PROBLEM-SOLVING
An area of theory which attempts to model directed thinking (thinking which is goal-oriented and rational), often with a view to computer simulation. In cognitive psychology and Artificial Intelligence, the term ‘problem’ has a wide interpretation– referring to a desired state of knowledge towards which behaviour is directed. It thus potentially includes language production and comprehension. Many of the ideas in this field are influenced by studies of expertise (especially that of chess players) and by an early AI computer program, the General Problem Solver (GPS), which modelled rational thought.
There are two important factors in the problem-solving process. The first is a representation of the critical features of a task. Representation includes an initial state, agoal state and possible paths to achieving the goal. A set of operators based on world knowledge determine which paths are permissible and likely to achieve the goal. Together, the states and operators constitute what is known as the problem space.
The second important component is search. Much enquiry in this area has focused on the strategies adopted by individuals in searching a problem space. They tend to adopt two main approaches. An algorithm is a systematic procedure which gives a guaranteed solution. However, it entails evaluating all possible paths that might lead to the goal state, and is thus potentially very time-consuming. By contrast, a heuristic is a set of general guidelines which steer the search process so that a complete search is not necessary. Heuristics do not guarantee a solution but they enable the problem solver to economise on effort. Individual differences in problem-solving are partly attributable to differences in the ability to select and apply appropriate heuristics.
One way of formalising the knowledge that underlies problem solving is in terms of production systems. These are sets of rules (‘productions’) for solving a problem, often expressed in an IF/THEN form. They are used to represent procedural knowledge (e.g. in Anderson’s ACT models of expertise) and have also been used in accounts of syntactic parsing:
Problem-solving theory has implications for other areas of language acquisition and use. For example, it can model the way in which listeners resolve ambiguity in the signal or can provide a framework for analysing the strategies employed by second language learners.
See also: Model
Further reading: Robertson (2001)
