Tuesday, October 9, 2012
The Science of Chess: Fernand Gobet and the Einstellung effect
Since this blog is about Chess and Science, it is only natural to discuss the nexus between the two: scientific studies that examine some aspect of Chess. The game of chess is used by some social science, psychology and neuroscience researchers as a model or tool to examine memory, expertise, decision making skills, or some other process that two combatants engage in over the chessboard.
A few chess players may be already familiar with an example of this type of work, perhaps the highly cited studies by Adriaan de Groot. In some of the studies conducted by de Groot, chess players of different skill levels were tasked with position recall and were told to analyze a position while thinking out loud. This has led to conclusions about the way expert players organize their memory, and how they search and evaluate a position. (I'll probably revisit de Groot in a future post)
A contemporary researcher in this field is Fernand Gobet, a professor at Brunel University, West London. (See his professional homepage). Gobet has dual qualification to study chessplayers, since he is both a cognitive psychologist and an international master (a qualification not unique to the researchers in this field). He has published a number of interesting articles on a variety of aspects of the game, ranging from memory, visualization skills of chess players, and even gender differences between players. Of particular note, Dr. Gobet has attempted to simulate certain aspects of expert memory and skill using commuter models.
Below, I'll give you my reaction to a paper by Dr. Gobet that deals with the interesting Einstellung effect. While I am scientist, cognitive psychology is not my field, so the experts will have to forgive me if I do not assess the work correctly. Also, I provide below only a brief sketch of the work; I might someday blog about the individual topics in more detail, as I learn more about them.
I'd love to hear from you, dear reader, if you have some insight onto to the work of Dr. Gobet or others. In particular, please feel free to let me know if I have gotten something wrong. You can do so by leaving a comment.
The Einstellung effect
“When you see a good move, look for a better one”
The Einstellung effect, as I read it, describes a sort of inflexibility that is tied to (or partially a result of) experience and expertise in a subject. This phenomenon is also known as fixation. It is a rather common sense notion that a person can become blind to innovation because of the success of prior solutions. In other words, expertise and knowledge may lead to overconfidence in one's methods and an unwillingness to consider alternative solutions to a problem. I think that this effect would grow in strength proportional to the prior success rate of the fixated solution.
This effect can be observed in many different fields, including chess. Traditionally, fixation in chess players has been analyzed by tasking participants to solve tactical chess problems (Forced mates). In a study by Saariluoma et al, a series of checkmate puzzles featuring a smothered mate solution were presented to a collection of players (Saariluoma, 1990). Then, problems are introduced in which a smothered mate is still possible, but is not the ideal solution. The extent to which these players still selected the smothered mate in the 2-solution problem quantifies the Einstellung effect. In essence, the participants became fixated on one particular type of solution (the smothered mate motif) through the first series of puzzle, preventing them from finding the other solution in the 2-solution problems (even when the smothered mate was not possible). This effect was demonstrated further by providing players with an 'extinction' problem: a chess puzzle that did not have the smothered mate solution, only the shorter mate from the 2-solution problem.
The Einstellung effect that was observed in several of Saariluoma's experiments was dramatic. The more familiar solution blocked out the more correct one to such an extent that when presented with the 2-solution problem, only 11% of participants found the optimal solution (67% were reported to find the more familiar solution. I assume this means 22% failed to solve the problem.). Importantly, the optimal, but less familiar solution was not intrinsically difficult. This is evident because 69% of players exposed only to problems in which the smothered mate was impossible (the extinction problem, containing only the alternative mating pattern) had found the correct solution. Interestingly, the participants that found the optimal solution also were the strongest out of the group (most of the participants were not very strong, with ratings less than 2100)
The studies by Gobet and colleagues extended the work by Saariluoma and attempted to account for the skill level of the players. This study did not including the priming steps, in which beginners are trained upon a series of puzzles that feature the smothered mate theme immediately before the challenge. Instead, the previous knowledge each participant has is harnessed, since the smothered mate solution present in the puzzle is a very typical version of this mating motif, while the shorter solution relies on awkward pins and lateral Queen movement (These problems are said to be inspired by the Saariluoma study, although I cannot verify this as I do not have access to that paper). Unlike in the Saariluoma study, which primarily used beginners, players representing a range of skill levels were chosen, and differences between the groups were noted. I think that this is a better experimental design, because it tests a more general chess expertise, rather than priming the participants with the smothered mate motif.
Interestingly, the power of the Einstellung effect varied with chess strength. Very strong players (Grandmasters) all chose the most optimal solution to the problem (Bilaic, 2008a). This is despite the fact that some admitted noticing the smothered mate pattern first. This admission was later verified through observation of the eye movement of the players during the test (Bilaic, 2008b). Taken together, this suggests that very strong players were flexible enough to overcome any fixation effect exerted by their previous knowledge (which did exert some effect, as evidenced by the eye movement measurements). This may partly be a consequence of a larger repertoire of tactical motifs and chess knowledge (perhaps the optimal solution was not altogether that unfamiliar to the grandmasters).
Slightly less strong players (International Masters and lower) suffered from the fixation effect during these tests. In fact, as is noted in the discussion to the paper, this effect was strong enough to lower their problem solving ability to that comparable with much lower rated players. When 'distracted' by an familiar pattern, IMs performed on par with Class A players, and Candidate Masters performed on par with Class C players. This is good news for all the untitled players out there! Just guide your opponent into positions in which the more natural move is suboptimal (of course, easier said than done). The effects observed also scaled along with problems of different difficult. If the problem was easier, the effect shifted, and IMs found the optimal solution in all problems, but lower masters still suffered from the Einstellung effect (The reverse was also true; a sufficiently difficult problem could exert this effect on Grandmasters as well; Bilaic, 2008a). Finally, the researchers noted that every player analyzed the position for approximately the same amount of time, regardless of their skill level or success in solving the problem.
This effect was not confined to checkmate puzzles. Additional experiments from the study by Gobet and colleagues found similar effects when the problem involved a win of material (inferior but familiar tactic) versus checkmate (optimal solution). An attempt to extend the analysis to positional problems, where the solution consists of the correct strategic plan, failed because the problems chosen were too difficult (Bilaic, 2008a). Hopefully, this important question will be revisited in the future with the appropriate test problem.
Greater insight might be gained by using more complicated problems, such as the aforementioned failed attempted at using a positional puzzle. As a naive reader of this work, and a tournament chess player, I can't help but wonder how this effect might influence the outcomes of the games I play. Test problems in which the most natural solution does not produce a checkmate or a winning advantage, but rather a draw in a difficult position, would be instructive (the optimal solution should still produce a win). Furthermore, the authors do not address how their work might impact the development of expertise. It would be helpful to know if there are any specific training methods or approaches which might overcome the Einstellung effect, other than keeping an open mind and a fresh perspective.
The studies by Saariluoma, Gobet, and others, produce interesting insight onto the Einstellung effect. All of these studies, however, examine aspects of chess in isolation, with controlled parameters. So is there a way for the average player to benefit from this research, or is it purely academic (which, I remind the reader, isn't necessarily a bad thing). I believe that Emanuel Lasker said it best: "When you see a good move, look for a better one". These studies are a reminder that Lasker's axiom should not only be applied to our chess games, but to any field in which we have expertise. After all, what good is knowledge and expertise if it prevents us from finding the optimal solution to a problem, whatever that problem may be. Especially in Chess and Science, the ability we have to solve a problem should never outpace our ability to question our own solutions.
Bilaic, M., et al (2008a) Inflexibility of experts—Reality or myth? Quantifying the Einstellung effect in chess masters. Cognitive Pyschology, Vol 56 Pg 73-102
Bilaic, M., et al (2008b) Why good thoughts block better ones: The mechanism of the pernicious Einstellung (set) effect. Cognition, Vol 10 Pg 652-661
Saariluoma, P. (1990). Apperception and restructuring in chess players’ problem solving. In K. J. Gilhooly, M. T.
Keane, R. H. Logie, & G. Erdos (Eds.). Lines of thinking: Reflections on the psychology of thought (Vol. 2,
pp. 41–57). Oxford, England: John Wiley & Sons.
Stay tuned! As I digest the rest of Dr. Gobet's work, as well as learn about other researchers, I will share this information with you, my dear readers.