WELLCOME TO OUR SCIENCE CORNER!!
You've may not be aware of it, but since the end of the nineteenth century, chess has been a popular topic of research in cognitive psychology, the field of scientific psychology that studies perception, memory, learning, and thinking.
Some of the most striking phenomena characterizing expert behaviour were first discovered by De Groot (1978) and Chase and Simon (1973) in their work on chess, and later replicated by researchers in other domains. There are at least two reasons why chess was chosen. Its competitive nature makes it possible to precisely differentiate between skill levels, which is not the case in many domains of expertise studied in psychology, such as medicine and physics. Chess also offers a nice balance between simplicity of the rules and complexity of the space of possible games; this enables both the design of elegant experiments and a fruitful cross-fertilization with mathematics and artificial intelligence.
In this section you'll find summaries of cognitive psichology articles that investigate expertise and knowledge and skill acquisition using chess.
Perceptual and semanthical units.
the brain’s tendency to stick with familiar ideas can literally blind us to superior solutions
Engage in appropiately challenging tasks that are chosen with the goal of improving a particular skill.
THE MAGICAL NUMBER SEVEN
TRAINING IN CHESS: A SCIENTIFIC APPROACH Fernand Gobet & Peter J. Jansen
This article is definitelly worth reading if you're a chess player willing to improve.
There's quite a few thing psychology discovered about chess player worth knowing:
1. Chess players have a highly efficient mode of (high-level) perception. They can access the key elements of a position rapidly.
2. Chess players show a remarkable memory for chess positions and games. This ability typically does not extend beyond chess.
3. Chess knowledge is encoded at several levels, in particular at a low, perceptual level, where patterns of pieces are stored, and at a high, conceptual level, where information about plans, evaluation, etc., is stored. These various types of encoding, with rich indexing and a high level of cross-referencing, account for chess players’ excellent professional memory.
4. Chess players search highly selectively. It is rare that they analyze more than one hundred positions in the search tree before choosing a move.
5. There is no difference between the search algorithm of class A players (Elo 1800-2000) and that of Grandmasters.
6. Masters lose relatively little of their skill when they play simultaneous games or speed chess.
The author revisit the training based on the scientific evidence collected and they also debunk a series of widelly spread myths.
The recomendations made by the authors can be summarized as follows:
- Teach from the simple to the complex
- Teach from the known to the unknown
- The elements to be learnt should be clearly identified
- Use an improving spiral, where you come back to the same concepts and ideas and add increasingly more complex new information
- Focus on a limited number of types of standard problem situations, and teach the various methods in these positions thoroughly
- Repetition is necessary. Go over the same material several times, using varying points of view and a wide range of examples
- Avoid spending too much time on historical and anecdotal details
- At the beginning, don’t encourage students to carry out their own analysis of well-known problem situations, as they do not possess the key concepts yet
- Encourage students to find a balance between rote learning and understanding
- Encourage students to keep information in a central filing system or a database
- The ability to look ahead possible moves is made possible by knowledge. Training this ability per se by exercises aiming at improving short-term memory or visualization (e.g. by playing blindfold chess) is not recommended
CHESS PLAYERS' THINKING REVISITED Fernand Gobet
This article from 1998 studies the differences in thought process between chess masters and amateurs.
Traditionaly, to the article date, there were 2 main approaches to chess expertise. Pattern recognition vs. Search Depth. Simply, or bluntly put each theory can be summarized as follows:
- Pattern Recognition: Masters recognize a position easily and retrieve from memory information related to it easily. Thus they are faster and do not need to search or navigate thounsands of options.
- Seach Depth: Masters search deeper, thus they are able to find better moves.
The article tries to settle which of both approaches is correct analyzing new data, based on their ability to predict the correct experiment result.
Based on their assumptions, when comparing amateurs to masters performing different tasks (simplified from the ones in the article), their predictions are:
|Pattern Recognition theory
|Search Deph theory
|choose better moves
|select move faster
|search more nodes
|Master equal or less
|Master equal or less
|need of time
The results of the experiment corroborate the Pattern Recognition approach, but also confirm that there is a difference in mean depth search (meaning that on average master calculate a little bit deeper). Thus it seems that both approaches instead of being mutually exclusive are complementary views of the same process that can be integrated.
The author's proposal is to integrate both approaches in Template Theory. Rooted in Pattern Recognition approach, Template Theory states that memory does store Templates, rather than chunks.
Template = chunks with slots + associated moves and plans + associated templates
PERCEPTION IN CHESS William G. Chase Herbert A. Simon
Since De Groot's experiment chess masters' perception has been in the spotlight. And it seems pretty much clear that masters seem to perceive chess positions as chunks of several RELATED pieces.
In this article from 1973 the authors revisit De Groot's experiment trying to lay out a solid base for the chunking theory.
- Perception task: reconstruct a chess position while it remains visible, measuring the number of glances to complete the task and the accuracy of the reconstruction
- Memory task: reconstruct a chess position after a brief exposure to it, using piece placement timing and order to uncover clusters of related pieces and the accuracy of the reconstruction
Once again, the master outperformed amateur and beginner in LEGAL positions. But not in random (NON-legal) positions.
This led to think that memory of positions depends on recognizing familiar configurations of pieces (a.k.a. chunks).
Several hypothesis where used:
- pieces placed after a single glance are part of the same chunk
- pieces placed with time intervals of less than 2 seconds between them are part of the same chunk
Chunking theory states that perception and long term memory are developed all together and thus experts and novices do not perceive the same thing when exposed to the same situation.
DE GROOT'S EXPERIMENT
During the mid sixties De Groot made a ground breaking study in expertise using chess players.
As one of the experiments used for the study has been widely repeated and discussed it is worth to go through it here so that the reader is familiar with it.
During the study De Groot presented players of different strength (from amateurs to, GMs, including a World Champion) several piece setups over a chess board (for a restricted amount of time, 5 seconds).
The masters showed:
- Greater ability to recall and reproduce the LEGAL chess positions shown
- Similar ability to recall positions (compared to amateurs) when exposed to RANDOM (NON-legal) positions
This showed a great remarkable short term memory for LEGAL chess positions in masters.
Strikingly, this was not the only experiment conducted, but for all other tasks and processes evaluated (i.e. number of moves evaluated, thought process, search depth, etc.) masters showed the same performance that amateurs, weaker players.
From that moment on chess mastery has been tied to perception and memory enconding.
CHUNKING MODELS OF EXPERTISE: IMPLICATIONS FOR EDUCATION Fernand Gobet
Chunking theories explain how knowledge is preceived and created. Chunks, perceptual/semantic units, are encoded for later retrieval in LTM (Long Term Memory).
In this article explores how 2 different models of chunking theory (EPAM & CHREST) may suggest how to conduct education more efficiently.
The implications for education can be summarized as follows:
The role of practice and the cost of acquiring knowledge
Dedication is essential to acquire knowledge to create and grow the perceptual and memory discrimination network. Thus deliberate practice is key, but as the author puts it Practice needs to be tailored to the goal of improving performance, as many invalid or irrelevant chunks can be acquired, which may be counterproductive.
This raises the question about the value of several online tactical trainers that expose the user to random exercises. Are they
The role of perception in acquiring knowledge
Relationship between abstract knowledge and perception is under discussion but it seems failry clear that: perceptual skills, anchored in concrete examples, play a central role in the development of expertise, and that conceptual knowledge is later built on such perceptual skills. So it is important to develop perceptual chunks but without over emphatizing on it.
The role of teachers and tutorial systems
Their role should be directed to ...acquisition of perceptual chunks, an important role for teachers (both human and artificial) is to direct learners’ attention to the key features of the material to learn because Presenting components of the right size and difficulty will help students direct attention to the important features of the material, and in turn help the acquisition of perceptual chunks that are appropriate....
The question of transfer
Chunking theories and expertise investigation as a whole, have demonstrated that transfer is low and just restricted to situations ...when there is an overlap between the components of the skills required in each domain.
Order effects in learning
Not all chunking theory simulators reached the same conclussions, but ...simulations with CHREST showing that changes within the ordering of the learning set have a rather strong impact on the structure of the discrimination network and on the speed of information retrieval. Thus the order in which curriculum is thaught seems to be relevant.
Productions are rule based, condition-action pairings. It seems fairly demonstrated that chunks are often encoded as conditions to actions. Two conclussions are reached in the study:
- ...the importance of balancing the acquisition of the condition and action parts of productions
- One possible reason behind the success shown by some tutoring systems... is that, in addition to providing individual feedback, these systems, by forcing students to solve problems, encourage the creation of more discriminated conditions in the productions they are acquiring.
Schemata, or templates, are generalizations of chunks. Chunks with slots.
...without variation, schemata cannot be created. For example, in the case of elementary mathematics, presenting a narrow range of problems will hamper the acquisition of a sufficient variety of chunks and links connecting them, and, consequently, schemata are not likely to be formed.
Declarative, procedural, and conceptual knowledge
Relation between different types of knowledge is tricky. And it may seem more conceptual than real, as chunking models have not shown clear distinction between Declarative knowledge (knowing what) and Procedural knowledge (knowing how) creation: ...the learning of both types of knowledge occurs incrementally and implicitly Conceptual: ...to give sufficient basis to conceptual knowledge it is necessary to acquire a richly-connected network of links joining productions and schemata, which are accessible through perceptual chunks.
Acquiring multiple representations
In fact, learning multiple representations requires duplicating the same information in different formats. Although redundancy is certainly an important aspect of human memory and understanding, CHREST draws our attention to the fact that it also has a cost, in particular with respect to the time spent in learning
The use of multiple representations is only one of many learning devices that have flourished with the advent of modern educational technology. ...Chunk-based models actually warn us against any excess of optimism in the use of new technologies, as long as they do not help circumvent the key limiting constants of human cognition (i.e. attention, STM, and learning rates).
The role of individual differences and talent
There are vast individual differences in people’s cognitive abilities (Ackerman, 1987; Sternberg, 2000), both at the novice and expert levels.
First, while individual differences tend to be diluted by large amounts of practice, they play a large role in the early stages of studying a domain, which characterizes much of classroom instruction. Second, as seen earlier, taking into account individual differences may lead to better instruction, because instruction can be optimized for each student, including feedback on progress, organization of material, and choice of learning strategies to be taught.
THE ROLE OF DELIBERATE PRACTICE IN CHESS EXPERTISE
Neil Charness, Michael Tuffiash, Ralf Krampe, Eyal Reingold and Ekaterina Vasyukova
In this study try to stablish which of the following three activities: studying alone, tournament practice (competitive play) and formal instruction is a stronger predictor of chess expertise.
After couple of surveys were conducted over a 375 chess players group from different countries, they found out that hours devoted to chess study alone were the strongest predictor of chess success.
THE RELATIONSHIP BETWEEN COGNITIVE ABILITY AND CHESS SKILL:A COMPREHENSIVE META-ANALYSIS Alexander P. Burgoyne, Giovanni Sala, Fernand Gobet, Brooke N. Macnamara, Guillermo Campitelli & David Z. Hambrick
THE GEOMETRY OF EXPERTISE Maria Juliana Leone & Mariano Sigman
RAPID CHESS: A MASSIVE-SCALE EXPERIMENT Diego Fernandez Slezak1, Pablo Etchemendy and Mariano Sigman