Switch to Desktop Edition

What do I know about cognitive models? Careful selection

2013-7-1516: 34 reading: 10 179 label: model.

In modern science, the concept of "model" can be seen everywhere. Open a textbook of cognitive psychology, and you can see many models from beginning to end. Many famous cognitive scientists and cognitive neuroscientists are also famous for putting forward some famous models. So, what is a model? What is the significance and value of the model? How to build a model? This is something that many beginners are very concerned about. I want to talk about some views in this article based on my own experience. Your knowledge and experience are limited and incorrect. Everyone is welcome to criticize and correct me.

My understanding of models began in middle school. At that time, I loved chemistry, especially Mendeleev and Madame Curie. I was one of their "fans" in China. Mendeleev put forward the periodic table of elements, arranged all the elements discovered at that time in a periodic table according to their valence, and predicted some elements that had not been discovered at that time. A few years later, these elements were discovered by other chemists, and the result was the same as Mendeleev's prediction. From the point of view of modern science, the periodic table of elements is a chemical model, which successfully deals with the relationship between various elements, and has good prediction ability, which can guide future research. I read in the newspaper some time ago that there is also a very famous model in modern physics-the "standard model" of particle physics, which divides elementary particles into quarks, leptons and bosons. The model predicts the existence of 62 elementary particles, among which the Higgs boson, the "God particle", is the last elementary particle that has not been proved to exist. According to the report, there is some experimental evidence that this new type of particle may exist.

After I went to college, I studied psychology and came into contact with many psychological theories, many of which can be expressed by a simple formula. For example, the old behaviorism advocates studying the relationship between stimulus and response, and their theory can be expressed as: S-R, while the new behaviorism advocates that there are intermediary variables (O) between stimulus and response, and behavior is not a direct function of stimulus, but determined by these intermediary variables. Their theory can be expressed as follows: psychologists who study motivation in S-O-R O-R have developed various formulas to explain the role of driving force or inducement in human behavior. For example, Hull believes that the individual's behavioral potential (P) is determined by both driving force (D) and habit intensity (H). Their relationship can be expressed as: P=DxH. From the perspective of modern science, these formulas are just some models. The birth of these theoretical models has a certain experimental basis. For example, the discovery of quasi-sports and the problem-solving behavior experiment of orangutans are the experimental basis of gestalt psychology; The study of fear is the experimental basis of behaviorism; The analysis of dreams is the experimental basis for the birth of psychoanalysis. The above formula expresses the core idea of the theory in the most general way and deals with the relationship between behavior, psychology and stimulus variables. Based on the above theoretical model, people can predict possible behaviors under certain conditions and design experiments to test their hypotheses.

In the mid-1980 s, he began to systematically contact cognitive psychology, and then taught this course for many years. Only then can we understand that modeling is an important feature of cognitive psychology and modern science. 1995 or so, the book Cognitive Psychology, co-edited by teacher Zhang Biyin and me, was originally published by Donghua Bookstore in Taiwan Province Province in traditional Chinese characters, and later by Zhejiang Education Publishing House in simplified Chinese characters. In this book, I systematically expounded my views on the model.

"Building a cognitive model is actually building a cognitive theory (partial or whole). With the help of cognitive model, people can effectively describe the relationship between various components or elements of human cognition, the stages of information processing and the characteristics of transition from one stage to another. Can explain all kinds of facts and phenomena found according to experiments; It can predict the experimental results of cognitive psychology, guide people to further experimental research, and also test the model through simulation on the computer. " In other words, the role of the model is to "describe", "explain", "predict" and "guide" further research.

The establishment of cognitive model is mainly based on the analogy principle. Starting from the observed facts, some abstract and organized concepts are obtained through reasoning, that is, all kinds of facts, phenomena and their relationships are brought into an abstract and organized system.

"A good cognitive model should be: 1) able to express the structure and process of cognition, that is, the stages of cognitive experience, the relationship between stages, what internal processing has been experienced, and how knowledge is represented in cognitive activities; 2) Can explain the regulation of various central control factors (such as needs, emotions, goals, plans, values, etc.). ) about the cognitive system; 3) Ability to accommodate and explain observed facts and phenomena (at least partially); 4) Being able to predict the emergence of new facts and phenomena, that is, having the function of guiding further observation and experiment; 5) It is necessary to absorb the research results of neighboring disciplines, especially the research results of neuroscience, brain science and artificial intelligence. The cognitive model does not directly seek to support its physiological basis, but it should not contradict the findings of neurophysiology. "

At the same time, some of my graduate students have also started specific research on modeling. In a study of 1994, we established a connectionist model of cognition and pronunciation (CMPR), which proposed three main hypotheses, namely, processing hypothesis, representation hypothesis and learning hypothesis. Using the method of distributed representation, the frequency effect, regularity and consistency effect of pictophonetic characters in Chinese character recognition and naming are simulated through training. In another study (1997), we established a semantic-based computational vocabulary decision model (CLDM) by combining distributed representation and local representation. This model simulates the frequency effect and semantic priming effect in Chinese vocabulary recognition. In the future, based on the experimental results of Chinese compound words, we put forward the inner connection model (IIC model for short) (1999), and further verified the model with reverse words as experimental materials in the follow-up research. These studies have helped me to further realize that modeling must be based on existing observations and experiments, have my own theoretical assumptions, and constantly test my theoretical assumptions with new experimental data.

In modeling research, some graduate students have also gone through some detours. There is a graduate student transferred from physics major. He has good training in mathematics and physics, and has also read some physiological works, such as Pavlov's works. He boldly put forward the physical model of psychology. But this model lacks the support of enough experimental results, and he doesn't like to do experiments to test his model. I wanted him to design some experiments to test his hypothesis, but he said, "What I want to do is to propose a model, and it is my brother and sister's business to test the model." He is a bit stubborn and difficult to accept other people's opinions, including teachers. As a result, three years later, the research on the model made little progress, and finally he had to give up. Another student who graduated from the psychology department also put forward a model and wrote to solicit opinions on his model, including writing to Mr. Qian Xuesen. However, the model is too broad to be operated and proved, and finally it has to be abandoned. These two experiences and lessons made me understand a truth: the model must be based on experimental research and must be tested and proved through further research. Models without experimental basis are doomed to be short-lived. In any field, whether a model can exist depends on the degree of being tested.

In our research team, some people like to do experiments, but they don't pay much attention to the construction of theoretical models, which is not good. Research innovation should include theoretical innovation. In a sense, this is a more important and valuable innovation. A person may have done many good experiments, but only theoretical innovation is a more important contribution and can leave a wider impact on the world. Psychologists in China have done a lot of good work in experimental research, but they lack important theoretical contributions with world influence, which is my regret. I didn't do it myself, but I hope that future psychologists in China can display their talents and do valuable work in this field.

The model should be constantly revised and improved. In this respect, Chomsky's research in linguistics sets a good example for us. Since the mid-1950s, Chomsky has constantly revised and developed his own language model. His earliest theoretical model was called classical theory (1957- 1965), and was later revised into standard theory (1965- 1970) and extended standard theory (1970-/kloc-0). Chomsky attaches great importance to the new progress in linguistics and pays close attention to the new achievements in psychology. He doesn't stick to his point of view, nor is he satisfied with the status quo. This is his greatness and respect, and it is also the place where his theoretical model has vitality. It takes a lot of theoretical courage to put forward a model, and even more theoretical courage to correct your own model. We must face new experimental findings, dare to give up our original incorrect assumptions and constantly put forward new assumptions, and have the spirit of self-criticism in academic research. Everyone is lazy. When people become famous, they tend to maintain their original assumptions, even though they know they are wrong, they are unwilling to change them. This is a taboo for researchers. Every time I introduce Chomsky's theoretical model in class, I will emphasize that it is a characteristic of Chomsky's scholarship to be brave and diligent in correcting his model, and it is also where we should learn from him.

With the rise and development of cognitive neuroscience, modeling is a common method to study the mechanism of behavioral brain. This includes functional model, structural model, model with unified structure and function, etc. In Ruglia's words, the latter model can also be called the functional model of the brain. Like cognitive model, brain model is also a theory or hypothesis, which is based on existing findings and can be used to predict and guide further research. Compared with the traditional cognitive model, the modern brain model is more likely to reveal the internal mechanism of psychological activities at the levels of behavior, neural network, cells and molecules, predict possible phenomena, and thus realize the regulation of behavior and psychological activities. In recent years, we have seen many theoretical models based on meta-analysis, which are not only based on our own experimental research, but also on a large number of experimental studies by others, so they are more convincing. It has become an important mission of cognitive neuroscience to test and modify the model and constantly build new models according to new findings.