What is Cognitive Science?

Cognitive science is an iconic emerging research category of world science in the 20th century. As a cutting-edge cutting-edge discipline that explores the working mechanism of the human brain or mind, it has attracted widespread attention from scientists around the world. It is generally believed that the basic viewpoints of cognitive science were initially scattered in some separate special disciplines from the 1940s to the 1950s, and have been greatly developed since the 1960s.

[rèn zh k xué]

According to E. Sheener, the director of the Institute of Cognitive Science at the University of Alden, the term "cognitive science" was used by Longgate Dickins in 1973, and in the 1970s Only gradually became popular in the late 1950s. In 1975, the Alfred P. Sloan Foundation (a private research funding agency in New York City) began to consider supporting interdisciplinary research programs in cognitive science. The foundation's funding continues to this day. The institutionalization of this new discipline has played an important role. Because the Sloan Foundation has played a decisive role in advancing cognitive science by organizing the first cognitive science conference and establishing research programs.

Mind is the function of the brain and nerves, and the bridge between the brain and the mind is knowledge.

Cognitive science is the theory and doctrine of mental research. In 1975, thanks to the investment of the famous Sloan Fund in the United States, American scholars

Cognitive science is not yet mature, and as an independent discipline, it has not been sufficiently unified and integrated. There is also great disagreement about what cognitive science is. On October 1, 1978, the "State Council on Cognitive Science" submitted a report to the Sloan Foundation. (Schiller, translated by Shi Qi, 1989) defined cognitive science as "the interaction Study of Principles. " The report authors then developed this definition in two directions. The first is episodic: it lists the branches of human cognitive science and the cross-links between them. The sub-fields listed are computer science, psychology, philosophy, linguistics, anthropology, and

The understanding of the scope of cognitive science can also be viewed from the content of cognitive science. The main contents involved in cognitive science so far include perception (including pattern recognition), attention, memory, language, thinking and appearance, Consciousness etc. This seems to be the concern of psychologists, but it is also the concern of philosophers, linguists, computer scientists, neurophysiologists, and anthropologists. It is just that researchers with different professional backgrounds have adopted different research methods for these same issues. Chinese scholar Li Boyo pointed out that artificial intelligence, cognitive psychology and psycholinguistics are the core disciplines of cognitive science, and neuroscience, anthropology and philosophy are peripheral disciplines of cognitive science.

Because of the complexity of cognitive systems, multi-dimensional research is needed. Cognitive science needs to use the tools and methods used by multiple disciplines to conduct comprehensive comprehensive research on cognitive systems in a complete sense. It can be said that the achievements of cognitive science so far are closely related to its interdisciplinary research methods. However, interdisciplinary research methods have also brought many problems and challenges to cognitive science.

Cognitive science is the science that studies the process of human perception and thinking information processing, including from sensory inputs to complex problem solving, intelligent activities from individual humans to human society, and the nature of human intelligence and machine intelligence. Cognitive science is the result of the interdisciplinary development of modern psychology, information science, neuroscience, mathematics, scientific linguistics, anthropology, and even natural philosophy.

The rise and development of cognitive science indicates that the research on human-centered cognitive and intelligent activities has entered a new stage. The study of cognitive science will enable human beings to understand and control themselves, and raise human knowledge and intelligence to unprecedented heights. The phenomena of life are intricate and complicated, many problems have not been well explained, and the content that can be learned from them is also numerous and multifaceted. How to extract the most important and critical problems and corresponding technologies from them is a goal that many scientists have long pursued. To solve many difficulties faced by mankind in the 21st century, such as the large demand for energy, environmental pollution, depletion of resources, and population expansion, it is not enough to rely on existing scientific achievements. We must learn from biology and find a new path for technological development.

Cognitive science perception

The expression of perceptual information is the basic problem of perceptual research, and the basis for studying other levels of cognitive processes. Where does the perception process begin? Are those variables in the external physical world of psychological significance? What is calculated as a perceptual computing model? These questions surrounding the expression of perceptual information are the questions that must be answered first to establish any perception and theories and theoretical models related to perception, whether human or computer. The study of perceptual information expression can have different levels of problems, including problems such as perceptual organization, perceptual learning, perceptual dynamic memory, and face recognition.

Combining experimental research in cognitive neuroscience with computer vision research at the level of computational theory, the level of knowledge expression of the brain, and the level of computer implementation, a new theory (or idea) and a solution to the above scientific problems will be proposed.

Cognitive science learning

Learning is a basic cognitive activity, a process of accumulation of experience and knowledge, and a process of grasping and understanding external things in a related manner in order to improve the performance of system behavior.

The neurobiological basis of learning is the plasticity change of the synaptic structure between the nerve cells, which has become a very active research field in contemporary neuroscience. The synaptic plasticity condition is that the synaptic connection is strengthened when the presynaptic fibers and the associated post-synaptic cells are excited simultaneously. In 1949, Canadian psychologist Hebb proposed the Hebb learning rules. He envisioned that the synapses involved in the learning process would change, resulting in the enhancement of synaptic connections and the improvement of transmission efficiency. Hebb learning rules become the basis of connected learning. A neural network is a widely parallel interconnected network of simple units with adaptability. Kohonen proposed a self-organizing mapping network. Haken proposed a cooperative associative memory network based on the law of synergistic formation structure and competition for development. Amari proposed the use of differential manifolds and statistical reasoning to study neural networks. Based on Amari theory, Shi Zhongzhi and others proposed a neural field model, which consists of a field organization model and a field effect model.

Perceptual learning is learning that takes place at the perceptual level. It mainly studies how to obtain relevant abstract data from raw data input by low-level sensors. Perceptual learning mainly considers the methods of transforming from unstructured and semi-structured information to structured information through visual and auditory learning. It also studies the semantic description of images and its rapid extraction technology. It also studies the attention mechanism and metacognition in perceptual learning.

Cognitive learning theory believes that there is a corresponding thinking process behind human behavior. The change in behavior is observable, and the inner activity of the learner can be inferred through the change in behavior. In the theory of cognitive learning, such as the theory of meaningful learning (also known as assimilation theory) proposed by Ausubel, the core idea is that obtaining new information mainly depends on the existing concepts in the cognitive structure; meaning learning is through new information and learning Only the existing conceptual interactions in the cognitive structure of the person can occur; as a result of this interaction, the assimilation of the meaning of the new and the old knowledge is caused. The information processing learning theory proposed by Gagne likens the learning process to a computer's information processing process. The learning structure is composed of a perception register, short-term memory, long-term memory, controller, and output system. The cognitive process can be divided into selective reception , Monitor, adjust, repeat, refactor. A very critical part of this information processing process is executive control and expectations. Execution control refers to the impact of existing learning experience on the current learning process, and expectation refers to the impact of the motivation system on the learning process. The entire learning process is carried out under the action of these two parts.

Introspective learning is a learning process of self-reflection, self-observation, and self-knowledge. With the support of domain knowledge and example libraries, the system can automatically select and plan machine learning algorithms to better knowledge discovery of massive amounts of information.

Implicit learning is the process of unconsciously acquiring complex knowledge that stimulates the environment. In implicit learning, people don't realize or state what rules govern their behavior, but they learn them. After the mid-1980s, implicit learning has become the hottest and most concerned topic in the field of psychology, especially in the field of learning and cognitive psychology, and has become one of the most important topics that will have a profound impact on the development of cognitive psychology. . Implicit learning has the following three characteristics:

Implicit knowledge can be generated automatically without the need to consciously discover the explicit rules of task operations;

Implicit learning is general, and it is easy to generalize to different symbol sets;

Implicit learning is unconscious, and the knowledge acquired implicitly cannot be expressed in a language system.

Cognitive science language

During human evolution, the use of language has differentiated the functions of the two hemispheres of the brain. The emergence of the language hemisphere clearly distinguishes humans from other primates. Some studies have shown that the left hemisphere of the human brain is related to serial, sequential, and logically analyzed information processing, while the right hemisphere is related to parallel, visual, and non-timed information processing.

Language is a system composed of speech as the shell, vocabulary as the material, and grammar as the rule. Language is usually divided into two categories: spoken and written. The expression of spoken language is sound, and the expression of text is image. Spoken language is much older than words. Individuals learn a language before they learn spoken words, and then learn words.

Language is the most complex, systematic, and widely used symbol system. Language symbols not only represent concrete things, states or actions, but also abstract concepts. Chinese is significantly different from the Indo-European languages with its unique lexical and syntactic systems, text systems and phonetic tone systems. It has a unique style of tight combination of sound, form and meaning. Concepts are thinking forms that reflect the unique attributes of things. Concepts are closely related to words. The emergence and existence of concepts must depend on words. Words can express other things because people have corresponding concepts in their heads. Therefore, words are the language form of concepts, and concepts are the ideological content of words.

Studying Chinese from the three levels of nerves, cognition and computing gives us an excellent opportunity to open the door to intelligence. Research on Chinese cognitive psychology has a long history, and has achieved world-class research results. However, most of these studies focus on Chinese characters and vocabulary, and higher-level syntax and sentence processing need to be further explored. The study of the entire speech chain is not systematic enough, especially not much is known about the language processing mechanism of the brain. In the field of intelligent systems, China attaches great importance to Chinese computer information processing. It has invested a lot of funds to support the research and development of computational linguistics, machine translation, and natural language understanding systems, and has achieved a large number of important results. But on the whole, there are many outstanding problems in the intelligent processing of language information. The solution must be based on the research of cognitive science and guided by new theories to make a breakthrough.

In 1991 Mayeux and Kandel proposed a new language information processing model based on the Wernicke-Geschwind model. The linguistic information of auditory input is transmitted from the auditory cortex to the corner back, then to the Wernicke area, and then to the Broca area. The language information of visual input is directly transmitted from the visual joint cortex to the Broca area. Visual perception and auditory perception of a word are processed independently of each other by different sensory patterns. These pathways each reach the Broca area independently, as well as more advanced areas related to linguistic meaning and expression. The working mechanism of each step of the language processing pathway in the brain needs further study.

Use mathematical methods to study the language and find the forms, models, and formulas of the language structure, so that the grammatical rules of the language can be as systematic and formal as mathematical symbols and formulas, and can be used to generate unlimited sentences. The famous American linguist Chomsky proposed the formal grammar of language in 1956, which established the theoretical foundation for language information processing. In 1996, Yip and Sussman proposed that the use of two-way constraint propagation mechanism in the rules of phonetics can explain how auditory signals at the neural level correspond to symbols of the thinking level.

It is worth mentioning that machine translation involves multiple disciplines such as linguistics, computer science, cognitive science, and mathematics, and is a cutting-edge interdisciplinary discipline. This challenging research field has been ranked first in the world's top ten scientific and technological problems in the 21st century. However, judging from the existing achievements, the quality of the machine translation system is still far from the ultimate goal; and the quality of machine translation is the key to the success of the machine translation system. ^[1]

Professor Zhou Haizhong, a Chinese mathematician and linguist, once pointed out in his dissertation "Fifty Years of Machine Translation": To improve the quality of machine translation, the first problem is to solve the problem of the language itself, not the programming; Machine translation system is definitely unable to improve the quality of machine translation. In addition, without human beings understanding how the human brain performs fuzzy recognition and logical judgment of language, it is impossible for machine translation to reach the level of "faithfulness, elegance, and elegance". These are the bottlenecks that restrict the quality of machine translation.

Cognitive science memory

Memory is the reflection of the human brain on things that have happened in the past, and it is the maintenance of newly acquired behaviors. Because of memory, one can maintain past reflections, make current reflections based on previous reflections, and make reflections more comprehensive and deeper. That is, with memory, people can accumulate experience and expand experience.

There are three types of human memory: sensory memory, short-term memory, and long-term memory. After the stimulus stops, its effects do not disappear immediately, and afterimages can form. Visual afterimages are most noticeable. The afterimage can be said to be the most direct and original memory. The afterimage can only exist for a short time. For example, the most vivid visual afterimage does not last for dozens of seconds. This is sensory memory. The time interval of short-term memory of 100 million yuan is longer than that of sensory memory. However, the storage time is only about a minute, or even shorter. Long-term memory refers to information storage with a holding time of more than one minute. Human memory can be divided into process memory and propositional memory. Process memory is a skill for maintaining related operations, which is mainly composed of perceptual motor skills and cognitive skills. Propositional memory stores knowledge represented by symbols and reflects the essence of things. Propositional memory is further divided into episodic memory and semantic memory. The former is a form of memory that stores personal events and experiences. The latter is to store knowledge of the nature of events that individuals understand, that is, to remember knowledge about the world.

In 1974, Baddeley and Hitch proposed a three-system concept of working memory based on experiments to simulate short-term memory impairments, replacing the original concept of "short-term memory" with "working memory". Baddeley believes that working memory refers to a system that provides temporary storage space and information necessary for processing for complex tasks such as speech understanding, learning, and inference. Working memory systems can store and process information at the same time. The concept of time memory only emphasizes that the storage function is different. Working memory is divided into three sub-components, namely the central execution system, the visual-spatial preliminary processing system, and the voice loop. A large number of behavioral studies and a lot of evidence in neuropsychology have shown the existence of three subcomponents, and the understanding of the structure and action forms of working memory is constantly enriched and improved. People find that working memory is closely related to language comprehension ability, attention and reasoning, etc. Working memory contains the secret of intelligence.

Cognitive science attention

Since the mid-1950s, with the rise of cognitive psychology, people have re-recognized the importance of attention in information processing in the human brain, and proposed several attention models. Among them are the attention filtering model and attenuation model, which belong to the perceptual selection model. These two models locate the attention mechanism at the perceptual stage of information processing and realize information selection before recognition. In contrast to the perceptual selection model is the response selection model, which believes that the role of attention is not to select the stimulus, but to choose the response to the stimulus. The model believes that all information can enter the advanced processing stage, but only the most important information will cause the response of the central system. The two types of models have different emphasis. The perceptual selection model emphasizes concentrated attention, while the response selection model focuses on distributed attention. The focus of the two debates is the position of attention mechanism in information processing. The central energy model of attention is generated in this context. The theoretical basis of the model is the limited processing capacity of the information system. It avoids the problem of the position of attention mechanism in information processing, and makes the experimental results of perceptual selection model and response selection model formally unified; however, it does not reveal the information processing process involved in attention.

With the rapid development of brain imaging technology and neurophysiological research, efforts to separate attention networks from other information processing systems have become a reality. The use of positron tomography (PET) and functional magnetic resonance imaging (fMRI) technologies can more accurately measure changes in cerebral blood flow (rCBF) in various regions of the brain when completing specific attention tasks, thereby determining the function of each attention subnetwork Structural and anatomical positioning. In the early 1980s, the feature integration model proposed by Treisman tightly combined the internal processes of attention and perceptual processing, and used the "spotlight" to visually compare the spatial selectivity of attention. According to this model, the visual processing process is divided into two interconnected stages, namely the pre-attention and concentrated attention stages. The former performs fast and automatic parallel processing on simple features such as color, orientation, and movement of visual stimuli, and each feature is separately encoded in the brain to generate a corresponding "feature map". Each feature in the feature map constitutes the appearance of attention. Pre-attention processing is a "bottom-up" information processing process that does not require focused attention. The location of each feature in the feature map is uncertain. To obtain object perception, you need to focus your attention and scan the "location map" through the "spotlight" to integrate the features that belong to the search target organically. Achieve dynamic assembly of features. In 1989, Gray pointed out that concentrated attention can cause the synchronous release of neurons related to the attention event, and the synchronous release usually manifests as a synchronous oscillation of about 40 weeks. This finding provides neurophysiological evidence for a feature integration model of attention.

According to the existing research results, Posner divides the attention network into three subsystems: the front attention system, the rear attention system and the alert system. The anterior attention system mainly involves the frontal cortex, anterior cingulate gyrus, and basal ganglia. The posterior attention system mainly includes the superior parietal cortex, the thalamic occipital nucleus, and the superior colliculus. The alert system mainly involves the input of blue norepinephrine to the cortex located in the right frontal region of the brain. The functions of these three subsystems can be summarized as directional control, guided search and alertness.

Cognitive science awareness

Consciousness is perhaps one of the greatest mysteries and highest achievements of the human brain. Since the establishment of modern psychology in 1879, consciousness has become the main research object of psychology. According to James, psychology is the science of consciousness. However, due to methodological issues, it is impossible to conduct specific scientific research on consciousness. The behaviorist psychology that emerged in the 1920s did not recognize the existence of consciousness. The cognitive psychology that appeared in the 1950s raised the issue of consciousness and studied consciousness from perception and awareness. Great progress has been made in the study of perception, but the research on awareness and other issues is still in its preliminary stage.

It is currently very difficult to give a unified and precise scientific definition of consciousness. Different areas have different understandings of consciousness. Nobel laureate Crick believes that consciousness involves the neural mechanism that combines attention and short-term memory, which can be studied scientifically [4]. Crick's amazing hypothesis about consciousness and through visual attention and short-term memory Specific suggestions for visual awareness have aroused widespread interest among a large number of cognitive psychologists, neuroscientists, and computational neuroscientists.

In the late 1980s and early 1990s, there was a major discovery in the research of visual physiology: the phenomenon of synchronous oscillation was recorded from the release of different neurons. This synchronous oscillation phenomenon of about 40Hz is considered to connect nerves between different image features. Signal. Crick and Koch proposed a model of 40Hz oscillation for visual attention. And it is speculated that the 40Hz synchronous oscillation of neurons may be a form of "bundling" of different features in vision. As for "free will," Crick thinks it has to do with consciousness and involves actions and the execution of plans. Eccles, another Nobel laureate, is passionate about the study of consciousness. In the book "Ego and Brain" co-authored with the philosopher Popper, he published a philosophical view of the "three worlds". It is believed that world 1 includes all material worlds (including the brain), world 2 includes the human spiritual world, and world 3 includes human social, linguistic, scientific, and cultural activities. In his later works, based on the structure and function of the nervous system, he proposed the "dendron" hypothesis. The dendron is the basic structural and functional unit of the nervous system. There are an estimated 400,000 dendrites in the human brain. He went on to propose the "psychon" hypothesis that the mental child of World 2 corresponds to the dendritic child of World 1. Because the microstructure in the dendrites is close to the quantum scale, quantum physics may be used for consciousness issues.

Consciousness is a complex issue, and we should find an entry point and further study it in combination with the currently available technical means. Research consciousness can use awareness and non-awareness as entry points to find the difference between neural related substances in brain activity.

In fact, there is also an emotional system, which has been actively researched in recent years. The immune system and intelligence are also closely related. Due to space limitations, we will not discuss them here.