5

Often, humans will claim to "understand" something. When pressed, they will define understanding as something like:

  • Knowledge
  • Conception within the mind
  • Comprehension
  • Awareness of meaning

However, these all seem specious to me. Knowledge isn't well-defined. Minds might not exist, and the best definitions imply that minds are not confined to human bodies, but sprawl out into notes; can a notepad understand something? Comprehension seems like a synonym for understanding, when it's not used as a descriptor for standardized testing. And finally, awareness of meaning seems to both presuppose a semantics, and also some deeper ill-defined entailment from semantics; it's easy to construct scenarios where a human claims to understand something, and then takes action which contradicts that claim.

So, what's going on here? Is there an operational definition of understanding? Can humans understand anything?

13
  • 1
    This can be moved to Skeptics SE if it is not a question of philosophy. However, given that this is parallel to one of Socrates' more famous claims (that humans do not know anything), it seems like a philosophical venue is more appropriate.
    – Corbin
    Commented Sep 2, 2023 at 15:33
  • sir, the nail's head is flush with the board, sir!!
    – Hudjefa
    Commented Sep 2, 2023 at 17:50
  • 1
    Indeed complete understanding is extremely hard, in some sense understand very often leads to not understand where one begins to truly understand, as the ancient Flower Adornment sutra warned long ago: Seeing what is true as true, Seeing what is not true as not true: Such ultimate understanding Is the reason for the name of Buddha. The Buddhadharma cannot be realized: Understanding this is called realizing the Dharma. All Buddhas cultivate in this way. Not a single dharma exists... Commented Sep 3, 2023 at 5:38
  • Yes obviously, but detailed answers need to elucidate the different concepts you are mixing. "Awareness of meaning": humans communicate, in many cases with problems but in most cases with success. Thus, whatever meaning is, we usually "exchange" it. "Conception within the mind": this expresses a solipsism-like position. If "minds might not exist", then we (we who?) have no issue. Commented Sep 5, 2023 at 13:00
  • 1
    Obligatory Sep post plato.stanford.edu/entries/understanding
    – emesupap
    Commented May 9 at 18:21

5 Answers 5

2

ChatGPT raises an interesting question. I would argue that it doesn't understand anything it says. It is just exploiting statistical regularities of text in the vast corpus on which it was trained. Normally we might use the fact an entity could explain how it knows something as evidence of understanding, but in this case it is just exploiting the statistical regularities of text in a huge training corpus that happens to contain examples of texts that explain how people know things.

There is still some scope by finding out if an entity can use it's knowledge to synthesize something new, or to solve a novel problem. ChatGPTs success in that direction is at best mixed. It has a tendency to generate things like mathematical proofs that look right (because they are based on proofs of similar concepts) but which contain obvious errors (because ChatGPT is just predicting symbols without understanding what they mean, so it can't check the proof step-by-step for errors).

I'd say that humans do understand things, but a practical definition of what that means isn't straightforward.

2
  • 1
    When I was a kid (so, before the widespread availability of computers) there was a popular demonstration that you could build something that 'learned' the winning strategy of a game using Dixie cups and paperclips. And, it did work. I wrote a short computer program that would always either win or at worst draw in Tic Tac Toe. Provably. So no, understanding, cognition, knowledge, consciousness... are not needed to succeed at things or there would be no life on earth.
    – Scott Rowe
    Commented May 9 at 22:53
  • 1
    @ScottRowe indeed, just because it doesn't understand what it says, that doesn't mean ChatGPT isn't useful or a very impressive achievement - it is, but it does show that "understanding" is not easily measured even if we intuitively know what it is!
    – user6527
    Commented May 10 at 8:16
2

"In this sort of predicament, always ask yourself: How did we learn the meaning of this word ("good", for instance)? From what sort of examples? In what language-games? Then it will be easier for you to see that the word must have a family of meanings." -Wittgenstein, in Philosophical Investigations

There is a tendency with these kind of complex words with many ways of using them, to jump in looking for some kind of noumena to it, some kind of essence. To think, can we do with the word in practice what we intuit or assume we are doing with it, and so as you do to hold up contradictions or inconsistencies, and declare it meaningless. But language in practice is different, we use the word to tell us useful things, it has functions in communication, and we readily understand what competant use of the word looks like. Another classic example:

"What then is time? If no one asks me, I know what it is. If I wish to explain it to him who asks, I do not know."

  • Saint Augustine

It is very common to identify a hierarchy, like say the DIKW pyramid

The DIKW pyramid, source https://en.wikipedia.org/wiki/DIKW_pyramid

I think of the military phrasing, that intelligence is information plus assessment. There is a process of sorting and organising, and then situating yourself towards information, in generating knowledge from it.

I like Vervake's terms, of salience landscape, and organising what we know so that we can get cognitive grip on what we know. Discussed here: Wisdom and John Vervaeke's awakening from the meaning crises?

Identifying the points where we can influence events, how they can have positive or negative impacts on us, is critical. We build models of the world and events, fitting together the informayiin or observations we have to improve them, whuch we can relate to induction and Bayesian reasoning. We look for errors and false predictions to distinguish between models, and we look for midels that could in principle be falsified, but we also have to acknowledge in this demain the models do not come from induction, they come from insights, they are limited to what we can imagine being the case.

It's widely acknowledged cosmological models have been dhaped by our highest technology, like looms and quern stones for Ancient Greeks and Vikings, aquaducts shaping the medical theory of humours, a clockwork or steam angine universe, or now the flow of information in the computer age. Our knowledge is limited by what models we can imagine. Newton's theory gave true knowledge, except in the presence of strong gravitational fields and high speeds, so later we had a special case, or were using limited or unjustified assumptions, that obscured a more general and usually more complex picture. Cartwright makes the case all models are wrong, are only valid in so as the premises, the guiding assumptionsare correct, and we always seek simplifying ones that give models that are more tractable or have other benefits, over direct unmediated experiences. Discussed here Thought experiments and empiricism

I think understanding what meaning is, is crucial to identifying and organising what knowledge is, and for that we have to look at the social generation of language and the Private Language Argument: According to the major theories of concepts, where do meanings come from?

Otherwise we slip into substance dualism, the idea words have meaning seperately to our use of them. I argue even mathenatics relies on intersubjectivity, here: The Unreasonable Ineffectiveness of Mathematics in most sciences

10
  • 1
    "Social generation of language" is a bankrupt concept. Meaning derives from use, yes, but the objective world is the ultimate authority on whether a claim is right or wrong, not what people say to each other about it. It is possible for people as a group to be wrong. Even a unanimous group of people can all be wrong. if people were the ultimate authority, that could not happen.
    – causative
    Commented Sep 2, 2023 at 17:43
  • Furthermore, the mind of the individual is the only place that judgments as to truth, falsehood, or meaning take place. A social group decides nothing on its own, except as the union of individual judgments made by individual minds.
    – causative
    Commented Sep 2, 2023 at 17:59
  • "Even a unanimous group of people can all be wrong" But can there be meaning, seperate to people? "the mind of the individual" How do you reconcile that with the Private Language Argument? Intersubjectivity is not the same as group-decided reality, eg. the existence of money is decided by the group but then also has stubborn & awkward realities that go beyond the group's decisions.
    – CriglCragl
    Commented Sep 3, 2023 at 8:38
  • The private language argument is complete nonsense. An individual may investigate or think about whatever he wants, with or without sharing it with others. If he makes up a word to use by himself, its meaning derives from whatever his intention was. The truth or falsehood of claims involving that word derive directly from the object of his investigations, as judged by himself, not from some other person telling him he's right or wrong.
    – causative
    Commented Sep 3, 2023 at 9:04
  • 1
    One is not limited to using language to conjure pictures into others' heads; one can also use language to conjure pictures into one's own head. Both uses of language are meaningful. And the second use need not involve any second person.
    – causative
    Commented Sep 3, 2023 at 21:58
1

To understand a domain means to be able to quickly and correctly solve a variety of cognitive problems that relate to that domain. For example, to understand quantum physics means to be able to quickly solve a lot of problems in quantum physics. To understand politics means to be able to quickly (and correctly) answer a broad variety of questions about politics, such as being able to say how well a candidate would do if they campaigned on X policy instead of Y policy.

Solving the cognitive problems may involve deductive, heuristic, analogical, or any other kind of reasoning. It may also involve consulting outside sources, but you could be penalized for this.

Understanding is a somewhat fuzzy concept because:

  • How quickly are you able to solve the problems? If you spend a lot of time puzzling out the answer with multiple dead ends, you don't understand it as well as someone who immediately knows what to do.
  • Which inferences count as "relating" to that thing? Technically, biology "relates" to quantum physics, in that biological systems are built out of quantum systems. But for someone to understand quantum physics, we don't also require that they understand biology. There is a notion of how closely related a problem is to a domain.

We could try to write a formula to quantify how well you understand a domain. Let:

  • D be the domain
  • p be a cognitive problem, from the set of all possible problems P
  • R(D, p) be the relevance of the cognitive problem to the domain, between 0% and 100%
  • q(p) be the quality of the answer you give to the problem p, on a scale from 0% to 100%.
  • t(p) be a score based on how long it takes you to give the answer to p. t(p) is 100% for a very rapid answer, 50% for an answer that takes a fairly long amount of time, 0% if you never answer.
  • s(p) be a score based on how much you had to consult outside sources. s(p) is 100% if you could answer right off the top of your head, 0% if you just asked someone else who gave the answer, and some score in between based on how much you personally did the cognitive work to produce the answer, versus using reference sources, calculators, or other people.

Then your understanding of the domain could be given as the average of q(p)t(p)s(p), weighted by R(D, p):

(∑_{p∈P} R(D, p) q(p) t(p) s(p)) / ∑_{p∈P} R(D, p)

Note, that this average is taken over all possible p, whether you actually solved that particular problem or not. It is based on what you would be able to do if you were hypothetically given p to solve.

Also note that all the quantities involved in it are unspecified and open to interpretation. The formula is more of a guide to thinking about the factors that go into understanding, than something you would actually calculate a number for. (Although, you might use something like it to grade students in a course, or score how well a neural network solves problems.)

1
  • Gosh, I guess I understand a lot about computer programming after 45 years. Very satisfying.
    – Scott Rowe
    Commented May 9 at 22:56
0

Understanding is illusory and ill-defined. Humans merely feel like they understand. Humans may also feel as if others understand, as an instance of the mind-projection fallacy.

14
  • 1
    @MichaelHall which is why it is correct. (what else could define 'understanding'?)
    – Scott Rowe
    Commented May 9 at 22:42
  • 1
    @ScottRowe, if anyone expects to find happiness here they are in the wrong place! Philosophy seems like a bottomless pit of despair that creates mostly baseless confusion over the mundane. At least this site does... And this question and downvoted self-answer months later is the perfect example. Commented May 9 at 23:05
  • 1
    @MichaelHall "It is always darkest just before the dawn." I have a barrel of laughs here. Remember Rule Number 6!
    – Scott Rowe
    Commented May 9 at 23:10
  • 2
    I think that you feel that I understand. As you may know, I could well be a dog or a robot, and I think I could be a p-zombie. Interestingly, this actually sharpens my answer; I'll add a link to the mind-projection fallacy.
    – Corbin
    Commented May 10 at 15:40
  • 2
    @MichaelHall I was thinking about "The Fish Slapping Dance". Iyanla vanZant said, "Some people need to be slapped."
    – Scott Rowe
    Commented May 11 at 1:15
-1

"Is there an operational definition of understanding?"

Yes.

Operational, pragmatic, common definitions are available in most dictionaries, and for most people, this is sufficient. (the question fails to point out how they may be deficient in any meaningful way)

"Can humans understand anything?"

Yes.

"Anything", is an absolute statement. If the answer were no, then it would mean that humans are incapable of understanding anything at all. Yet all who read and post here demonstrate understanding in the form of reading comprehension, as well as an understanding of grammar by writing in the English language.

"Can a notepad understand something?"

No.

Notepads have never demonstrated any capacity for understanding.

3
  • 1
    This answer is good enough. Too bad it doesn't understand itself, but conscious written words are maybe taking things a bit too far.
    – Scott Rowe
    Commented May 9 at 22:41
  • 1
    This is a sophist's answer. You can't even demonstrate that we aren't p-zombies; I have yet to see you substantiate your claim, and your approach seems to be one of pearl-clutching disbelief that anybody could observe the universe differently from you.
    – Corbin
    Commented May 10 at 14:13
  • 1
    There's literally a Wikipedia page explaining the fallacy you've committed; it's the argument from ridicule. It's fine to think that philosophy is silly, but you ought to try showing why.
    – Corbin
    Commented May 10 at 14:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .