Your network blocks the Lichess assets!

lichess.org
Donate

Networks for expertise in the chess player's brain

Thanks for your continued research.

Just some notes on the problem of chess and Cognitive science, your quote "Chess is the drosophila of artificial intelligence", means chess serves as the ideal, simplified model organism for testing and developing AI algorithms, chess use for cognitive science is unclear. As after the Cognitive Revolution (~1956 based on Shannon Information Theory). most of the testing on chess was from people more interested on Computer Algorithms than the human mind, as the founders of psychology were more skeptical of the use of chess for helping solve problems of mind, including expertise.

The problem of studying experts, as chess was original chosen to study expertise due to the rating system and the clearly differentiated levels of expertise, where a 400 rating point difference means the superior player will win over 90% of the time. Meaning from novice to master there are at least 5 clear skill levels where each skill level can beat the lower skill level almost every game. As Michelene Chi (novice in Chase & Simon chess research) pointed out in the 1980s defining two approaches to the study of expertise: the absolute and the relative. The absolute approach studies exceptional individuals who are known experts and usually assumes that greatness or creativity arises from chance and unique, innate talent.

A useful study for psychology and expertise, the path from novice to intermediate is much more informative than expert performance. Study experts themselves provides little insight. Any conclusion from the 'neuro correlates of consciousness' would have to test novices to master, with the stronger focus on the differences in the early skill levels. The trends of experts would only be useful is shown to be true on a level to level basis.

And as quoted, "Chess is the drosophila of artificial intelligence", not human intelligence or psychology, the studies on music, education and athletics has replaced the chess studies. Although K. Anders Ericsson early studies included chess 1993, later expertise studies started excluding chess for various reasons. Now longitudinal studies on hundreds of thousands of students, athletes, musicians has completely negated the chess research.

Although chess still somewhat remains the drosophila of artificial intelligence, like for Google Alpha and Deep Mind, but even in AI, chess has taken a back seat.

Chess players are in a bad situation for cognitive science, as most professional chess players and trainers tend to reject the scientific evidence and psychological studies (for obvious reasons), that hope more serious intellectuals on LICHESS will struggle with.

Thanks for your continued research. Just some notes on the problem of chess and Cognitive science, your quote "Chess is the drosophila of artificial intelligence", means chess serves as the ideal, simplified model organism for testing and developing AI algorithms, chess use for cognitive science is unclear. As after the Cognitive Revolution (~1956 based on Shannon Information Theory). most of the testing on chess was from people more interested on Computer Algorithms than the human mind, as the founders of psychology were more skeptical of the use of chess for helping solve problems of mind, including expertise. The problem of studying experts, as chess was original chosen to study expertise due to the rating system and the clearly differentiated levels of expertise, where a 400 rating point difference means the superior player will win over 90% of the time. Meaning from novice to master there are at least 5 clear skill levels where each skill level can beat the lower skill level almost every game. As Michelene Chi (novice in Chase & Simon chess research) pointed out in the 1980s defining two approaches to the study of expertise: the absolute and the relative. The absolute approach studies exceptional individuals who are known experts and usually assumes that greatness or creativity arises from chance and unique, innate talent. A useful study for psychology and expertise, the path from novice to intermediate is much more informative than expert performance. Study experts themselves provides little insight. Any conclusion from the 'neuro correlates of consciousness' would have to test novices to master, with the stronger focus on the differences in the early skill levels. The trends of experts would only be useful is shown to be true on a level to level basis. And as quoted, "Chess is the drosophila of artificial intelligence", not human intelligence or psychology, the studies on music, education and athletics has replaced the chess studies. Although K. Anders Ericsson early studies included chess 1993, later expertise studies started excluding chess for various reasons. Now longitudinal studies on hundreds of thousands of students, athletes, musicians has completely negated the chess research. Although chess still somewhat remains the drosophila of artificial intelligence, like for Google Alpha and Deep Mind, but even in AI, chess has taken a back seat. Chess players are in a bad situation for cognitive science, as most professional chess players and trainers tend to reject the scientific evidence and psychological studies (for obvious reasons), that hope more serious intellectuals on LICHESS will struggle with.

Thanks for reading! A few comments below inserted into your discussion:

@DIAChessClubStudies said in #2:

Thanks for your continued research.

Just some notes on the problem of chess and Cognitive science, your quote "Chess is the drosophila of artificial intelligence", means chess serves as the ideal, simplified model organism for testing and developing AI algorithms, chess use for cognitive science is unclear. As after the Cognitive Revolution (~1956 based on Shannon Information Theory). most of the testing on chess was from people more interested on Computer Algorithms than the human mind, as the founders of psychology were more skeptical of the use of chess for helping solve problems of mind, including expertise.

I think that skepticism is certainly warranted, but I also think that chess offers a nice sort of "sandbox" for studying expertise so long as one is wary about what kinds of generalization are meaningful to make. My current feeling is that the strength of chess engines makes this even more true - besides the properties of players we have access to through rating data, we also have much better data about games and chess positions than we used to.

A useful study for psychology and expertise, the path from novice to intermediate is much more informative than expert performance. Study experts themselves provides little insight. Any conclusion from the 'neuro correlates of consciousness' would have to test novices to master, with the stronger focus on the differences in the early skill levels. The trends of experts would only be useful is shown to be true on a level to level basis.

I think this is a great insight. I would love to see more work comparing performance across those earlier stages of learning! I suspect that there are a few things working against this as a trend in cognitive research, though. For one, we often try to target moderate to large effect sizes and the difference between a novice and an expert thus seems more appealing than the presumably smaller differences between beginner and intermediate players. I think the other factor is just that the outstanding abilities of experts are more salient to people - that doesn't mean they're more interesting or important scientifically, though!

There are a few studies that have included intermediate players or defined expertise with a more inclusive ELO threshold, but it would be great to see more of this kind of thing.

And as quoted, "Chess is the drosophila of artificial intelligence", not human intelligence or psychology, the studies on music, education and athletics has replaced the chess studies. Although K. Anders Ericsson early studies included chess 1993, later expertise studies started excluding chess for various reasons. Now longitudinal studies on hundreds of thousands of students, athletes, musicians has completely negated the chess research.

I think "negated" is a bit strong, no? I think that work still has meaning even if other domains have become more prominent.

Chess players are in a bad situation for cognitive science, as most professional chess players and trainers tend to reject the scientific evidence and psychological studies (for obvious reasons), that hope more serious intellectuals on LICHESS will struggle with.

There are a few trainers out there that do incorporate cognitive science into their practice, but I do think it would be great if there was more work to bridge the gap between cog. sci. research and chess instruction. Thanks again for your comments and best wishes!

Thanks for reading! A few comments below inserted into your discussion: @DIAChessClubStudies said in #2: > Thanks for your continued research. > > Just some notes on the problem of chess and Cognitive science, your quote "Chess is the drosophila of artificial intelligence", means chess serves as the ideal, simplified model organism for testing and developing AI algorithms, chess use for cognitive science is unclear. As after the Cognitive Revolution (~1956 based on Shannon Information Theory). most of the testing on chess was from people more interested on Computer Algorithms than the human mind, as the founders of psychology were more skeptical of the use of chess for helping solve problems of mind, including expertise. I think that skepticism is certainly warranted, but I also think that chess offers a nice sort of "sandbox" for studying expertise so long as one is wary about what kinds of generalization are meaningful to make. My current feeling is that the strength of chess engines makes this even more true - besides the properties of players we have access to through rating data, we also have much better data about games and chess positions than we used to. > A useful study for psychology and expertise, the path from novice to intermediate is much more informative than expert performance. Study experts themselves provides little insight. Any conclusion from the 'neuro correlates of consciousness' would have to test novices to master, with the stronger focus on the differences in the early skill levels. The trends of experts would only be useful is shown to be true on a level to level basis. I think this is a great insight. I would love to see more work comparing performance across those earlier stages of learning! I suspect that there are a few things working against this as a trend in cognitive research, though. For one, we often try to target moderate to large effect sizes and the difference between a novice and an expert thus seems more appealing than the presumably smaller differences between beginner and intermediate players. I think the other factor is just that the outstanding abilities of experts are more salient to people - that doesn't mean they're more interesting or important scientifically, though! There are a few studies that have included intermediate players or defined expertise with a more inclusive ELO threshold, but it would be great to see more of this kind of thing. > > And as quoted, "Chess is the drosophila of artificial intelligence", not human intelligence or psychology, the studies on music, education and athletics has replaced the chess studies. Although K. Anders Ericsson early studies included chess 1993, later expertise studies started excluding chess for various reasons. Now longitudinal studies on hundreds of thousands of students, athletes, musicians has completely negated the chess research. > I think "negated" is a bit strong, no? I think that work still has meaning even if other domains have become more prominent. > > Chess players are in a bad situation for cognitive science, as most professional chess players and trainers tend to reject the scientific evidence and psychological studies (for obvious reasons), that hope more serious intellectuals on LICHESS will struggle with. There are a few trainers out there that do incorporate cognitive science into their practice, but I do think it would be great if there was more work to bridge the gap between cog. sci. research and chess instruction. Thanks again for your comments and best wishes!

Excellent blog, very interesting! Thank you :)

Excellent blog, very interesting! Thank you :)

@somethingpretentious said in #5:

Excellent blog, very interesting! Thank you :)

Thank you for reading!

@somethingpretentious said in #5: > Excellent blog, very interesting! Thank you :) Thank you for reading!

Hi,
found your description and likely (not read yet) the linked publications very interesting. Thanks for summarizing

Perhaps it is mentioned somewhere (but didn’t find it):

how large are the variations (error bars) ?
Perhaps a follow up study(?) would investigate the variation for people with different elo ranges in this
(saw only “experts” and novices mentioned).

anyway many thanks for having put time into writing this article. michael

Hi, found your description and likely (not read yet) the linked publications very interesting. Thanks for summarizing Perhaps it is mentioned somewhere (but didn’t find it): how large are the variations (error bars) ? Perhaps a follow up study(?) would investigate the variation for people with different elo ranges in this (saw only “experts” and novices mentioned). anyway many thanks for having put time into writing this article. michael

I am curious how the expertise for 3+0 compares to the expertise for 3+2. Some players play both time controls, so there should be an abundance of data (although it may be laborious to analyze).

I am curious how the expertise for 3+0 compares to the expertise for 3+2. Some players play both time controls, so there should be an abundance of data (although it may be laborious to analyze).

@michaeld said in #7:

Hi,
found your description and likely (not read yet) the linked publications very interesting. Thanks for summarizing

Perhaps it is mentioned somewhere (but didn’t find it):

how large are the variations (error bars) ?
Perhaps a follow up study(?) would investigate the variation for people with different elo ranges in this
(saw only “experts” and novices mentioned).

anyway many thanks for having put time into writing this article. michael

Thanks for reading! The figures depicting the correlations between the different theoretical RDMs and the brain-based RDMs have error bars on them (and are flagged for statistical significance) so that tells you a little bit about the variation they measured. That represents the last stage in a rather long analysis pipeline, though, so you might want to know about the variability in the raw data: How variable were the BOLD measurements made during the scan, for example? This data isn't in the main text and depending on the analysis plan, one might average over trials or participants to get to a single value you'll work with from then on, or you may use tools like linear mixed-models to explicitly estimate different sources of variability.

I definitely agree that trying to cover more of the intermediate range of ability would be excellent.

@michaeld said in #7: > Hi, > found your description and likely (not read yet) the linked publications very interesting. Thanks for summarizing > > Perhaps it is mentioned somewhere (but didn’t find it): > > how large are the variations (error bars) ? > Perhaps a follow up study(?) would investigate the variation for people with different elo ranges in this > (saw only “experts” and novices mentioned). > > anyway many thanks for having put time into writing this article. michael Thanks for reading! The figures depicting the correlations between the different theoretical RDMs and the brain-based RDMs have error bars on them (and are flagged for statistical significance) so that tells you a little bit about the variation they measured. That represents the last stage in a rather long analysis pipeline, though, so you might want to know about the variability in the raw data: How variable were the BOLD measurements made during the scan, for example? This data isn't in the main text and depending on the analysis plan, one might average over trials or participants to get to a single value you'll work with from then on, or you may use tools like linear mixed-models to explicitly estimate different sources of variability. I definitely agree that trying to cover more of the intermediate range of ability would be excellent.

@Toadofsky said in #8:

I am curious how the expertise for 3+0 compares to the expertise for 3+2. Some players play both time controls, so there should be an abundance of data (although it may be laborious to analyze).

Are you thinking of comparing people who are especially good with increment vs. without? I could see that being really interesting and looking for specific time control expertise in general would be cool. For this study, the trickiest bit would be getting enough of players from the groups you're interested in to come in for the fMRI.

@Toadofsky said in #8: > I am curious how the expertise for 3+0 compares to the expertise for 3+2. Some players play both time controls, so there should be an abundance of data (although it may be laborious to analyze). Are you thinking of comparing people who are especially good with increment vs. without? I could see that being really interesting and looking for specific time control expertise in general would be cool. For this study, the trickiest bit would be getting enough of players from the groups you're interested in to come in for the fMRI.