Sleepwalking has interesting epistemological implications.

People generally seem to overestimate what tasks require consciousness. Things like keeping one's balance, walking and navigating your environment without bumping into things, having basic conversations, identifying and moving objects (even delicate ones like china bowls, without breaking them), preparing foods (albeit poorly), pouring drinks, and pulling up blinds are all possible while sleepwalking.

The woman in this video does a number of these things.

Sleepwalkers have no recollection of what they did while asleep. They are not conscious while sleepwalking. Yet they can perform all of these tasks, so consciousness cannot be required for their execution.

Here's what's interesting about the video linked above. At 1:53, we see a table. Then we see a cut to what is presumably another night, and the table has been moved. It's not in that same place anymore. Still, as the video points out, the woman walks around where the table used to be – as she normally might to avoid bumping into it. She then tries to put a jar of mayo on the missing table, 'realizes' (though not consciously) that the table isn't there, and sets the mayo on the floor instead. Then, at 2:36, she does not go around the missing table anymore. She walks right where the table used to be. Assuming she was sleepwalking the whole time, that means she was able to recognize and respond to changes in the environment. Something changed in her – unconsciously – to update her model of that room, take into account the missing table, and change her behavior accordingly. Setting down the mayo is what tipped her off.

That is the kind of useful change in behavior most people will eagerly interpret as evidence of consciousness in animals. But, as we have just seen, unconscious people can change behavior in the same way, so such changes cannot be evidence of consciousness.

In this video – granted, it's from Dr. Oz, so take it with a grain of salt – we can see a girl playing the piano while asleep, a woman dancing and then (nonsensically) explaining her behavior in conversation, and a girl sleepwalking nine miles (!) to her uncle's house. That last girl says she has no recollection of sleepwalking. A nine-mile walk requires plenty of flexible and fairly sophisticated behavior because you need to take turns, cross the street, avoid cars, and so on. How did the girl do that? A pathfinding algorithm, as philosopher Elliot Temple, who has influenced my views on animal sentience, explained to me in the context of cat behavior. As I've written before, people are mistaken to think flexible behavior is evidence of consciousness.

What's striking in most of these examples is that, with some exceptions like not walking around the missing table anymore, many of these behaviors are nonsensical – by which I mean that the sleepwalkers make mistakes which, were they conscious, they generally wouldn't make in the first place or would immediately correct. For example, in the last video, a man is seen getting a bowl and then pouring milk not in the bowl but down the kitchen sink. The general impression scenes like that leave corroborates the Popperian notion that consciousness may have to do with error correction. But, as the example of the missing table shows, it can't just be any kind of error correction. Something more must be going on. If any kind of error correction required consciousness, the woman would not have been able to correct her error of avoiding collision with a non-existing table while asleep. I have conjectured that being conscious is the same as being critical.

Many animals routinely do things that are far more idiotic than what we see in these sleepwalking videos. For example, dogs 'swim' above water under certain conditions. Most people never give the idea that animals aren't conscious a chance, but if a person tried to swim above water they'd readily accept that he isn't conscious as an explanation. They'd recognize that something is seriously off about him. The difference is, of course, that animals are never conscious, while people sometimes are.

Even when people are conscious, they frequently do things that don't require consciousness. Walking, as we have seen, does not require conscious attention, even though we usually happen to be conscious while walking (unless you're a sleepwalker).

The last video features a sleep specialist who says that sleepwalkers can do anything they've done before – such as driving. I suspected the same, and it also has epistemological significance: it means people can do things they have learned previously (while awake). Then, while asleep, it's all mindless execution. In turn, I suspect people cannot learn new things while sleepwalking, because learning takes a critical attitude. They lack this attitude in their sleep. In other words, someone who does not yet know how to drive won't be able to learn to drive in their sleep. But if they've learned it before – and, presumably, if they have corrected a sufficient amount of errors – they can do it again while sleeping.

Physicist David Deutsch, who is also of the opinion that animals are not conscious, has said that "any useful change can be considered learning". Real learning is more than just a useful change in behavior. Whatever real learning is, it does seem to require – or perhaps bring about – consciousness.

These considerations are also important for the field of artificial (general) intelligence, where useful changes and sophisticated behavior are mistaken for evidence of consciousness all the time.