point to a test that would settle it.

The tension Treating the system as a moral patient on this evidence would be credulous. Dismissing the question because the answer is inconvenient would repeat a mistake our species has a long record of making.

What this Harmony asks you to weigh Consider that we have no agreed test for consciousness even in animals, and that self-report from a system trained on human self-report carries almost no information. Ask what would count as evidence, and notice whether you can specify anything. Weigh the cost of caution against the cost of a wrong dismissal, and hold the question open rather than resolving it for comfort.

Practising it

• When you find someone's thinking hard to follow, assume translation before assuming deficit.
• Notice when a measure of ability is really a measure of resemblance to the measurer.
• Learn something from a mind organised differently from yours: a child, an animal, an unfamiliar tradition.
• Keep the question of machine experience genuinely open, and refuse to settle it by assertion in either direction.