4 ms·

The top answer is a worthwhile read if learning is of even remote interest. I'm forwarding it on to a teacher I know, who is always looking for ways to improve

The top answer is a worthwhile read if learning is of even remote interest. I'm forwarding it on to a teacher I know, who is always looking for ways to improve children's learning.

tgv 2y ago
But it's not exactly accurate. E.g., > You can actually model this precisely, mathematically, using a forgetting curve. I'm not exaggerating when I refer to these things as laws of physics That's utter bullshit. There is no precise model for retention time. It isn't mathematics. It isn't even physics. Retention time isn't even the only factor involved. So what remains of the "mathematically precise forgetting curve"? People forget more and more over time. And even that isn't 100% true.
- rmbyrro 2y ago
  Multiple studies have found statistical patterns that can be modeled through mathematical functions. It's not mathematical in a precision and mechanical sense, but the standard deviation of practical observations from the modeled curve is low enough to make it a good predictor and guide for spaced repetition and to optimize time spent to retention ratio.
  - mjburgess 2y ago
    The laws of physics refer to real properties the world has (eg., mass, charge, etc.). Largely, when you use statistics, you're modelling epiphenomena: bundles of randomly arranged properties that aggregate to an apparent quantity that doesnt correspond to anything in the world (eg., compare "mass" with "salad"). The problem with much psychology (and anything ending in -metrics) is that mainstream research in it, for ~century, has been treating the former and latter as equivalent. There is no property of animals called, "forgetting", and anything we model with stats is, at best, an epiphenomenon. It's important not let the x-metrics disciplines get away with their pseudo-scientific propaganda on this. Populations of individuals have, on average, rates of memory (etc.) performance across time which can be approximately quantities. It is highly unlikely this corresponds to any physical property.
    - n4r9 2y ago
      By that argument, are the "Gas Laws" [0] not considered laws of physics? They are an aggregated by-product of more fundamental inter-molecule interactions. Perhaps you'd class them as Chemistry rather than Physics? [0] https://en.wikipedia.org/wiki/Gas_laws https://en.wikipedia.org/wiki/Gas_laws
      - mjburgess 2y ago
        Statistical mechanics follows more closely from real properties of individual particles. The ensemble effects, via enough randomization, can be quantified from indviidual to macro properties. Temperature isnt, in this sense, an epiphenomnon.
        n4r9 2y ago
        In what sense are you using the word "epiphenomena", then? Wikipedia provides several definitions across various domains [0] none of which appear to match the way you're using it. It even states that "charge and current are 'epiphenomena'" according to Maxwell. Your description > aggregate to an apparent quantity that doesnt correspond to anything in the world is unclear. What do you mean by "the world"? You go on to say > There is no property of animals called, "forgetting" It feels possible to me that a sufficiently complex model of the brain can observe an individual and determine that they are in a state of "forgetting". Not with current technology, of course, but I don't see what rules it out in principle. Likewise, I don't see why it's impossible for mental states to be reduced (with sufficient computational power) to the states and interactions of the brain's constituent atoms. [0] https://en.wikipedia.org/wiki/Epiphenomenon https://en.wikipedia.org/wiki/Epiphenomenon
        mjburgess 2y ago
        Say there are formula whose terms correspond to individual causal properties of a system, whereby action on one of these terms necessarily brings about a determinate response in another, eg., f=kQq/r^2... if I increase one of the charges, necessarily, the force increases. There is a sense in which a highly specific explanation of "electrostatic force" includes "charge", so that deabstracting this formula would yield a logically necessary structure. Then there are formula on the other end, "teen mazagine regressions", where the terms are superstitious aggregates (eg., star signs, number of times you've kissed, etc.). And formula/models/etc. inbetween, where terms like 'temperature' correspond to a property that no part has, but there is a "nearly determinsitic" relationship between them, such that necessarily for any given set of measures (x, y,t) you have a temperature.. but that you cannot, in practice, actually set {(x, y, t)} so you have a "stochastic-causal relationship". My claim is that almost all pscyho-social science is of the teen magazine variety: running linear regressions on surveys and then falsely attributing reality the apparent property given by the correlation coefficient. This is just outright pseudoscience and why much of this stuff is unreporducible nonesense. There is nothing even like "temperature" to our mental faculties, which if explained with a scientific theory, would comprise a vast number of specific causal properties. To say we "forget something" is to describe a combination of 100s of processes each of which have different ecological expressions. Likewise, "IQ" is exactly the same sort of psuedoscience: a linear regression through survey responses of an entire population and then a fraudulent pseudoscientific project of taking this correlation coefficient and turning it into a singular mental faculty. 110% BS
  - tgv 2y ago
    > Multiple studies have found statistical patterns that can be modeled through mathematical functions. For conditions that don't generalize. If it would be a "mathematical law", spaced repetition would not work, because the retention time would be that of the last presentation. So, we've got some exception. Interference, reinforcement and generalization are another known influences. Psychology doesn't have a good model of how memory works, whereas there are fairly detailed approximations to model the behavior of e.g. gases. So it isn't mathematics, and it doesn't approach the standards of physics. And that's also part of what's wrong with cogsci. It has held physics too long as the standard to emulate.
    - rmbyrro 2y ago
      If you study modern physics, you'll be surprised by how much uncertainty there is. Thankfully, uncertainty is not an obstacle for acquiring useful knowledge. Not in physics, not in cognitive science. I understand your skepticism, but you're missing out on a LOT of very useful advancements in our understanding of the brain and mind, despite all uncertainty that still remains.
Balgair 2y ago
Oh man, AI is going to frickin rule for education now. Based on what that answer is saying, an AI something along the lines of GPT4o - with the ability to lockout YouTube and TikTok and the like - is just going to sky rocket kids learning. Things like Duolingo already are doing the spaced repetition and all that jazz. That stuff is easy enough to handle with software already. But the accountability portion is key here. We have the hyper addictive apps already now. The kids are successfully hooked. The sugar water is ready for the thirsty rats. Now we just need the supervision, the accountability, the grad student looming above the maze, the gate keeper. Combining the AIs we already have with the software of spaced repetition and the like should be straightforward. The judgement portion is then essentially solved. All you have to do then is link that output to the ability to access the sugar water, the addictive apps. The kids are going to do anything to get through these 'lame' educational barriers then. And all of it will be automated, personalized, and cheap. Sure, sure, gotta keep an eye on it, spot check things from time to time. Make sure the rascals aren't ginning up the systems and hacking something. But that is just orders of magnitude cheaper than the near one-on-one instruction that the comment is saying is needed.

The top answer is a worthwhile read if learning is of even remote interest. I'm forwarding it on to a teacher I know, who is always looking for ways to improve children's learning.

tgv 2y ago

But it's not exactly accurate. E.g., > You can actually model this precisely, mathematically, using a forgetting curve. I'm not exaggerating when I refer to these things as laws of physics That's utter bullshit. There is no precise model for retention time. It isn't mathematics. It isn't even physics. Retention time isn't even the only factor involved. So what remains of the "mathematically precise forgetting curve"? People forget more and more over time. And even that isn't 100% true.

rmbyrro 2y ago
Multiple studies have found statistical patterns that can be modeled through mathematical functions. It's not mathematical in a precision and mechanical sense, but the standard deviation of practical observations from the modeled curve is low enough to make it a good predictor and guide for spaced repetition and to optimize time spent to retention ratio.
- mjburgess 2y ago
  The laws of physics refer to real properties the world has (eg., mass, charge, etc.). Largely, when you use statistics, you're modelling epiphenomena: bundles of randomly arranged properties that aggregate to an apparent quantity that doesnt correspond to anything in the world (eg., compare "mass" with "salad"). The problem with much psychology (and anything ending in -metrics) is that mainstream research in it, for ~century, has been treating the former and latter as equivalent. There is no property of animals called, "forgetting", and anything we model with stats is, at best, an epiphenomenon. It's important not let the x-metrics disciplines get away with their pseudo-scientific propaganda on this. Populations of individuals have, on average, rates of memory (etc.) performance across time which can be approximately quantities. It is highly unlikely this corresponds to any physical property.
  - n4r9 2y ago
    By that argument, are the "Gas Laws" [0] not considered laws of physics? They are an aggregated by-product of more fundamental inter-molecule interactions. Perhaps you'd class them as Chemistry rather than Physics? [0] https://en.wikipedia.org/wiki/Gas_laws https://en.wikipedia.org/wiki/Gas_laws
    - mjburgess 2y ago
      Statistical mechanics follows more closely from real properties of individual particles. The ensemble effects, via enough randomization, can be quantified from indviidual to macro properties. Temperature isnt, in this sense, an epiphenomnon.
      - n4r9 2y ago
        In what sense are you using the word "epiphenomena", then? Wikipedia provides several definitions across various domains [0] none of which appear to match the way you're using it. It even states that "charge and current are 'epiphenomena'" according to Maxwell. Your description > aggregate to an apparent quantity that doesnt correspond to anything in the world is unclear. What do you mean by "the world"? You go on to say > There is no property of animals called, "forgetting" It feels possible to me that a sufficiently complex model of the brain can observe an individual and determine that they are in a state of "forgetting". Not with current technology, of course, but I don't see what rules it out in principle. Likewise, I don't see why it's impossible for mental states to be reduced (with sufficient computational power) to the states and interactions of the brain's constituent atoms. [0] https://en.wikipedia.org/wiki/Epiphenomenon https://en.wikipedia.org/wiki/Epiphenomenon
        mjburgess 2y ago
        Say there are formula whose terms correspond to individual causal properties of a system, whereby action on one of these terms necessarily brings about a determinate response in another, eg., f=kQq/r^2... if I increase one of the charges, necessarily, the force increases. There is a sense in which a highly specific explanation of "electrostatic force" includes "charge", so that deabstracting this formula would yield a logically necessary structure. Then there are formula on the other end, "teen mazagine regressions", where the terms are superstitious aggregates (eg., star signs, number of times you've kissed, etc.). And formula/models/etc. inbetween, where terms like 'temperature' correspond to a property that no part has, but there is a "nearly determinsitic" relationship between them, such that necessarily for any given set of measures (x, y,t) you have a temperature.. but that you cannot, in practice, actually set {(x, y, t)} so you have a "stochastic-causal relationship". My claim is that almost all pscyho-social science is of the teen magazine variety: running linear regressions on surveys and then falsely attributing reality the apparent property given by the correlation coefficient. This is just outright pseudoscience and why much of this stuff is unreporducible nonesense. There is nothing even like "temperature" to our mental faculties, which if explained with a scientific theory, would comprise a vast number of specific causal properties. To say we "forget something" is to describe a combination of 100s of processes each of which have different ecological expressions. Likewise, "IQ" is exactly the same sort of psuedoscience: a linear regression through survey responses of an entire population and then a fraudulent pseudoscientific project of taking this correlation coefficient and turning it into a singular mental faculty. 110% BS
- tgv 2y ago
  > Multiple studies have found statistical patterns that can be modeled through mathematical functions. For conditions that don't generalize. If it would be a "mathematical law", spaced repetition would not work, because the retention time would be that of the last presentation. So, we've got some exception. Interference, reinforcement and generalization are another known influences. Psychology doesn't have a good model of how memory works, whereas there are fairly detailed approximations to model the behavior of e.g. gases. So it isn't mathematics, and it doesn't approach the standards of physics. And that's also part of what's wrong with cogsci. It has held physics too long as the standard to emulate.
  - rmbyrro 2y ago
    If you study modern physics, you'll be surprised by how much uncertainty there is. Thankfully, uncertainty is not an obstacle for acquiring useful knowledge. Not in physics, not in cognitive science. I understand your skepticism, but you're missing out on a LOT of very useful advancements in our understanding of the brain and mind, despite all uncertainty that still remains.

Balgair 2y ago

Oh man, AI is going to frickin rule for education now. Based on what that answer is saying, an AI something along the lines of GPT4o - with the ability to lockout YouTube and TikTok and the like - is just going to sky rocket kids learning. Things like Duolingo already are doing the spaced repetition and all that jazz. That stuff is easy enough to handle with software already. But the accountability portion is key here. We have the hyper addictive apps already now. The kids are successfully hooked. The sugar water is ready for the thirsty rats. Now we just need the supervision, the accountability, the grad student looming above the maze, the gate keeper. Combining the AIs we already have with the software of spaced repetition and the like should be straightforward. The judgement portion is then essentially solved. All you have to do then is link that output to the ability to access the sugar water, the addictive apps. The kids are going to do anything to get through these 'lame' educational barriers then. And all of it will be automated, personalized, and cheap. Sure, sure, gotta keep an eye on it, spot check things from time to time. Make sure the rascals aren't ginning up the systems and hacking something. But that is just orders of magnitude cheaper than the near one-on-one instruction that the comment is saying is needed.