lichess.org/@/ndpatzer/blog/science-of-chess-achtung-einstellung/CuxWuZYP
thoughts. when I feel like I might be taking too much space there. (fluctuating notion)
thoughts. when I feel like I might be taking too much space there. (fluctuating notion)
Ok. I do like the openned door to mutliple mechanisms interacting with each other. I was just worried about having close some door on what is pattern recognition proper (without having some chess understanding of what a pattern might be and if there were categories, that can be formalized from the board. .
> Chess players and their varying tendency to fall for Einstellung-like traps strike me as an interesting opportunity to see how a rapid visual recognition mechanism and a more deliberative process involving spatial transformation and/or other means of extrapolating from a specific position end up working in tandem (or not).
I think, I am not sure that generalization problem from machine learning might not be a better model, as it does include time evolution of learning (a part of cognitive science I would hope). While such E. effect might have experimental context value, it might not be an explanation in itself. What I am trying to point to, is not necessarily against what you might be pursuing in the remaining of the blog left for me to keep digesting (chunk by chunk), I hope it does not rest on a frozen concept of pattern recognjition and problem solving. As both seem not clear or dissected enough with respect to chess information of the board.
I am not aware of all the chess targetting cognitive science publications, of course, I do have my own bias, I would say from late exposure in my past formation, to machine learning ideas. But it seems to me that they had to make sure to be compatible with the board information, down to its core rules. And then, the focus of the method development woud be not just about how expert do their things, but how one evolves (or travels in some to-be-learned wilderness of the chess world) from experiencing many games over generations of RL batches of initially self-play, pitting previous learned clones, with only one allowed to learn, and the other "serving" games to its improving clone according to the same previous batche many games experience set).
That I find already closer to dissecting what the words pattern recognitoin might lump, and maybe even problem solving, for the exploration (part generative model of control by the policy NN).
Could E. effect be better framed as a learning from data problem. I am also suggesting, that the dominant and not very explicit theories of learning in chess, that I might have been witnessing or hearing about, are not very aware of the need to train such skill. As the emphais is on things like "woodpecker" and speed of execution in time pressured memory retrival and performance tasks.
not on how does one generalize a solution learning in one exemple, to "similar" enough, but then not to dissimilar enough..
if there is no sense of how to generalize, I would think, that ROT taught as survival crutches without their board level argumentation (pre-condition) would tend to train E effect into the learners. The more an advice is not questioned for its dependencies and critical thinking reception, I would think the more the Occam razor least imagination effort of 360 genrealization would prevail. And the possibility that rating improvement would still happen with a rigid application of ROTs through many games, over some stretch of time, would not let one suspect the overgeneralization of the pre-condition orphaned action rule. I wonder If I am paraphrasing you without having read you.. sorry. if that is so.. but hey. paraphrasing does not hurt.
by neglecting a study of existing learning theories, even if never really revised in a serene truth seeking collaboration, hype and all getting in the way, maybe, of the market wanting it that way, as rewarding enough in some short term.
The art of generalization, needs also a better first exposure learning theory, I would say.. or suggest from own subjective theory of self trying to learn some chess, my way. (not very rating efficient, well, not how I measure progress, I mean). I also have some other clues. But I think it is more honest to call this a hypothesis. We all have our own blind spots right next to our expertise, domains. Perhaps a missing concept is exactly that. Expertise over what exactly? does the experience set so far matter?
> Chess players and their varying tendency to fall for Einstellung-like traps strike me as an interesting opportunity to see how a rapid visual recognition mechanism and a more deliberative process involving spatial transformation and/or other means of extrapolating from a specific position end up working in tandem (or not).
I think, I am not sure that generalization problem from machine learning might not be a better model, as it does include time evolution of learning (a part of cognitive science I would hope). While such E. effect might have experimental context value, it might not be an explanation in itself. What I am trying to point to, is not necessarily against what you might be pursuing in the remaining of the blog left for me to keep digesting (chunk by chunk), I hope it does not rest on a frozen concept of pattern recognjition and problem solving. As both seem not clear or dissected enough with respect to chess information of the board.
I am not aware of all the chess targetting cognitive science publications, of course, I do have my own bias, I would say from late exposure in my past formation, to machine learning ideas. But it seems to me that they had to make sure to be compatible with the board information, down to its core rules. And then, the focus of the method development woud be not just about how expert do their things, but how one evolves (or travels in some to-be-learned wilderness of the chess world) from experiencing many games over generations of RL batches of initially self-play, pitting previous learned clones, with only one allowed to learn, and the other "serving" games to its improving clone according to the same previous batche many games experience set).
That I find already closer to dissecting what the words pattern recognitoin might lump, and maybe even problem solving, for the exploration (part generative model of control by the policy NN).
Could E. effect be better framed as a learning from data problem. I am also suggesting, that the dominant and not very explicit theories of learning in chess, that I might have been witnessing or hearing about, are not very aware of the need to train such skill. As the emphais is on things like "woodpecker" and speed of execution in time pressured memory retrival and performance tasks.
not on how does one generalize a solution learning in one exemple, to "similar" enough, but then not to dissimilar enough..
if there is no sense of how to generalize, I would think, that ROT taught as survival crutches without their board level argumentation (pre-condition) would tend to train E effect into the learners. The more an advice is not questioned for its dependencies and critical thinking reception, I would think the more the Occam razor least imagination effort of 360 genrealization would prevail. And the possibility that rating improvement would still happen with a rigid application of ROTs through many games, over some stretch of time, would not let one suspect the overgeneralization of the pre-condition orphaned action rule. I wonder If I am paraphrasing you without having read you.. sorry. if that is so.. but hey. paraphrasing does not hurt.
by neglecting a study of existing learning theories, even if never really revised in a serene truth seeking collaboration, hype and all getting in the way, maybe, of the market wanting it that way, as rewarding enough in some short term.
The art of generalization, needs also a better first exposure learning theory, I would say.. or suggest from own subjective theory of self trying to learn some chess, my way. (not very rating efficient, well, not how I measure progress, I mean). I also have some other clues. But I think it is more honest to call this a hypothesis. We all have our own blind spots right next to our expertise, domains. Perhaps a missing concept is exactly that. Expertise over what exactly? does the experience set so far matter?
Also I would not call a trap, necessarily, an exception surprise from an orphaned rule of action learned. (be it from woodpeckering or wrong attribution of eternal truth to shared teaching).
yet you do have a point. A trap must have some temptation. I wonder why there is not some characterization of such aspect.
rather, as in opening theory, we get catalogs of opening data, called trap. but this might actually be an old version of a transversal catalog. A common behavior or outcome quality shared across the tree of decisions known so far.
I think some of those might even be "transposition" traps.. But furthering charaterizations for trap catalogs per their temptation nature?
is it opening long line groove induced. or board information induced. perhaps a mixture. If one was trully trained from board features first, would they fall there too.
on the other hand.. similarity (undefined, left to intution tuning and life long exprience, that might have ended up traversing the hard way more than one or 2 "répertoires") of boar features..
well. again. this is about first exposure, or quality of expreince and contrast in outcomes from applying same action to different enough yet similar input.. soemthing like that..
i should keep reading now. another day though.
each paragraph triggers my thoughts. they were preexisting, I guess. so, the blog is acting like a structuring canvas, so I can be my natural debate high attention self (curse but also inspiring how I feel, like something is happening, not just time passing by, live not just survive). I don't care how I look, to those who actually come to read here. There must be something for them.
But someone else well structure shared thoughts is enabling me to pace my own things in echo to that. I just wished I could accumulate the reading and my reactions. into a well combed presentation in return.. but no.. good at storms and stuck in them.
yet you do have a point. A trap must have some temptation. I wonder why there is not some characterization of such aspect.
rather, as in opening theory, we get catalogs of opening data, called trap. but this might actually be an old version of a transversal catalog. A common behavior or outcome quality shared across the tree of decisions known so far.
I think some of those might even be "transposition" traps.. But furthering charaterizations for trap catalogs per their temptation nature?
is it opening long line groove induced. or board information induced. perhaps a mixture. If one was trully trained from board features first, would they fall there too.
on the other hand.. similarity (undefined, left to intution tuning and life long exprience, that might have ended up traversing the hard way more than one or 2 "répertoires") of boar features..
well. again. this is about first exposure, or quality of expreince and contrast in outcomes from applying same action to different enough yet similar input.. soemthing like that..
i should keep reading now. another day though.
each paragraph triggers my thoughts. they were preexisting, I guess. so, the blog is acting like a structuring canvas, so I can be my natural debate high attention self (curse but also inspiring how I feel, like something is happening, not just time passing by, live not just survive). I don't care how I look, to those who actually come to read here. There must be something for them.
But someone else well structure shared thoughts is enabling me to pace my own things in echo to that. I just wished I could accumulate the reading and my reactions. into a well combed presentation in return.. but no.. good at storms and stuck in them.
I did not think that the E. effect was about shorter or faster solution, but actually "easier" solution the way I read it first time like a few weeks ago.
Maybe the chess unwritten rule that shorter is better, another rule of thumbs, that I already discussed, is dominating as the notion of easier.
I thought it was more about using a known solution, to a new problem context. Notice here a possibly missing level of concept, how can it be a solution, are we not talking about a method of solution, being applied to find a solution. And then there might have been another method of finding another solution, and, that one is shorter, but it might have been harder on your cognitive load. So, while not giving the full problem constraints, how can one make sure which constraint not part of the problem, was actually being solved also?
Where does cognitive load come into the picture in this version of the E.effect.
I am not sure that if one is allowed to add constraints to the problem after proposing the task, that comparing 2 solutions with a new criterion, oh, and find better, the shortest one, no matter how hard, and no matter the cognitive load, to the point you might drop a kitten, and hence maybe provide for a wrong solution during the allotted time. I understand wanting to reduce the number of variables by saturating certain putative compartment (of high level model, same level as theory being tested), using time pressure. But I think care has to be taken about the baby in the bathwater. Could our assumptions be more than saturating a confounding factor, could it be restricting the scope of the experiment, or even bring own biases (worse than ignoring an explaining variable might be to make it impossible to detect even as noise first. wild guess here, salt please).
Did I interpret wrong. I thought it was about creativity, not hard cognitive load is best. I am twisting the point to show that there can be other constraints that could be value as better. As energy conservation. Or certainty, or self-aware error checking while solving, that might end up with a different optimality of task performance.
Hidden constraints are often how experiments change their role in future theories. It would be the same experiment and results, but after having refine the theory, it would be reinterpreted for its missing explicit context and the new experiment would shed light on alternate context. I am thinking of another blog about cognitive offloading. 2 papers not that far apart, showing exactly what I just said. well with my editorial version. They might not be presenting it that way, but my "analysis " of what their titles are saying (or their conclusion section, reverberated in their abstract), are doing.
So, there might be some slipping around the notion of better. I find the no solution versus a solution a more clear-cut notion of E effect. should I read all that you read? or can discussion go faster than that?
my partial take home from that recent fly by, was about notions of creativity, constraints, and expertise interacting in not always constructive ways, or being different independent variables to consider. I might have confused a bunch.
It is also possible that more than one thing is being lumped into the constellation so far having been swooped under the vocable, chess is not the only place where such in construction theory efforts might need further dissection. But contrary to psychology, chess has not the excuse of not being able to control the task. It could ask many questions, it does not, because it would not provide a performance boost within the remaining lifetime of the eager improver. who improves only its rating and own play, at any cost of understanding. or understanding later, if needed to get over some rating climb reduction.. must be tired.. that is old news from me.. also probably wrong, about the nature of plateaux. not about generalization not having been trained properly, or the E. effect if those are different things.. but hey, since nobody seems to have made that a known problem, i think asking is warranted.
Maybe the chess unwritten rule that shorter is better, another rule of thumbs, that I already discussed, is dominating as the notion of easier.
I thought it was more about using a known solution, to a new problem context. Notice here a possibly missing level of concept, how can it be a solution, are we not talking about a method of solution, being applied to find a solution. And then there might have been another method of finding another solution, and, that one is shorter, but it might have been harder on your cognitive load. So, while not giving the full problem constraints, how can one make sure which constraint not part of the problem, was actually being solved also?
Where does cognitive load come into the picture in this version of the E.effect.
I am not sure that if one is allowed to add constraints to the problem after proposing the task, that comparing 2 solutions with a new criterion, oh, and find better, the shortest one, no matter how hard, and no matter the cognitive load, to the point you might drop a kitten, and hence maybe provide for a wrong solution during the allotted time. I understand wanting to reduce the number of variables by saturating certain putative compartment (of high level model, same level as theory being tested), using time pressure. But I think care has to be taken about the baby in the bathwater. Could our assumptions be more than saturating a confounding factor, could it be restricting the scope of the experiment, or even bring own biases (worse than ignoring an explaining variable might be to make it impossible to detect even as noise first. wild guess here, salt please).
Did I interpret wrong. I thought it was about creativity, not hard cognitive load is best. I am twisting the point to show that there can be other constraints that could be value as better. As energy conservation. Or certainty, or self-aware error checking while solving, that might end up with a different optimality of task performance.
Hidden constraints are often how experiments change their role in future theories. It would be the same experiment and results, but after having refine the theory, it would be reinterpreted for its missing explicit context and the new experiment would shed light on alternate context. I am thinking of another blog about cognitive offloading. 2 papers not that far apart, showing exactly what I just said. well with my editorial version. They might not be presenting it that way, but my "analysis " of what their titles are saying (or their conclusion section, reverberated in their abstract), are doing.
So, there might be some slipping around the notion of better. I find the no solution versus a solution a more clear-cut notion of E effect. should I read all that you read? or can discussion go faster than that?
my partial take home from that recent fly by, was about notions of creativity, constraints, and expertise interacting in not always constructive ways, or being different independent variables to consider. I might have confused a bunch.
It is also possible that more than one thing is being lumped into the constellation so far having been swooped under the vocable, chess is not the only place where such in construction theory efforts might need further dissection. But contrary to psychology, chess has not the excuse of not being able to control the task. It could ask many questions, it does not, because it would not provide a performance boost within the remaining lifetime of the eager improver. who improves only its rating and own play, at any cost of understanding. or understanding later, if needed to get over some rating climb reduction.. must be tired.. that is old news from me.. also probably wrong, about the nature of plateaux. not about generalization not having been trained properly, or the E. effect if those are different things.. but hey, since nobody seems to have made that a known problem, i think asking is warranted.
"Simpler" I think. Not easier. While both need more specification to make that a satisfying explanation of the concept of E effect and the priors of the experiments. Or phenomenology.
I think one way is to think about this: expertise in hammering leading to everything looking like nails.
I think something else is floating, in my trying to understand my subjective dissonance. Population of solvers, and scope/specificity of tasks. Also, the definition of the problem, as being a single individual performance task. Something about small brains in a population of many brains. And population distributions of learning/cognitive styles, that one formulation with its common sense hidden assumptions might distort or transform, having different theory building consequences (or experimental conclusions/interpretation consistent with the working version) than another formulation with different set of unspecified might end up with. I do find a strong "similarity" at some level between the machine learning point of view or theory, and this different path of wording and associated theory from cognitive sciences.
It would not be the first time that there would be some detours, looking at the same thing. After all, theories are not recreating the complex reality we do not understand, but would like to.
They (theories in general) are things we can share that help us understand the target objectified reality not yet understood or technological controlled** within their scope of cerning (frenchism) and discerning toward explanation and prediction outside the experiments. This need for prediction is used to keep the wheel of theory and experiments going.
Interlude from some balcony:
What now, "generalization" again? I hear? What kind of Don Quichotte are you? Who's is talking right now, a staged intruder into this paragraph? a figment of my imagination, can I respond. but I was also the author.. well. it was fun to imagine and right without any warning. sorry. (now talking to the reader...).
concerns in no order (when do I ever do that.. rarely, after many iterations, not yet there, this is new. discovery of similar concept from somewhere I was not listening to for ages..
That the experimenters knows in advance that a task has a "simpler" solution and also another, that under some convention of criteria, could be called more complex. But that the effect being sought, or experimentally setup to observe or not, comes from the priors of the subjects. At first, that might also be unknown. And we find a diversity of behaviors given a well-formed task delivery.
Which is also a soft belly point, toward disambiguating conclusions and possible different notions of the E. effect (if I am subjectively allowed to have my current, possibly misconceived, working understanding).
ok i kept 2 deleted versions of this both starting with simpler rather than easier. (but in the end both needing specs, or the conclusion is not satisfying to me. What is that murky E. effect. is it only definable in each experiment, or it is something that can be shared. is it having a consensus meaning. (then what is the common sense context of the population for who it appears to be so).
The tyrany of one lifetime and many specialisation and their languages efficiency shortcuts. how to cut through that ?
third the charm. doubtful.. but better maybe.
** (that is for the impatient, like the improvers seeking only techonological advantage in some narrower pursuit, subset of that reality),
I think one way is to think about this: expertise in hammering leading to everything looking like nails.
I think something else is floating, in my trying to understand my subjective dissonance. Population of solvers, and scope/specificity of tasks. Also, the definition of the problem, as being a single individual performance task. Something about small brains in a population of many brains. And population distributions of learning/cognitive styles, that one formulation with its common sense hidden assumptions might distort or transform, having different theory building consequences (or experimental conclusions/interpretation consistent with the working version) than another formulation with different set of unspecified might end up with. I do find a strong "similarity" at some level between the machine learning point of view or theory, and this different path of wording and associated theory from cognitive sciences.
It would not be the first time that there would be some detours, looking at the same thing. After all, theories are not recreating the complex reality we do not understand, but would like to.
They (theories in general) are things we can share that help us understand the target objectified reality not yet understood or technological controlled** within their scope of cerning (frenchism) and discerning toward explanation and prediction outside the experiments. This need for prediction is used to keep the wheel of theory and experiments going.
Interlude from some balcony:
What now, "generalization" again? I hear? What kind of Don Quichotte are you? Who's is talking right now, a staged intruder into this paragraph? a figment of my imagination, can I respond. but I was also the author.. well. it was fun to imagine and right without any warning. sorry. (now talking to the reader...).
concerns in no order (when do I ever do that.. rarely, after many iterations, not yet there, this is new. discovery of similar concept from somewhere I was not listening to for ages..
That the experimenters knows in advance that a task has a "simpler" solution and also another, that under some convention of criteria, could be called more complex. But that the effect being sought, or experimentally setup to observe or not, comes from the priors of the subjects. At first, that might also be unknown. And we find a diversity of behaviors given a well-formed task delivery.
Which is also a soft belly point, toward disambiguating conclusions and possible different notions of the E. effect (if I am subjectively allowed to have my current, possibly misconceived, working understanding).
ok i kept 2 deleted versions of this both starting with simpler rather than easier. (but in the end both needing specs, or the conclusion is not satisfying to me. What is that murky E. effect. is it only definable in each experiment, or it is something that can be shared. is it having a consensus meaning. (then what is the common sense context of the population for who it appears to be so).
The tyrany of one lifetime and many specialisation and their languages efficiency shortcuts. how to cut through that ?
third the charm. doubtful.. but better maybe.
** (that is for the impatient, like the improvers seeking only techonological advantage in some narrower pursuit, subset of that reality),
ok. i think that the delete 2 versions of above, while overlapping, had their own slighlty different points I might want to share and keep at same level, in my committing stuff in here. bad or good. I might toy with language machine and throw such set of 3 "neighboring" blurb for it to chew on, and spit edulcorated versions of any of them. maybe with enough AND and ORs and iterations I could be helped into spewing both something I agree with, that has still some information sharing value, or some ideation proposal (my fun is to create things I find new hypothesis of association, and if some logic glue emotional seal (eureka!) can be found that is my soul comforting food). We all need to project some of ourselves in the future, and positively (survival bias) so. i.e. one can always dream...
If I recall right. what is missing from above. is where does reinforcement psychology comes from and whether it was about percpetion cognition or action cognition (somewhere between those poles of that dichotomy I am considering here, might be a bunch of other things, but if I try to have blurry vision, and most ignorant self at work here, I would find those interesting to use as explicit handles, just to see what it might give with the same data.. ).
Behaviorism, observable behaviors as only observer subjective objectivity (toying ... sorry). Its conceptual school of thought came from ethology (I gathered or conflated by my own imagination filling blanks as usual). So, they could not talk to their subject. So, for fear of anthropomorphizing too much, they are well full ignorant. in the theory with experiment wheel of science building. I would say the therapeutic art of those times might have needed a behaviorist splash in the pond, possibly as much as MFTL might have been needed to break some lack of awareness about the task's specifics about chess learning "theories" (yes, again, and that is not a catalog of data either). Luckily we can re-center after some dust settles, into keeping the parts that still are useful theory of something (learning, or chess board whereabouts when we hand move something on it).
the other one deleted. did it have another aspect I want to keep here. ? forgot. going to look.. but i did think it before rambling previous paragrpahs. this post was supposed to be readble. mind slapping myself. bad... bad.. brain.
If I recall right. what is missing from above. is where does reinforcement psychology comes from and whether it was about percpetion cognition or action cognition (somewhere between those poles of that dichotomy I am considering here, might be a bunch of other things, but if I try to have blurry vision, and most ignorant self at work here, I would find those interesting to use as explicit handles, just to see what it might give with the same data.. ).
Behaviorism, observable behaviors as only observer subjective objectivity (toying ... sorry). Its conceptual school of thought came from ethology (I gathered or conflated by my own imagination filling blanks as usual). So, they could not talk to their subject. So, for fear of anthropomorphizing too much, they are well full ignorant. in the theory with experiment wheel of science building. I would say the therapeutic art of those times might have needed a behaviorist splash in the pond, possibly as much as MFTL might have been needed to break some lack of awareness about the task's specifics about chess learning "theories" (yes, again, and that is not a catalog of data either). Luckily we can re-center after some dust settles, into keeping the parts that still are useful theory of something (learning, or chess board whereabouts when we hand move something on it).
the other one deleted. did it have another aspect I want to keep here. ? forgot. going to look.. but i did think it before rambling previous paragrpahs. this post was supposed to be readble. mind slapping myself. bad... bad.. brain.
deleted blooper 2 (the previous to last post being then the 3rd iteration possibly of same nature, until I consult some language machine, and shake well toward some voila).
Once upon a time....:
Maybe simpler solution not seen rather than "easier". As "better" "easier" is not really well defined.
But yes, any such word need some specifics. Need specifics to what end, might one ask, adoping the same approach I am suggesting, or hammering (c'est selon, haha, talking of subjectivity). Not to dismiss what has been done, to keep going from it, see a bit through it toward more refined conclusions, and one should guess by now, more questions.. It seems that answers are just stepping stones for the real stuff: better questions. (I am warped that way, maybe even under some E. effect, that I might have surfed for some while from questions to questions, and sometimes, enough answers I could share would fall off, and that got me hooked on the method, well, a long time scale it does feel like a mehod, but who knows, it might be hindsight ratoinalization, but I would not know, having only one lifetime (occams razor on that).
I guess I am swirling around that plateau of possibly circular concept set, if not being careful about the information flow. or basis.
question, task, subject, object. but also the gist of something that can be view both thorugh high level cognitive science model and experimental discovery, their words E. effect. and on another hand, the machine learning offshoot from a more ancestral psychology theory of learning (or was it memory, I think coming from behavioral school of thought, it was more about observable behavior, only the observers would handle words of perception, the subjects were are contributing as observable behavior, so it would be such measurable end-point that would make the task information flow back to the thoery from that kind of empirical source. I would say, more learning than memory as the outcome observable. (not that it does not "hide", perception, or decisoin, but the question that mattered was the dynamic between the behaving changes and the environment information feedback as a result of the behavior, under the convened and still accepted notoin of "affective/emotional/basic need satisfied/pain/negative known as such events, if not physical pain (observers, and intended scientific audience would likely convene on that, just being picky here, go read some texto, if that is not readable).
Once upon a time....:
Maybe simpler solution not seen rather than "easier". As "better" "easier" is not really well defined.
But yes, any such word need some specifics. Need specifics to what end, might one ask, adoping the same approach I am suggesting, or hammering (c'est selon, haha, talking of subjectivity). Not to dismiss what has been done, to keep going from it, see a bit through it toward more refined conclusions, and one should guess by now, more questions.. It seems that answers are just stepping stones for the real stuff: better questions. (I am warped that way, maybe even under some E. effect, that I might have surfed for some while from questions to questions, and sometimes, enough answers I could share would fall off, and that got me hooked on the method, well, a long time scale it does feel like a mehod, but who knows, it might be hindsight ratoinalization, but I would not know, having only one lifetime (occams razor on that).
I guess I am swirling around that plateau of possibly circular concept set, if not being careful about the information flow. or basis.
question, task, subject, object. but also the gist of something that can be view both thorugh high level cognitive science model and experimental discovery, their words E. effect. and on another hand, the machine learning offshoot from a more ancestral psychology theory of learning (or was it memory, I think coming from behavioral school of thought, it was more about observable behavior, only the observers would handle words of perception, the subjects were are contributing as observable behavior, so it would be such measurable end-point that would make the task information flow back to the thoery from that kind of empirical source. I would say, more learning than memory as the outcome observable. (not that it does not "hide", perception, or decisoin, but the question that mattered was the dynamic between the behaving changes and the environment information feedback as a result of the behavior, under the convened and still accepted notoin of "affective/emotional/basic need satisfied/pain/negative known as such events, if not physical pain (observers, and intended scientific audience would likely convene on that, just being picky here, go read some texto, if that is not readable).
I found the first, it might repeat some of above. when confronted to my own excessive vertical non-sparse filling with strings of characters I could then only see as a wall of shapeless text nightmare, for me to fix, or anyone to read, I thought trying again aiming at a concise version would be better. see previous post for that result. . and then back (back again, skip a post that is), for the 3rd.
Twice now, upon you courageous persevering reader (or language machine intermediate, maybe to the aided reader, consider reading both the averaged machine ironed version and the source, for differences, maybe first the machcine smooth third person smooth purring prose, and then with that in mind scan for contrast in my raw version, that would be my plan for my own future iteration attempts).
I was going to try some clever non-native English distorted formulation like: Twice upon some time.. but i give up. .
Here the zeroth iteration..
(I did say, some time ago, that I might be error prone with counting stuff, but I could explain algebra linguistics if enough symbols in them, a paradox).
Here, now, really.. wait for it.. and now you can look, just a line spacing below... The usual string flow, really, nothing new.
It might have been simpler being missed by expertise second nature "reflexes". I think the adage is about, when becoming a hammering expert, everything starts looking like a nail. Notice the emphasis on the action expertise.
Chess is certainly not built over ages and generations of human beings having a bite at it (lacking English, so that will do, could have been, having a "go" at it), toward that kind of expertise.
And yet, that is pretty much what woodpecker (or other intensive unspecified position world being covered repetitious training, meant to drive action patterns into "muscle" memory.
I find that the perception skills are forgotten in that exaggeration, a hypothesis of mine about what woodpecker is hiding as a theory of learning. All I know is that it has been a hypothesis of hindsight explanation of some titled expert attributing own expertise made official by tournament performance enough times (over some range of rating climb?, could it be of lifetime morsel, dedicated daily study, was it the only exposure to chess, and would it work for everyone, and last, does it work over many repertoires. How far the repertoires? Well, how could we know. How distant they are. Besides taking the whole population habits in sequential moves signature dataset clusters made graphs, assuming somehow that it would be representing some familiarity "logic", but that is a bit of an assumption, whether repertoire shifts are contiguous or not.
And here, I am doing nonsensical, because we don't have ways to say many of the things I mentioned, not in a common sensory way. In a hand wavy way, appealing to something we have been doing for ages, main-"line". (why line? is a sequene of SAN syllable representting a sequence of position, when their composition is actually instanciating the game from the initial position where it is applied (I am toying here, a bit).
Why does experts start changing "répertoire". or anyone for that matter. The swarm effect at high level for some under the top dreamer of high level glory.. or someone convince someone that this line, preferrably secret or not in vogue lately, would take enough people by surprise, that some rating mileage would be gotten until some population propagation of that deep line knolwedge. Or is it, maybe like the graph data analysis I recall (lichess blog some while ago, also a github repository for the work behind, using gephy as computing tool), an assumptoin of board based simlarity, disctance, or familiarity.
not constrast I would assume. would someone jump of the promise of a winning long and deep sequence just for the transient hike in novelty wins? I guess, that kind of work at the populaton level, being linked to the board information has been rare.
move seqeunce data being more dominant, it might have been that kind of information that was more salient in population data analysis (see another blog about tournament surprises, being about move sequence, invididual move surprise metric). Nothing wrong with that, like nothing wrong with a tree of decision projection of the world of chess positions and games. There are tools that have been developped.. Could a population get stuck in population tools experise in some sort of Effect.
I think the most abstract analogy from my corner of the mind woods, is doing gradient descent on an unknown complex surface (big space), using local curvature information only, and possible confusing some local optimum, as the global optimum.
Well, if not knowing the complexity (or needed degree of expressivity) that a target world to capture the mechanistic and probability law (all lumped per the formalism I am thinking of, a priori, a perhaps assuming I can separate some signal of mechanistic nature from noise of not having been taking all the factors of the ultimate truth into factor (chess not locally having such hidden factors, only about population of game outcomes (i.e. 2 player variable, and human calculation breadth limits, and expertise/knowledge/ignorance which I would call fog at range or whatever is beyond each person raw calculation breadth, if subtracting the knowledge/experience induce bias of pruning (it might be a champion bias too, the best so far, or engine familiy biases, they too have some fog and biases about it, I think DeepMind ensemble learning paper suggest that it might be a direction of modelling that shows promise (under my interpretation, but not only, I think I am not inventing anything, only I don't know much, so within my perception, I am making such hypothesis of understanding.
Twice now, upon you courageous persevering reader (or language machine intermediate, maybe to the aided reader, consider reading both the averaged machine ironed version and the source, for differences, maybe first the machcine smooth third person smooth purring prose, and then with that in mind scan for contrast in my raw version, that would be my plan for my own future iteration attempts).
I was going to try some clever non-native English distorted formulation like: Twice upon some time.. but i give up. .
Here the zeroth iteration..
(I did say, some time ago, that I might be error prone with counting stuff, but I could explain algebra linguistics if enough symbols in them, a paradox).
Here, now, really.. wait for it.. and now you can look, just a line spacing below... The usual string flow, really, nothing new.
It might have been simpler being missed by expertise second nature "reflexes". I think the adage is about, when becoming a hammering expert, everything starts looking like a nail. Notice the emphasis on the action expertise.
Chess is certainly not built over ages and generations of human beings having a bite at it (lacking English, so that will do, could have been, having a "go" at it), toward that kind of expertise.
And yet, that is pretty much what woodpecker (or other intensive unspecified position world being covered repetitious training, meant to drive action patterns into "muscle" memory.
I find that the perception skills are forgotten in that exaggeration, a hypothesis of mine about what woodpecker is hiding as a theory of learning. All I know is that it has been a hypothesis of hindsight explanation of some titled expert attributing own expertise made official by tournament performance enough times (over some range of rating climb?, could it be of lifetime morsel, dedicated daily study, was it the only exposure to chess, and would it work for everyone, and last, does it work over many repertoires. How far the repertoires? Well, how could we know. How distant they are. Besides taking the whole population habits in sequential moves signature dataset clusters made graphs, assuming somehow that it would be representing some familiarity "logic", but that is a bit of an assumption, whether repertoire shifts are contiguous or not.
And here, I am doing nonsensical, because we don't have ways to say many of the things I mentioned, not in a common sensory way. In a hand wavy way, appealing to something we have been doing for ages, main-"line". (why line? is a sequene of SAN syllable representting a sequence of position, when their composition is actually instanciating the game from the initial position where it is applied (I am toying here, a bit).
Why does experts start changing "répertoire". or anyone for that matter. The swarm effect at high level for some under the top dreamer of high level glory.. or someone convince someone that this line, preferrably secret or not in vogue lately, would take enough people by surprise, that some rating mileage would be gotten until some population propagation of that deep line knolwedge. Or is it, maybe like the graph data analysis I recall (lichess blog some while ago, also a github repository for the work behind, using gephy as computing tool), an assumptoin of board based simlarity, disctance, or familiarity.
not constrast I would assume. would someone jump of the promise of a winning long and deep sequence just for the transient hike in novelty wins? I guess, that kind of work at the populaton level, being linked to the board information has been rare.
move seqeunce data being more dominant, it might have been that kind of information that was more salient in population data analysis (see another blog about tournament surprises, being about move sequence, invididual move surprise metric). Nothing wrong with that, like nothing wrong with a tree of decision projection of the world of chess positions and games. There are tools that have been developped.. Could a population get stuck in population tools experise in some sort of Effect.
I think the most abstract analogy from my corner of the mind woods, is doing gradient descent on an unknown complex surface (big space), using local curvature information only, and possible confusing some local optimum, as the global optimum.
Well, if not knowing the complexity (or needed degree of expressivity) that a target world to capture the mechanistic and probability law (all lumped per the formalism I am thinking of, a priori, a perhaps assuming I can separate some signal of mechanistic nature from noise of not having been taking all the factors of the ultimate truth into factor (chess not locally having such hidden factors, only about population of game outcomes (i.e. 2 player variable, and human calculation breadth limits, and expertise/knowledge/ignorance which I would call fog at range or whatever is beyond each person raw calculation breadth, if subtracting the knowledge/experience induce bias of pruning (it might be a champion bias too, the best so far, or engine familiy biases, they too have some fog and biases about it, I think DeepMind ensemble learning paper suggest that it might be a direction of modelling that shows promise (under my interpretation, but not only, I think I am not inventing anything, only I don't know much, so within my perception, I am making such hypothesis of understanding.
yes.. I associated de novo (i.e. as I wrote) lichess blogs in that one.. more chess theory of learning or behaving. some population data reviewing.
I should honor the blogs with links.. for they were (and might still be) stimulating. . food for that mind.. (third person, danger danger).
I should honor the blogs with links.. for they were (and might still be) stimulating. . food for that mind.. (third person, danger danger).
Quantifying the complexity and similarity of chess openings using online chess community data | Scientific Reports
www.nature.com/articles/s41598-023-31658-w?fromPaywallRec=false
Quantifying human performance in chess - Scientific Reports
www.nature.com/articles/s41598-023-27735-9.pdf
Author Correction- Quantifying human performance in chess - Scientific Reports
www.nature.com/articles/s41598-023-31622-8
Experience shapes non-linearities between team behavioral interdependence, team collaboration, and performance in massively multiplayer online games | Scientific Reports (I must have been in wide mode, or this came from some AI).
www.nature.com/articles/s41598-024-57919-w?fromPaywallRec=false
looked at abs. it must have been perhaps citing one of the others, and I can see an interest for its methodolical purpose.. the video game being more like chess, a sort of methodological control group, for the methodology theory that is the target of paper. for other fields with high level models of their objects of interest (my words, but don,t take that see the quote of the intro (not abstract). I would say it might be science of science research.
> Social sciences are suffering a “replication crisis”1. Disciplines like psychology, team research, organizational behavior, and behavioral economics have faced mounting concerns related to the reproducibility2, replicability3, generalizability4, and overall trustworthiness of their findings5 usually because of external validity issues in experimental research and limited sample sizes in observational studies1. Therefore, we need complementary data sources that can help researchers not only replicate previous results but also to expand them6. In this paper we argue that video games, as interactive digital experiences, have transcended mere entertainment to serve as a unique platform for researchers to delve into intricate domains of human behavior7.
some unexpected interactions from mullti-agent cooperation.. did not read details.. trade off notions.. typical of when we are not dealing with linear combinations of factors but non-linear response over the input domain possibly AND type not just about each factor independently having nonlinear response. In biology and derived (shorcut) sciences mentioned above, nothing can be linear. there are constraints everywhere and infinite growth has obvious diminishing returns.
BTW: It seems as a species we are clueless about what life has "known" (yeah, man , it knows shit). But luckily we have some immutable truths from pre-Gutenberg propagated written things that we can keep reading as if it were of literal knowledge value, and fill all our blind spots of knowledge with that, as the simplest occam razor thing to compute logically from.
There will allways be some fog that such brand of Occam's razor simple bias assumption (when in doubt that is the simplest, otherwise what does simple actually mean, sure one, can rig the question to already defined domain of parameter space, and then have Bayesian argument to select the simplest according to that. Am I too skeptical? or, more likely, not getting it? But really I think chess is a fine platform for thinking about that.
So i might be making an analogy with the purpose of using the video game population data, there as a methodological object of theory and experiements, to shed light on other disciplines processes. Where it is difficulty to see above own premises, notion of external referential maybe? I might be seeking what I want.. be warned. but I know that, part of the process, of hypothesis with clear surgical method of progress on its vaility as more information is found. always smelling something circular, though, when thinking too hard about something.. one brain never enough.
Chess as control problem where conditions and local verifiability of 100% logical consequences (transitively through all the plies, in eachgame, and for all games). Yet, it does not compute from knowing the initial condition to the terminal outcome (initil condition including the player ID variables, itself being object of modelling scientific pursuit, I would add, beyond mere variable pool referential system giving some average competitive pairing win expectations, the true strict interpretation of all ELOS and glickos, strenght if you are desperate have an estimate, but really that would be in ones beholding eyes.).
www.nature.com/articles/s41598-023-31658-w?fromPaywallRec=false
Quantifying human performance in chess - Scientific Reports
www.nature.com/articles/s41598-023-27735-9.pdf
Author Correction- Quantifying human performance in chess - Scientific Reports
www.nature.com/articles/s41598-023-31622-8
Experience shapes non-linearities between team behavioral interdependence, team collaboration, and performance in massively multiplayer online games | Scientific Reports (I must have been in wide mode, or this came from some AI).
www.nature.com/articles/s41598-024-57919-w?fromPaywallRec=false
looked at abs. it must have been perhaps citing one of the others, and I can see an interest for its methodolical purpose.. the video game being more like chess, a sort of methodological control group, for the methodology theory that is the target of paper. for other fields with high level models of their objects of interest (my words, but don,t take that see the quote of the intro (not abstract). I would say it might be science of science research.
> Social sciences are suffering a “replication crisis”1. Disciplines like psychology, team research, organizational behavior, and behavioral economics have faced mounting concerns related to the reproducibility2, replicability3, generalizability4, and overall trustworthiness of their findings5 usually because of external validity issues in experimental research and limited sample sizes in observational studies1. Therefore, we need complementary data sources that can help researchers not only replicate previous results but also to expand them6. In this paper we argue that video games, as interactive digital experiences, have transcended mere entertainment to serve as a unique platform for researchers to delve into intricate domains of human behavior7.
some unexpected interactions from mullti-agent cooperation.. did not read details.. trade off notions.. typical of when we are not dealing with linear combinations of factors but non-linear response over the input domain possibly AND type not just about each factor independently having nonlinear response. In biology and derived (shorcut) sciences mentioned above, nothing can be linear. there are constraints everywhere and infinite growth has obvious diminishing returns.
BTW: It seems as a species we are clueless about what life has "known" (yeah, man , it knows shit). But luckily we have some immutable truths from pre-Gutenberg propagated written things that we can keep reading as if it were of literal knowledge value, and fill all our blind spots of knowledge with that, as the simplest occam razor thing to compute logically from.
There will allways be some fog that such brand of Occam's razor simple bias assumption (when in doubt that is the simplest, otherwise what does simple actually mean, sure one, can rig the question to already defined domain of parameter space, and then have Bayesian argument to select the simplest according to that. Am I too skeptical? or, more likely, not getting it? But really I think chess is a fine platform for thinking about that.
So i might be making an analogy with the purpose of using the video game population data, there as a methodological object of theory and experiements, to shed light on other disciplines processes. Where it is difficulty to see above own premises, notion of external referential maybe? I might be seeking what I want.. be warned. but I know that, part of the process, of hypothesis with clear surgical method of progress on its vaility as more information is found. always smelling something circular, though, when thinking too hard about something.. one brain never enough.
Chess as control problem where conditions and local verifiability of 100% logical consequences (transitively through all the plies, in eachgame, and for all games). Yet, it does not compute from knowing the initial condition to the terminal outcome (initil condition including the player ID variables, itself being object of modelling scientific pursuit, I would add, beyond mere variable pool referential system giving some average competitive pairing win expectations, the true strict interpretation of all ELOS and glickos, strenght if you are desperate have an estimate, but really that would be in ones beholding eyes.).
Join the Dboing's Musings team, to post in this forum