Each of us knows directly only one mind: our own. We infer that other people have minds because they are built like us and behave like us. This inference grows harder the further we move from the human case, toward animals and machines.
Animal minds
Descartes held that animals are machines without thought, complex automata whose cries are like the sounds of a clock. Many of his contemporaries found this hard to believe, and it has few defenders today. In 1789 Jeremy Bentham, arguing that animals deserve moral consideration, wrote: “The question is not, Can they reason? nor, Can they talk? but, Can they suffer?” For Bentham, the key question about animals is whether they can suffer.
Evidence for animal consciousness comes from several directions: pain behavior, such as guarding injuries and trading off food against avoiding harm; shared brain structures and chemistry; and evolutionary continuity. In 2012 a group of scientists issued the Cambridge Declaration on Consciousness, which stated that many nonhuman animals have the brain basis for consciousness, including all mammals and birds and some others, such as octopuses.
The mirror test, developed by Gordon Gallup in 1970, checks whether an animal uses a mirror to inspect a mark on its own body. Chimpanzees pass; many other animals do not. But failing the mirror test shows little on its own, since animals may rely on other senses, such as dogs on smell. As Nagel's bat reminds us, other creatures may have rich experiences that are very unlike ours.
The Turing test
In a 1950 paper, “Computing Machinery and Intelligence,” the mathematician Alan Turing proposed setting aside the vague question “Can machines think?” In its place he proposed asking whether a machine could pass as human in a text conversation. In his imitation game, a judge exchanges typed messages with a hidden human and a hidden machine. If the judge cannot reliably tell which is which, the machine has done well. Turing predicted that within about fifty years machines would play well enough that an average judge would have no better than a seventy percent chance of a correct identification after five minutes. He also answered objections, including the claim that a machine could never be conscious.
The Chinese room
In 1980 the philosopher John Searle offered the Chinese room argument. Imagine a person who knows no Chinese, locked in a room with a rulebook in English. Chinese characters are passed in; by following the rules, the person passes back other characters. To outsiders the answers are as good as a native speaker's, yet the person understands nothing. Searle concluded that running a program is not enough for understanding: programs manipulate symbols by their shapes (syntax), while understanding requires meaning (semantics).
The best-known response is the systems reply: the whole system, not the person alone, understands, just as no single neuron understands English but you do. Searle answered that the person could memorize the rulebook and still not understand. Another reply says a computer connected to cameras and limbs in the world, a robot, might ground its symbols in real experience.
Today's questions
Modern conversational AI systems make these old thought experiments practical. Their fluent language shows that behavior in conversation can be produced in ways quite unlike human thinking, which makes behavior alone a shaky guide to inner experience. Experts disagree about whether such systems understand anything or could ever be conscious. A functionalist may think the right organization could produce a mind in any material; a critic in Searle's line thinks biology matters. Because the stakes include how we should treat both animals and machines, many suggest caution in both directions: not dismissing possible minds too quickly, and not crediting them on the strength of fluent words alone.
Worked example: three candidates for a mind
Consider three very different beings and ask, for each, what evidence bears on whether it has experiences.
A crab. Crabs have far simpler nervous systems than mammals. Yet in experiments, crabs that received mild electric shocks in one shelter later tended to avoid it, trading off the shelter's value against avoiding harm. Their protective behavior suggests more than reflex. The evidence is uncertain, but it is the kind Bentham's question asks us to weigh: can they suffer?
A conversational AI. It writes fluent paragraphs, tells jokes and says it is happy to help. On a Turing-style test it may do very well. But its words were produced by a system trained to predict text, and it lacks a body, pain receptors or an evolutionary history. Searle would say it is the Chinese room at scale. A functionalist might reply that we need to know its internal organization before deciding. Most experts think fluent words alone are not enough evidence.
A newborn baby. Babies cannot talk, would fail the Turing test and do not pass the mirror test until they are toddlers. Yet almost no one doubts that babies feel pain and comfort. This shows that our tests track particular abilities, not experience as such.
Notice the pattern. For the crab, behavior and biology point toward some experience despite the lack of language. For the AI, language is abundant but other evidence is missing. For the baby, everything but the tests points to a rich inner life. Thinking clearly about minds beyond our own means weighing several kinds of evidence, and holding conclusions with appropriate humility.