Mastodawn

tl;dr : Just because we can form an argument for something at the drop of a hat doesn't mean we used that argument to arrive at our conclusion.

We have reason to believe that brains are not using deduction when forming conclusions.

And we have no reason to believe brains do use deduction when forming conclusions.

Brains do not contain logic gates.
Brains do not contain anything that promotes binary division.
Brains have no problem whatsoever with contradictory information or beliefs.

It just doesn't seem to be the case, that logic plays a role in human thinking.

But it does play a role in human communication.

@androcat Note that you can train a simple neural net to do simple logic very reliably. Training a net to do an XOR operation was the test-case I used when writing NN software in the 1990s, because it wouldn't work if the propagation curve wasn't right.

We probably use way too many neurons to do logic, but we certainly can do it... with the right training.

@woozle For sure.

But looking at people actually doing that spontaneously, if they're not using pen and paper, they only reliably conclude what they want to conclude, and it's touch and go whether the premises they conjure actually match those outcomes.

If there is a "human logic", I believe it would contain a great many axioms that would distress a logician, and it probably wouldn't be free of circularity or reference to non-logical things like aesthetics.

@androcat It's true that we have great difficulty with anything complicated, and that's why we have computational tools.

I think it's important to note, though, that when it's said that "people aren't logical", that doesn't mean that we can't be logical when we recognize the importance of doing so, because we can use those tools where our internal logic is inadequate or undependable. The tools become an extension of our thinking.

@woozle I agree, like language also is, and many other technologies.

The human brain is great at picking up extensions... it's extremely general-purpose in that respect.

I suspect this adaptability is due in large part to not being dependent on stuff like language and logic - no limits to the interface except very general *predictability*

@androcat I know that I don't use words to think -- but it was my understanding that most people do, or at least believe that they do, hence phrases like "thinking out loud", "that voice in your head", etc.

I'm thinking that it seems entirely possible that many or most people might, in fact, use words for most of their reasoning processes. Consider how closely LLMs can now replicate human reactions (even on quite complex subjects) merely by analyzing word-sequences across volumes of data... and especially consider how many people there are who really seem to have trouble with basic logic, the same way that LLMs often do...

@woozle Well, humans routinely perform tasks that they would never, ever, be able to put into words.

It's quite hard to investigate how people think about things without triggering their Thinking For Speaking : If we ask them, we've already gone too far.
That's a "leading question" by definition.

I believe most people are simply mistaken about how they think about things, and I am not alone in that.

Also, it needs to be borne in mind that whenever people are talking about how they think, they're exactly in the frame of mind where they are Thinking For Speaking - so by their very nature, reports of how people think will be skewed towards Thinking For Speaking, even if in fact people never think in language outside that frame of mind.

@androcat So maybe "translating" from thinking into speech just comes more naturally to some people? I know I always found it difficult to put some thoughts into words, especially new thoughts, but most people seem to expect that it's easy to "tell me what you're thinking".

There's also the thing in teaching language -- or at least there was when I was in grade school, back in the Neolithic Era -- where the teachers don't want to give you a translation because they want you to learn to "think in $language" rather than thinking in English and just translating. Perhaps that was a fallacy?

Someone who's here now Jul 13, 2023

@woozle @androcat the idea is real when it comes to the language of the internal monologue. I'm saying this as someone for whom English is a second language and whose internal monologue is now around 80% in English and 20% in my mothertongue. If you don't have one (I'd like to know more about how you think), then the change wouldn't make sense to you, so I'm not sure how to explain this.

I know of a person who was deaf and whose ("main" layer of) thinking was in text and mental images of a SL

@Hrosts

For me, it depends what I'm thinking about.

If I'm doing carpentry -- designing something I want to build -- then I'm thinking visually, in 3d. If it gets complicated, I'll sketch it out so I don't lose track of details.

If I'm doing coding, I think maybe I'm thinking in abstract diagrams -- sets, hierarchies, flowcharts. I'll sometimes sketch those out too, when I'm trying to puzzle out something complicated, but the visual representation of the diagram is not what's in my head.

If I'm planning my day -- I think it's diagrams again. I'm not very good at planning, mind you; I often don't think of the consequences of a change in plans, which is why I don't like it when that happens.

If I'm writing music -- well, obviously I'm thinking in notes and chords and texture. I might also be thinking of sounds or effects I'd like to try, in which case I might be thinking of microphone placement, signal-paths, etc.

Any other areas of thought I should cover? Feel free to #AMA.

@androcat

Someone who's here now Jul 13, 2023

@woozle @androcat an obvious one would be - what about when thinking about a potential conversation? Like, if you were to decide to reply to me, and then went to kitchen to grab a glass of water while thinking through a potential reply. Also would it differ if the conversation was in voice?

@Hrosts @woozle That's when we use the "thinking for speech" mode.

But that is not how we think, as such. It's an additional layer of translation on top of the actual thinking.

@Hrosts @woozle As a personal example:
Independent of my (later) education in linguistics, I at some point decided to push a lot of my mental tasks out of conscious mind. Essentially, I chose to trust that my subconscious mind can do them better and faster than I can do consciously, and stopped keeping track of them on the conscious level.

Works fine (for me at least).
But I am often surprised to hear how I put things into words in discussions, because while I have been aware of what I felt to be right, I didn't have a verbal expression of it before I had to explain it to someone else.

@Hrosts @woozle According to (I believe) linguistic consensus "inner monologue" is just "thinking for speaking" rather than actually how we think.

Language is representative.
They are sounds or shapes or gestures that Represent meanings. And "meanings" are ... thought.

So it doesn't really make sense that we would need to think in representations of thought, that's a chicken-egg problem that is only resolved if we can think in thoughts.

Someone who's here now Jul 13, 2023

@androcat Yeah of course. I quite often think in images and "vibes" and affects(?), and ways I'm not sure I can properly remember or describe, but a huge chunk of my thought is taken by the monologue, so I'm talking about it

@androcat @Hrosts

Yeah, to me it definitely feels like conversation goes through a wordless internal-processing phase:

(1) receive sounds
(2) parse into words
(3) assemble words into meaningful phrases
(4) interpret phrases into internal understanding
(5) nonverbal: figure out how I want to respond
(6) translate response back into words (usually with a lot of editing)

This is also a large part of why I hate telephone calls: it takes me just a little bit too long to do all that to keep up.