Information processing (IP) uses the computer as a model of human learning. In the past the model has been a block of wax, a telephone switchboard, a radiator, and the engine, reflecting the technological advances of each era. The computer metaphor likens the brain or mind to the computer as a way to store and retrieve information. The basic theory used by computer scientists and the neurosciences is information processing, assuming that the brain and the computer have similar or identical faculties. The two general approaches to modeling human intelligence are artificial intelligence and connectionism.
Searle argues that computer programs are only syntactic, not semantic, and that brains create minds, not computers. A system that can cause minds must also have causal powers similar to a brain, which a computer does not. Therefore, before computers can be regarded as intelligent, they must do more than follow instructions. They must really think.
Descartes said, "I think, therefore I am." Although conscious awareness (sentience) has been accepted as proof of existence, a basic question is, What is thinking? Another basic question is, How do people think? What actually occurs in the synapses to enable thinking? Descartes did not say how people think, although he thought about it. He viewed the mind as a machine and thought the nerves were like the plumbing of fountains.
AI programmers develop serial computer programs that imitate human problem solving by first studying the performance of human experts. Then they develop long, complex linear programs, under executive control, to solve the problems like humans, such as playing chess or diagnosing diseases. By contrast, connectionism regards the brain to work in a parallel, non-linear, manner (parallel distributed processing or PDP) under the control of a central executive (to pay attention and monitor all the goings on). Although connectionism rejects serial processing, it really does not develop parallel programs because it just uses more computers to engage in simultaneous serial programs, which combine the results at the end rather than as a simultaneous interactive process. Many connectionists claim that their computer programs really think.
Differences Among Species
Scientists have traditionally focused on differences between humans and other animals and ignored their similarities, particularly intelligent behavior, tool use, and communication. As a product of mammalian phylogenetic development, the human brain contains ancient vestiges in common with other species from millions of years of evolution. Biological principles are surprisingly similar among all species, and visual, respiratory, locomotion systems are nearly identical across all mammalian species. Humans and chimpanzees share 98% of the same DNA. Recently, evidence that some animals use tools in the wild, and primates can learn language through signing, narrows the distinctions. But while most scientists have reluctantly admitted to similarities between humans and animals, others have eagerly accepted claims that computer programs can think just like humans. The essence of IP to these scientists (AI, connectionist, or PDP) is that cognition really exists in computer programs.
Computers as Minds
The assumption for believing computers create human minds is often attributed to the "Turing Test" proposed by Alan M. Turing. A founder of artificial intelligence, Turing asserted that if a machine can perform in a manner indistinguishable from an expert on some cognitive task, the machine can think. Turing made his name as a code breaker in Britain during World War II by successfully directing his team to break the German "Enigma" code. During this process he imagined the system of information processing that became fundamental to development of computers. The Turing Test is accepted by some today who believe that a computer can mimic the brain. The recent event of a computer defeating an international chess champion seemed to satisfy the question to some. Artificial intelligence was started by the Turing proposal in a paper that described thinking in computational terms. Psychologists and educators are more interested in this concept than AI workers, because AI is almost exclusively concerned with developing algorithms to solve real-world problems, problems that are almost always beyond the capacity of human brains. So AI research is interesting but it really has little to do with how the human brain functions or about intelligence.
There are many individuals who can be said to have contributed to information processing theory, but the basic human information processing (IP) model has one input modality (attention) and three memory stores (primary, secondary, and tertiary). Incoming information must be attended, it must be stored in a buffer, rehearsed, and transmitted to secondary and tertiary or long-term memory (LTM). Nearly all research in cognitive psychology has concerned understanding memory storage using this model or variations of it (Atkinson & Shiffrin, 1968).
Information processing views cognition as a template-matching process where the input (image, sound) is matched against the most similar template stored away in long-term memory. This may be true for certain kinds of learning and recall, but there may be different kinds of processing. If a linear process fails on a computer, it simply crashes. Although humans may use linear processes in some kinds of thinking, there is no equivalent of "crashing" in human thought. The brain does not simply stop but attempts to find what went wrong by starting over and taking a different approach. The brain also leaps to conclusions, with incomplete information or no information at all. The computer cannot account for this "fuzzy" kind of thinking, as when conclusions are drawn from perceptions, "feelings," intuition, and prejudice.
When memory is investigated, subjects are presented with information to be learned and are later tested to see how much and how long information is remembered. Within even a simple study, many variables can be considered (attention, discrimination, recognition, short-term recall, interference, forgetting, and so forth), becoming complex. The most obvious aspect of this model is that it consists entirely of explicit memory structures and processes, as if the brain were designed to store and recall information.
The human brain fails miserably at its most fundamental task, short term memory (STM). If information is not rehearsed or written down, it is lost (forgotten?) within a minute. The average STM or working memory is about 7 bits of information, with a standard deviation of 2. Long-Term Memory (LTM) is presumed to be stored in the brain without degradation for many years. Unlike STM, there appears to be no capacity limitation for LTM. Elderly people, who may no longer recognize their immediate surroundings, are able to recall facts and concepts from childhood with ease.
Other aspects of learning research have concentrated on such factors as massed practice, spaced practice, and over learning, as well as the efficiency of learning at different times in a sequence. Organisms (people, animals) tend to learn more in the beginning and later. Retention is better if learning is spread out over time rather than crammed (massed), and practice increases accuracy. Uses of self-generated methods, such as stories and mnemonics, increases efficiency. Awareness of one's own learning processes is referred to as metacognition.
According to Edelman, one of the outstanding challenges to any general brain theory is to explain how the brain is able to store information. In other words, what is the physical basis for such storage? How can a common network produce short-term changes affecting perceptual categorization and learning while also amassing long-term changes? How can the network store and distribute both kinds of information that are related both to input and output requirements? How can memory last a lifetime?
Most cognitive psychologists accept the premise that information is stored in the brain. Although there is some degree of disagreement over certain aspects of this model, the physical bases of memory are not questioned by most scientists. A central assumption is that perceptions and our experiences produce changes in the brain that represent these events. In other words, there must be some trace of the past experience that is stored in a permanent form, like a tape recorder or floppy disk. But according to Edelman, recent research has shown that memory cannot be stored in molecular patterns because protein turnover is rapid, fixed unchanging structure is rare in biological organisms, and there is no evidence to suggest that polynucleotide structures that can be repaired faithfully have any association with memory storage.
Noting that the time constraints of synaptic change do not necessarily relate directly to those of memory, Edleman maintains it is physically impossible for the same cellular tissue to store a trace. He concludes that there is no direct relationship between a physical event, its effect on synapses, and its ability to later be recalled. Maintaining that memory is, instead, a "form of recategorization based on global mappings," he suggests that long-term synaptic changes do not necessarily correspond to long-term memory. Edleman proposes switching the trace from long-term synaptic storage to categorical memory based on "less long lasting synaptic changes" in the network. Edelman rejects information processing, saying it is not a brain state. He explains brain functions as a selection process in evolutionary terms, where successful neuronal groups that are affected in responses to interactions with the real world are formed. Such a theory can explain both individual and group differences, because people's brains are shaped by their environments.
Research with humans has concentrated on the capacity limitations of the three stores or how long and how much information they can hold. Otherwise, a significant amount of research with computer programs has been conducted under the rubric of parallel distributed processing (PDP).
The approach of PDP is to imitate human cognitive processes as neurons in software. The belief is that
the software creates a human-like "excitatory network" that can discriminate and learn. This is accepted by theorists as knowledge representation, in that knowledge is stored as connections in the network By adjusting the strengths of connections, the network is said to learn.
The IP theory assumes that knowledge consists of meaningful (monolithic) units or chunks, whose organization in LTM scientists must discover. Connectionism assumes that knowledge is stored in connection weight patterns among distributed meaningless (subsymbolic) units. Most researchers who study a computer model of a hurricane do not believe it is actually wet inside the computer, but perhaps some do.
In the IP view, children are poor learners who improve as they grow older, and poor learners have structural deficits that interfere with the capacity to exercise executive control over short-term memory processes. But unlike the PDP approach, we are unable to "adjust the connection weights" of humans to correct for their problems, although some are suggesting that medications in the future may help. However, if medicine does something to the CNS to improve memory, is it the same thing as adjusting connection weights?
In the biofunctional view of learning some have denied that the brain's functional processes are explicit memory processes (Iran-Nejad, Marsh, & Clements, 1992), as proposed by long-term storage metaphors or by network metaphors (Rumelhart, Smolensky, McClelland, & Hinton, 1986). A distinction can be made between the brain and the mind, because mindful processes are not brain processes but their products, influenced by context and individual factors. The literature on unconscious information processing shows that information that enters the sensory store can be absorbed by the brain's learning processes without apparent awareness or attention (Marcel, 1983; Nisbett & Wilson, 1977), an idea originally proposed by Bartlett (1932). Explicit memory structures and processes of the IP model do not include context in the research. These functions are apparently impossible for a computer. If the computer chip is a brain, where is the mind? In the software?
Perhaps emotion should also be considered, for this is a significant part of human behavior. Emotion, intention, and motivation are aspects that are beyond computerization at the present time. Can a machine develop self-awareness? Consciousness? If so this may be the ultimate test of intelligence and of existence. But in doing so we have to disregard the physical properties--biological, real-world factors of living, thinking organisms. Perceptual limitations of the computer and weakness in non-linear tasks are the strengths of the human brain. The computer is a million times faster than a neuron, but the brain can solve perceptual problems much faster. While computers with greater power and speed are getting better at visual recognition, the equivalent power in a human brain would make the head too heavy to carry around on our shoulders. The reason is that scientists who create smart robots or computers have to build all the intelligence into the "executive" part of the system because there is no self organization. Building "intelligence" into computers and robots is difficult, something nature provides brains through the process of evolution; something that is distributed in sub-systems.
Information processing views cognition as a template-matching process where the input object is matched against the most similar template stored away in long-term memory. This may be true for certain kinds of learning and recall, but there may be different kinds of processing. In a linear process, as any computer user knows, a linear program will fail with one faulty routine. It simply crashes. Although humans use linear processes in thinking, there is no equivalent of "crashing" in human thought, a matter discussed in the section on mental models. The main point may be that humans routinely engage in many different kinds of thinking other than complex problem solving.
The Brain as a Biological Organ
The brain develops hemispheric asymmetry or lateral specialization, with linguistic functions typically controlled in the left hemisphere. It is from here that AI researchers originally attempted to imitate thinking, using verbal labels as in human speech. But persons with normal brains who have had the corpus callosum surgically severed (split-brains) still demonstrate all the functions of language, spatial orientation, and so forth. There is interhemispheric communication outside the body by means of cues and shaking the head. Certainly, although split-brained persons do not perform optimally, they apparently demonstrate that other systems of thinking occur in the brain, not just linguistic functioning of the left hemisphere. In fact, they seem to have two brains and, some might argue, two minds or one that knows it is fragmented and needs to cooperate with its severed self for some needs.
At the subcortical level, automatic body movements work, as a product of evolution, without direct executive control. In a cortical hierarchical view of the central nervous system, processes lower in the system, toward or at the level of the old brain, function autonomously and require no conscious attention, such as digestion, heartbeat, and respiration. Higher-order functioning vertical processes, or those at the same level, are apparently used together in some combinations to solve certain kinds of problems, but they probably work independently of one another until cooperation is necessary. They function together without executive control and in a systematic, self-organizing way.
In split-brain research, one hemisphere can be taught something or required to perform some task without the other side being aware of it. The "uninformed" side will later show surprise and denial when confronted with evidence that something was done. This argues against theories of excitatory and inhibitory states of the mind and supports the theory that independent subsystems exist in the brain. There are possibly many aspects of thinking, such as a finite quantity of independent, autonomous subsystems and numerous microsystems, creating a variety of ways that thinking and problem solving occur, both consciously and unconsciously.
Rather than simple formulas or wired neural sets, there may be qualitatively different kinds of thinking expected on the basis of shared attributes of mammalian brains. If the apex of human evolutionary development is abstract thought served by language, can it be denied that other kinds of thinking also exist in the brains of humans, as they most certainly do in animals. In some problems humans probably also think like animals.
Theoretically there are three kinds of intelligence used in the evolutionary process: dynamic, emergent, and intentional. Much research on information processing has focused on intentional learning. The main hypothesis is the longer information stays in short-term memory, the better it is remembered. How long, how much, and how well depends strictly on the work of the executive control process, or on active rehearsal of material to be memorized. Although this may be true for intentional learning, it may not be the way the brain actually functions, and, if it is, it is probably not the only way the brain processes information, especially given the close relationship to other mammalian brains. While we cannot know for certain, it seems unlikely that raccoons make mental lists or rehearse.As new information enters the senses there are reactions to very specific inputs. For most information, at least the majority that is not directly regarded by the executive, the process might be like discriminating or sorting the information into categories on the bases of gross discrimination, then to more refined classifications within categories. Apparently this level of discrimination is through a contentious process within microsystems that cease when agreement is reached about the nature of incoming stimuli. While it may be "misclassified" according to another brain, in each particular brain there are unique constructions that impose order on stimuli by sorting them according to unique, classifiable characteristics. New information may attract attention of the executive, as when there is some assimmilation-accommodation problem to solve. A "rookie" may be perplexed and not know what to do with it but an expert quickly identifies the problem for what it is, readily classifies it, and generates obvious solutions. This might be explained as the differences in richness of theme structures.
Each microsystem can act individually to compete for control of attention. Things that cause discomfort or peril get quick attention, such as pain. Attention is limited so the competition may granted to only a few at any given moment, and this may even be an internal state, such as daydreaming. At this point all other competing attractions are ignored, but the microsystems of the brain may continually deal with processing that is subsumed at an unconscious level. Hierarchically, each level of processing may function without conscious attention to the level beneath it or to other processes occurring at the same level. Traditional approaches to intelligence perceive complex intentional symbolic interactions under active executive control, and this is obviously not always the way the brain works.
By distributing intelligence as dynamic, emergent, intentional, the brain is capable of performing higher-level functions because it is not necessary or economical to use the central executive to pay attention to everything all the time and deliberately operate all subsystems, integrate them and be cognizant of the status of each one. In deed, the hormonal system, immune system, and many other systems of the body, many of them connect directly and indirectly with the central nervous system, function autonomously or semi-autonomously. Without significant evidence to the contrary, there is no reason to believe that this type of process simply stops at the level of the central nervous system. This is, at least, an alternative to the IP model of brain functioning.Learning Styles and Information Processing
In education the research into learning styles has been concerned with cognition, information processing, and affective (social, emotional, and cultural) influences on learning, motivation, and achievement. Various theories various and models have evolved to consider learning styles including locus of control, concrete versus abstract reasoning, reflective versus active engagement, domain specific preferences, modalities (auditory, visual, kinesthetic), cerebral processing (right versus left brain), metacognition, global/analytical impulsive/reflective field dependence/field independence simultaneous/sequential processing physical, environmental. The term "learning styles" has different meanings for people; for some, it is synonymous with "cognitive styles," and for others it refers to preferred approaches to learning based on modality strengths, and others believe it means hemispheric functioning, i.e., whether one is right-brained or left-brained.A number of classroom strategies have been used to match learning style with instructional style, including small-group activities, cooperative learning, mastery learning, performance-based assessment, and computers. Thus, there are many concepts and terms such as Witkind's field dependence/field independence, Kirby's simultaneous/sequential processing, innovative learners, analytic learners, dynamic learners, and many others.
The Myers-Briggs Type Indicator (MBTI) uses preferences based on Carl Jung's theory of psychological types to form 16 different learning style types.
Kolb's Learning Style Model presumes that learners have a preference that can be classifed as follows:
- extraverts
- sensors
- thinkers
- judgers
Herrmann Brain Dominance Instrument (HBDI) classifies learners this way:
- concrete, reflective
- abstract, reflective
- abstract, active
- concrete, active
The Felder-Silverman Learning Style Model uses these categories:
- Quadrant A (left brain, cerebral). Logical, analytical, quantitative, factual, critical;
- Quadrant B (left brain, limbic). Sequential, organized, planned, detailed, structured;
- Quadrant C (right brain, limbic). Emotional, interpersonal, sensory, kinesthetic, symbolic;
- Quadrant D (right brain, cerebral). Visual, holistic, innovative.
Curtis Carver and Richard Howard summarize the various approaches succinctly:
- sensing learners or intuitive learners
- visual learners or verbal learners
- inductive learners or deductive learners
- active learners or reflective learners
- sequential learners or global learners
Students have different learning styles--characteristic strengths and preferences in the ways they take in and process information. Some students tend to focus on facts, data, and algorithms; others are more comfortable with theories and mathematical models. Some respond strongly to visual forms of information, like pictures, diagrams, and schematics; others get more from verbal forms--written and spoken explanations. Some prefer to learn actively and interactively; others function more introspectively and individually.Curtis Carver and Richard Howard have experimented with hypermedia to teach with different presentations of material to appeal to different learning styles and make the following recommendations:These suggestions provide fertile ground for future research with multimedia.
- Teach theoretical material by first presenting phenomena and problems that relate to the theory (sensing, inductive, global).
- Balance conceptual information (intuitive) with concrete information (sensing). Intuitors favor conceptual information--theories, mathematical models, and material that emphasizes fundamental understanding.
- Make extensive use of sketches, plots, schematics, vector diagrams, computer graphics, and physical demonstrations (visual) in addition to oral and written explanations and derivations (verbal) in lectures and readings.
- To illustrate an abstract concept or problem-solving algorithm, use at least one numerical example (sensing) to supplement the usual algebraic example (intuitive). For example, when presenting Euler's method for numerical integration, instead of simply giving the formulas for successive steps, use the algorithm to integrate a simple function like y = x2 and work out the first few steps on the chalkboard with a hand calculator.
- Use physical analogies and demonstrations to illustrate the magnitudes of calculated quantities (sensing, global).
- Occasionally give some experimental observations before presenting the general principle, and have the students (preferably working in groups) see how far they can get toward inferring the latter (inductive).
- Provide class time for students to think about the material being presented (reflective) and for active student participation (active).
- Demonstrate the logical flow of individual course topics (sequential), but also point out connections between the current material and other relevant material in the same course.
While most educators accept the premise that there are learning styles, there really is limited research in this area to support many of the basic assumptions. While there has been considerable attention to cognition or information processing, the basis for learning styles may not be so much a matter of brain wiring as it is the cumulative effects of cultural values, socialization, and peer influence. While many educators tend to believe that family values influence learning, and there has been considerable evidence to demonstrate that SES has a higher association with achievement than school variables, cultural differences in learning style preferences develop through children's early learning experiences" (Cox & Ramirez, 1981, p. 63), and remarks have been elsewhere about the importance of each particular peer group in shaping attitudes toward schooling.
Students have different learning styles, perhaps, but the strengths and preferences for how information is processed may be more cultural or learned than wired in. It should be noted that gifted students typically have no preferred style and use all forms of input (auditory, visual, and so forth), so personality and social variables may be more important in learning than just a particular method of presentation. As noted throughout, the brain deals with whole information and works best in authentic contexts. The fact that some students have problem learning in school environments may have more to do with the nature of instruction than with the method of processing information, and information processing may be a fall back position of each student in attempts to make sense of instruction. As students become more proficient their learning styles may change.
Bartlett, S. F. (1932). Remembering: A study in experimental and social psychology. Cambridge, England: CambridgeCox, B., & Ramirez, M. (1981). Cognitive styles: Implications for multiethnic education. In J. Banks (Ed.), Education in the 80s: Multiethnic Education (pp. 61-71). Washington, DC: National Education Association.
Iran-Nejad, A., & Ortony, A. (1984). A biofunctional model of distributed mental content, mental structures, awareness, and attention. The Journal of Mind and Behavior, 5, 173-210.
.
Iran-Nejad, A., & Chissom, B. (1988, August). Active and dynamic sources of self-regulation in learning. Paper presented at the annual meeting of the American Psychological Association.
Iran-Nejad, A. (1989). A nonconnectionist schema theory of understanding surprise-ending stories. Discourse Processes, 12, 127-148.
Iran-Nejad, A., McKeachie, W. J., & Berliner, D. C. (1990). The multisource nature of learning: An introduction. Review of Educational Research, 60, 509-515.
Marsh, G.E., II. & Iran-Nejad, A. (1992). Intelligence: Beyond a monolithic concept. Psychonomic Journal, July.