Fluid reasoning is equivalent to relation processing

[Grey Bach]

Hello brain trainers and intelligence enthusiasts. What do you guys think of this study? It demonstrates that people's relation processing (as measured by three relational validation tasks) is statistically equivalent to Gf (as measured by some reasoning tests including RAPM). Given the simplicity and explanatory power of the RP tasks, it seems to me like practicing relational validation could be a very promising way to train Gf. I'm not a programmer, but perhaps Fredo or someone here (I would be willing to offer compensation) would be willing to implement the simple relational validation tasks used in the study in a gamified format.

https://www.sciencedirect.com/science/article/pii/S0160289620300672

[Fredo Corleone]

You can catch me on Discord or Telegram or LinkedIn. Let me know which one you prefer :)

[Fredo Corleone]

Monitoring, comparing and matching.

I believe Syllogimous v3 have all three.

I can replicate the same exercises administered in the study, I don't think it's hard to code them.

For my Syllogimous v4 I wanted to make something more ambitious: an engine to procedurally generate Advanced Raven Progressive Matrices.

[성문규]

That will be so sick omg

--
You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brain-trainin...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brain-training/56290e60-0081-4231-9567-2db8381fabcfn%40googlegroups.com.

[Leonardo]

If you read the study, you will find out that the task whose performance most closely relates to the one on progressive matrices is analogy.

Previously, on the same paper, you can also discover that said matrices are said to be ruled by the "application of logical operators such as AND, OR and XOR".

That intelligence is relations plus analogies plus logic operations has been clear for a while now. That is why I suggested the following improvement over the current version of syllogimous:

You can read more about this in the thread dedicated to this software. 

[Leonardo]

Search my comments here: https://groups.google.com/g/brain-training/c/0W8PEx2uIIM

[성문규]

I personally think it would be a really cool project regardless of how effective of a brain training method it would be. It’s like propagating logical rules through language but the language would be images, and theoretically an infinitely diverse one at that.

To view this discussion on the web visit https://groups.google.com/d/msgid/brain-training/0bf341fa-7e0b-4ace-b220-122ffcc48d51n%40googlegroups.com.

[Rotem Segev]

it looks super hard, to remember all these relations and to do a logic gate at the end. 

[Leonardo]

Given that intelligence is scarce, it is probably not a low-hanging fruit. 
If intelligence is the ability to do "super hard" things easily with your mind, maybe you need to train how to perform "super hard" things. 

[Grey Bach]

What I find promising about the study is the fact that something as complicated as Gf can be compressed into simple relation identification. It strikes me that if all the thought to be disparate elements of reasoning can be explained by performance in such simple tasks, then there is something about these tasks that reflects the core of how reasoning operates. As with all brain training protocol, the chief concern is near versus far transfer. We know that practicing even highly correlated subtests on an IQ test doesn't necessarily beget enhanced performance on anything but the tasks trained. So it seems to me that if we want to create a training regimen with the highest probability of success, it needs to target the fundamental skill on which general performance is predicated. The skill studied here seems like a good candidate.

[Leonardo]

We know that practicing even highly correlated subtests on an IQ test doesn't necessarily beget enhanced performance on anything but the tasks trained.

I don't know that. Proof? 

The same authors of this study talked about how ceiling effects limited their results. That is, the task was too simple for higher performing individuals. Which is the issue with training for people with higher than average IQ. People like many of us. 

[Grey Bach]

I’m just saying that, so often, abilities trained on a single demanding task generalize far weaker to a wide array of tasks including things that share a lot in common those trained. For example, n-back meta-analyses discuss how training improves performance in some closely related n-back tasks, but the effect significantly weakens on other wm tasks, cognitive control and Gf. However, yes, it is conceivable to me that practicing a bunch of IQ questions could potentially yield broader skill acquisition in some circumstances. My point is that I think when tasks as simple as what was studied here correlate so highly with fluid reasoning, it says something about the broad applicability of the underlying skill in human cognition. Fundamental skills are what I think we ought to be aiming to train.

Regarding the ceiling effects and high degree of accuracy, I understand your concern. However, what stops us from just coming up with some way to make the task trickier and more confusing than the one in the study? Perhaps we can use larger configurations of symbols with more interference and certain logical connectives rather than just greater/less signs.

[Leonardo]

 However, what stops us from just coming up with some way to make the task trickier and more confusing than the one in the study? Perhaps we can use larger configurations of symbols with more interference and certain logical connectives rather than just greater/less signs.

This is very similar to what I was suggesting, given that the study points out that analogy is the ability most related to IQ testing and that progressive matrices are transformed by logical operators. 

Anyways, being that this study is not about training and increasing IQ, I wouldn't interpret it as relevant for our purposes beyond some indications to deepen our theoretical understanding. It merely finds an association. High IQ people live longer, but that doesn't mean that by living longer you would have a higher IQ. I have always thought that people with high IQs, and I work on a foundation for gifted children and I was one myself, so I know what I'm talking about, do better almost everything. I'm sure that you could find who is gifted by the way they lace their shoes or something of the like; that doesn't mean that you could increase you cognition by lacing your shoes more efficiently. 

[Grey Bach]

I agree that changing one variable in a correlation won't necessarily drive change in the other variable. But if reaction time was measured as being statistically equivalent to Gf, wouldn't it at least be worth considering that a simple and reliable way to train reasoning could be to train yourself to respond quicker both internally and externally to your surroundings, given that speed is a well-defined construct, known to be an important part of efficient thinking? By the same token, I think it is at least worth investigating whether training relational processing, a crucial skill in reasoning, is possible and whether doing so will drive an increase in fluid reasoning as measured by more complex tasks.

[Leonardo]

We already train relation processing, but with a methodology backed up by research as a way to increase IQ. 

[Grey Bach]

Yes, but the RAI only correlates moderately to strongly with FSIQ, PIQ, VIQ etc. whereas the specific relational processing task I'm talking about is essentially isomorphic to Gf (r=0.95). I understand that the training regimen may not transfer, but it still strikes me that it's worth trying. What do the rest of you think? This is just my proposal and I certainly wouldn't want anyone to waste time on it if you guys don't think it's worth it.

[Fredo Corleone]

I think it's worth trying, do you guys decided what set of exercises to train?

[Grey Bach]

I would like to see all of them, but if you only want to do one or two that is fine as well and by no means do you have to implement all my ideas. I think that there should be harder options since so many of us here are far on the right of the bell. It would be good if the player could manually increase the problem size and speed up the time. For monitoring, I was thinking that there could symbols on both sides of the columns and trickier patterns for more interference. For comparing, you could add some logical connectives, rather than just having greater than or equal to. The player would have to check if the below statements are logically equivalent to any of the above statements. For matching, maybe have some of the edges could cross or something like that.

[Grey Bach]

Anyone else's ideas would be appreciated also!

[성문규]

Idk if this is relevant since you guys were super focused on relational logic training but I had a brain training game idea that trains pattern recognition and visuospatial working/short term memory:

Get a grid of a certain dimension(ex: 2x2, 5x5, 7x7 10x10, etc). The dimensions of the grid determines one aspect of the difficulty.

Provide N number of randomly/near-randomly generated "images"(some combination filled in squares on the grid), each in succession so that the player has to keep in mind all of the N "images".

The player then has to fill in all squares of the grid that were occupied by all N number of previous "images". This can either be done N images presented -> player fill in, and repeat, or it can be done N back style,

so the player fills in commonly occupied squares after every presentation of an image.

There are many ways to modify the difficulty here. It would for sure train visuospatial memory and even complex working memory. The difficulty can get ridiculous I think, so maybe it's best to move up like 2x2 -> 2x3 -> 3x3 -> 3x4 -> 4x4 etc. 

Anyhow I think it would be pretty cool, and definitely not too hard to code.

To view this discussion on the web visit https://groups.google.com/d/msgid/brain-training/62c0ae9d-9734-4170-a2da-f29aac9cf96bn%40googlegroups.com.

[성문규]

Or I guess the prompt could even be something like "fill in all squares which were only occupied twice in the past N images", etc. Lots of different types of prompts could be used, so there is no one strategy the player can use while committing to memory the N images.

[Leonardo]

Greysonb: don't get me wrong, I'm totally in for the ride. 

It think it may work if it really makes use of logic:

¬A ⇐ B; B ≡ C; ¬C → D; ¬C ≡ ⊤; E → (B > A); .

What of the following are true? C ≡ A; D → ¬C; (A < B) ← E; ¬C ≡ A. 

Of course, letters can be substituted by geometrical forms, colors or whatever. 

Training like this would make you a genius, but I don't think this is exactly what you want. Although it would be my humid dream. 

Fredo: Is there a release date for this v4? I'm really excited for trying it. 

itrn: this already exists, I know I have played it. I have little to no faith on memory training lately.

[Fredo Corleone]

No, there's no release date. I'm pretty busy working xD

[Leonardo]

At least integrate Syllogisms inside binary before you completely give up on v3. I know that you are busy. But please, Fredo; for the love of God, please. 

[Grey Bach]

Leonardo: I was thinking pretty much what you were thinking except asking the player to identify which statements below are logically equivalent to one of the above statements to keep the task similar in nature to comparing. Would that still keep the humidity level high?

[Karla]

You finally found the only interesting study that is worth reading about intelligence. I saw people here that get excited about bullshits, or discovers the hot water and makes stupid assumptions like cognitive flexibility is intelligence (holy crap). I've mapped all those cognitive functions and studies you are posting these days several years ago, also relating them to their neural substrate, and found or developed the necessary softwares for training them (for my personal use obviously). The software for relational training developed on this forum could also be useful, but as pointed out by some members some time ago, it has a lot of unnecessary redundancy and it does not tackle the real underlying factor everyone refer as general intelligence, that Jastrzębski pointed out. I was sincerely waiting for someone to investigate this hypothesis, that I was thought years ago, when I was interested in the field. I just want to give you an advice: enthusiasm is good, but IQ is by no means everything, other cognitive functions are just as important if not more, but above all a superior IQ would have not changed your life, and will not change your life if you will manage to get it.

Ok now that I have humiliated you enough by showing your ignorance, I give you a present; Jastrzębski published another paper that will suit you. In that paper he describes the only task you need to tackle the "general intelligence factor" and it is called graph mapping. You can also download the software that he developed as an IQ test alternative to classic IQ tests https://link.springer.com/article/10.3758/s13428-022-01846-z

I made a script in python to change the code and make it an adaptive training, but honestly I'm not using it because I have a life and a lot of things more interesting to do and enjoy than wasting my precious time playing games to raise my IQ. 

Il giorno venerdì 11 novembre 2022 alle 04:31:35 UTC+1 greysonb...@gmail.com ha scritto:

[Leonardo]

I didn't make that assumption. That is the title of the first study I linked. You made the assumption that I made that assumption, showing your profound stupidity, and expressing it in the most ill-mannered, rude, uneducated, yet highly amusing, way; stupid woman with name of a Czech whore. 
If you read all the research I shared in the thread in question, instead of stopping at the title, you will see how important it is to our purposes.

[Bo T]

Karla, There are so many stupid points in your post that I honestly literally don't know where to start. 

--

You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brain-trainin...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/brain-training/4b59d8d1-fc8a-45e5-a87e-dfab21c53ba5n%40googlegroups.com.

[Leonardo]

Where do these trolls come from? Are Karla and zoseph the same person with multiple account? Lately we have had a batch of extremely poor, inarticulate, insulting users. Are we being attacked by someone? Are, if considering Hanlon's razor, people just this stupid? 

It is difficult to explain posts like Karla's, zoseph's and this one:

thisemailb...@gmail.com

5:13 (hace 3 horas) 

do you have more receptive glowly eyes because of n-back?

a Dual N-Back, Brain Training & Intelligence

have you noticed your pupils being more dilated? 
do you look more intelligent?

[Karla]

Dear Leo,

my comment was harsh, but I didn't use inappropriate language or insults like you, with your explosion of nonsense... that just raises some other doubts about your mental stability, which is already in discussion judging by your messages on this groups.

Your reply is meaningless and useless, and you only replied on the less important part of my message (because you don't know how to reply, given the lack of arguments). By the way, I've read all the messages in that thread and is full of nonsense, and you were strongly supporting that claim initially because you change your mind every other seconds, just reading something and getting excited about nothing.

All those things you are posting these days are really old, you really don't know those cognitive functions and the relation between them after all your bla bla on this forum, after all your readings and you flawed reasonings? I just see that you spend your time boasting on this forum and wasting people time writing about nothing, spraying some copy-past around and then adding flawed assumptions already confuted by literature, which you clearly don't know  (and I'm not talking about that post, but about everything you write).

If you are just 1/100 intelligent as you are trying to unsuccessfully imply in your every message, you would see that what I wrote is really useful, more than all the useless nonsense messages you spread on this forum. But I think that your false confidence, fragility and instability is too great to make your eyes open. 

[Leonardo]

The exotic name, the deprecation... I cannot answer, I'm too turned on to think with clarity. All my blood flow is on my penis. 

[Bo T]

Leo may have a short temper sometimes, but you're a downright cunt karla. Plus, most of us have verified IQs of 135+, so stop condescending and contribute properly, or stfu. 

To view this discussion on the web visit https://groups.google.com/d/msgid/brain-training/9c9d30f5-382d-4b58-b0d7-d3b21329def6n%40googlegroups.com.

[Fredo Corleone]

Guys, calm down xD
I don't see the point in insulting each other like this.

Anyway, I would like to discuss what Karla, in particular: "The software for relational training developed on this forum could also be useful, but as pointed out by some members some time ago, it has a lot of unnecessary redundancy and it does not tackle the real underlying factor [...] the only task you need to tackle the "general intelligence factor" and it is called graph mapping".

I think Karla is referring to Syllogimous v3, I have to say that the app is redundant in many ways (infact I've pointed it out many times yet implemented feature requests), but it's actually very similar to graph mapping.

One example (the first opening the app):

GUEST is at East of HAMBURGER
FREEZER is at East of MOON
HAMBURGER is at North of FREEZER
GOVERNOR is at East of GUEST

MOON is at South-West of GOVERNOR

The above question is a graph (to know more https://en.wikipedia.org/wiki/Adjacency_matrix), everything is in some cardinal relation with everything else.

I would say that syllogisms are also graph based with a plus, the thing with them is that they can either contain or not, be contained (with various degrees of nesting) and be constrained to not contain other objects.

That said I read the paper about isomorphic graphs mapping/matching and I don't think it's much different from what Syllogimous can currently offer, I would rather develop rule based progressive matrices for the next Syllogimous than implement isomorphic graphs.

[Leonardo]

Certain parts are redundant logically, but not semantically. The famous RAI+ also had redundant parts. You already know my point of view. At the very least, the syllogisms should also be available inside of binary, and having some degree of negations, as in the RAI+ study, "CAJ is not different to BEJ. BEJ is the same as TUP. Is TUP the same as CAJ?" Would be interesting. Although from a programmer point of view it seems absurd, from a grammatical, semantic and neurological point of view, it makes sense. 

Sentential negation is a universal syntactic feature of human languages that reverses the truth value expressed by a sentence. An intriguing question concerns what brain mechanisms underlie our ability to represent and understand the meaning of negative sentences. We approach this issue by investigating action-related language processing and the associated neural representations. Using functional magnetic resonance imaging we measured brain activity in 18 healthy subjects during passive listening of sentences characterized by a factorial combination of polarity (affirmative vs. negative) and concreteness (action-related vs. abstract). Negation deactivated cortical areas and the left pallidum. Compared to abstract sentences, action-related sentences activated the left-hemispheric action-representation system. Crucially, the polarity by concreteness interactions showed that the activity within the action-representation system was specifically reduced for negative action-related vs. affirmative action-related sentences (compared to abstract sentences). Accordingly, functional integration within this system as measured by Dynamic Causal Modeling was specifically weaker for negative action-related than for affirmative action-related sentences. This modulation of action representations indicates that sentential negation transiently reduces the access to mental representations of the negated information.

https://www.researchgate.net/publication/23237861_Negation_in_the_brain_Modulating_action_representations

Negation is one of the most taxing actions in language processing and, most likely, one of the most limiting factors when it comes to understand relations. 

[Leonardo]

In the current EEG study, we ponder whether the reusing of response inhibition processes by negation is constrained to action-related contents or consists of a more general-purpose mechanism. To this end, we employed the same dual-task paradigm as in our prior study—a Go/NoGo task embedded into a sentence comprehension task—but this time including both action and non-action sentences. The results confirmed that the increase of theta power elicited by NoGo trials was modulated by negative sentences, compared to their affirmative counterparts, and this polarity effect was statistically similar for both action- and non-action-related sentences. Thus, a general-purpose inhibitory control mechanism, rather than one specific for action language, is likely operating in the comprehension of sentential negation to produce the transition between alternative representations.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6694754/

[Leonardo]

Inhibitory Control training has more transference to reasoning than Working Memory training:

Regarding transfer effects, results showed specific patterns of near transfer effects depending on the type of training. Interestingly, it was only the IC training group that showed far transfer to reasoning. 

https://www.frontiersin.org/articles/10.3389/fnhum.2016.00588/full

[Leonardo]

Believe me, what in code is undeniably redundant, it is not so for the brain. Semantic modifiers change the way the brain activates:

Brain connections of words, perceptions and actions: A neurobiological model of spatio-temporal semantic activation in the human cortex

We show that learning of semantic relationships between words and the objects and actions these symbols are used to speak about, leads to the formation of distributed circuits, which all include neuronal material in connector hub areas bridging between sensory and motor cortical systems. Therefore, these connector hub areas acquire a role as semantic hubs. By differentially reaching into motor or visual areas, the cortical distributions of the emergent ‘semantic circuits’ reflect aspects of the represented symbols’ meaning, thus explaining category-specificity. The improved connectivity structure of our model entails a degree of category-specificity even in the ‘semantic hubs’ of the model. The relative time-course of activation of these areas is typically fast and near-simultaneous, with semantic hubs central to the network structure activating before modality-preferential areas carrying semantic information.
https://www.sciencedirect.com/science/article/pii/S0028393216302421

[Leonardo]

Also, in the study from the OP analogy is the stronger prediction of progressive matrices performance (0.68). 

Thus why I insisted so much on this:

 A Boolean logic view of analogical proportions (a basic form of analogical statements of the form "a is to b as c is to d") has been recently proposed and extended to another logical proportion, namely paralogical proportion (stating that "what a and b have in common, c and d have it also"). When used in combination, these 2 proportions provide an enhanced power to complete IQ tests.

https://www.researchgate.net/publication/221203575_Analogy-Making_for_Solving_IQ_Tests_A_Logical_View

[Leonardo]

I have a degree in linguistics and I have spent my fair share of time programming, although almost exclusively low-level: logic gates, assembly, modular synthesis (because I'm also a musician); what may seem absurd from a pure logical and mathematical point of view, it is not from a neurological point of view. Trust me. I can produce as much evidence as you need to make my case. 

[Leonardo]

Sorry, I forgot to highlight the two reasoning matrices. Here is right:

Analogy as the Core of Cognition: https://www.youtube.com/watch?v=n8m7lFQ3njk&ab_channel=Stanford

I know that you believe that it is an absurd mess; but our brains are an absurd mess! There is no way around it. Relations + Syllogisms → Analogies → Boolean Gates = IQ, to the best of our knowledge. Again, the main study on the OP says so:  

matrices are said to be ruled by the "application of logical operators such as AND, OR and XOR"

By adding negation we are just training a fundamental feature of language and training Inhibitory Control. By using Carousel mode we are training Working Memory, by switching the TRUE and FALSE and having multiple semantic connectors we are training Cognitive Flexibility. 

Read all the research on cognitive flexibility and its importance for transfer here: https://groups.google.com/g/brain-training/c/bn3S4CqUQH4

With those few touches, we will be training all the fundamental aspects of cognition with just one program. I know that you don't particularly believe in furthering implementing these features, and that you are more motivated with your idea of matrices; but I think that we are close to perfection here and that we are very close to a breakthrough. 

[Leonardo]

More on the importance of negation:

The present study separated the aspects of logic from language and numerosity—mental faculties whose distinctness has been debated for centuries—and identified a new cytoarchitectonic area as correlate for an operation involving logical negation. A novel experimental paradigm that was implemented here in an RT/fMRI study showed a single cluster of activity that pertains to logical negation. It was distinct from clusters that were activated by numerical comparison and from the traditional language regions. The localization of this cluster was described by a newly identified cytoarchitectonic area in the left anterior insula, ventro-medial to Broca’s region. We provide evidence for the congruence between the histologically and functionally defined regions on multiple measures. Its position in the left anterior insula suggests that it functions as a mediator between language and reasoning areas.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6957563/

Negation is present in all natural and artificial languages, and children from an early age use and understand it. Yet, studies show that negative sentences are more difficult to process than affirmative sentences (Carpenter & Just, 1975; Clark & Chase, 1972; Farshchi et al., 2019; Greco, 2020; Kaup & Dudschig, 2020; Trabasso et al., 1971; Wason & Johnson-Laird, 1972; Wason & Jones, 1963). This asymmetry between both types of sentence is often known as the negation effect and has interested psychologists, linguists and philosophers alike.

Probably, the most accepted proposal in literature to explain the negation effect is its double processing. This idea stems from the logician Bertrand Russell who wrote that 'when I say, "this is not blue" I somehow consider it to be blue first and then reject it, considering it a colour other than blue' (Russell, 1948). This hypothesis has been taken up by different theories. According to the embodied cognition account, negation cannot be explicitly represented given that it is a linguistic operator that has no direct experiential simulation. As a result, the processing of a negative sentence (e.g., “the door is not open”) involves two steps during comprehension (Kaup & Zwaan, 2003): first, the representation of the negated situation (e.g., an open door) and then the actual situation (e.g., a closed door). Negation meaning would be implicitly captured in the deviations between these two simulations. Thus, according to this view, negation is difficult because, unlike affirmative sentences (e.g., “the door is closed”), which directly and exclusively represent the actual situation (e.g., a closed door), it calls for two mental simulations (Dudschig & Kaup, 2020a; Kaup et al., 2006). Recently, this two-step theory has been supplemented by mechanistic proposals that suggest that negation operates through processes associated with cognitive control functions; in particular, with conflict monitoring and inhibition (Dudschig & Kaup, 2018, 2020b; see also Beltrán et al., 2019, 2021; Liu et al., 2020a, 2020b).

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8660733/

[Fredo Corleone]

Then I should pay researchers to substantiate our beliefs and sell the program to the public and get rich! xD

[Leonardo]

I have more than 100 students. I swear to God that, if that is what you want, if you implement the changes I ask for, I will myself conduct an experiment with them to measure its cognitive impact. 

[Bo T]

I am waaay out of my depth when it comes to this stuff, but I vote for implementing the features solely based on Leo's enthusiasm about them lmao. 

To view this discussion on the web visit https://groups.google.com/d/msgid/brain-training/4b12e6b7-f4cf-4e00-96fe-06bc9f0fa9abn%40googlegroups.com.

[Grey Bach]

Fredo: Fair enough. If you think Syllogimous and graphing are too similar then stick to what is novel. What do you think about the other two tasks? Do you think the comparing task implemented in a way Leo or I were discussing has merit?

[Fredo Corleone]

I believe everything has it's unique flavor and usefulness.

OT try the following exercises http://reactivex.io/learnrx/ I think they are good food for your brains