Artificial Intelligence (AI)

Artificial Intelligence refers to computer programs capable of doing things that require intelligence when done by people (Best, 19860. Artificial intelligence is based on the fact that many tasks --from harmonizing music to medical diagnosis --can be reduced to a set of rules applied to a body of information. AI is valuable in situations were sped, vast memory, and persistence are required. In fact, AI programs are better at some tasks than humans.

Artificial intelligence provides a way to probe some of the oldest questions about the mind, such as how we comprehend language, make decisions, and solve problems. Increasingly, cognitive psychologists are using AI as a research tool in two basic ways:

Computer simulations Programs are used to simulate human behavior, especially problem solving. Here, the computer acts as a "laboratory" for testing models of cognition. If a computer program behaves as humans do (including making the same errors), then the program may be a good model of how we think (Mayer, 1983).

Expert systems. Programs that display advanced knowledge of a specific topic or skill. Expert systems have demystified some areas of human ability by converting complex skills to clearly stated rules that a computer can follow. Expert systems have been created to predict the weather, to analyze geological formations, to diagnose disease, to tell when to buy and sell stocks, to play chess, to read text, to do psychotherapy, and to perform many other tasks.

Experts and Novices. Working with artificial intelligence has helped especially to clarify differences between novices and experts. Research on chess masters, for example, shows that their skills are based on specific organized knowledge and acquired strategies. In other words, becoming a star performer does not come from some general strengthening of the mind. Master chess players don't necessarily have better memories than beginners (except for chess positions). And, typically, they don't explore more moves ahead than lesser players. What does set master players apart is their ability to recognize patterns that suggest what lines of play should be explored next. This helps eliminate a large number of possible moves. The chess master, therefore, does not waste time exploring unproductive pathways. Experts are better able to see the true nature of problems and to define them in terms of general principles (Rabinowitz & Glaser, 1985).

Expertise also allows more automatic processing, or fast, fairly effortless thinking based on experience with similar problems. Automatic processing frees attention and "space" in short-term memory that can be used to work on the problem. At the highest skill levels, expert performers tend to rise above a reliance on rules and plans. Their decisions, thinking, and actions become rapid and fluid. Thus, when a chess master recognizes a pattern on the chessboard, the most desirable tactic comes to mind almost immediately. (Dreyfus & Dreyfus, 1986).

What the preceding tells us is that experts in one area do not automatically become better problem solvers elsewhere. Nor do they become generally smarter. The same conclusion applies to artificial intelligence. Expert systems have been hailed as a possible remedy for human errors in tasks such as air traffic control, the operation of nuclear power plants, and the control of weapons systems. However, the truth is that expert systems are "idiot geniuses." They are very adept within a narrow range of problem solving, but they are "stone stupid" at everything else. Eventually, I may lead to robots that recognize voices and that speak and act "intelligently" (Best, 1986). But cognitive scientists are becoming aware that machine "intelligence" is ultimately "blind" outside its underlying set of rules. In contrast, human cognition is much more flexible. For example, u can understand words thet ar mizpeld. Computers are very literal and easily stymied by such errors. Humans are able to take into account exceptions, context, and interpretations as they think. We also make commitments and take responsibility for our actions. A rule-driven expert system processes information without regard for the meaning of actions. Expert systems may never be able to anticipate the infinite number of possible events that could occur. As a result, their actions might be disastrous in unanticipated situations.