Means-ends analysis (also known as "means/end analysis" or "MEA") is a cornerstone component of theories of human problem solving. It is a strategy in which a person identifies and then eliminates any differences between their current position in a problem and the goal state. Such theories of problem solving fall under the domains of cognitive psychology and artificial intelligence.
In cognitive science
Not all problems involve means-ends analysis. Yet such problems are common enough that basic theories like Allen Newell and Herbert Simon's influential "General Problem Solver" program of 1963 use means-ends analysis as the main heuristic in modeling human problem solving. James Greeno proposed three general categories of problems, while acknowledging that many other problems would fall outside of the typology or be described as a combination of problem types. His classification includes 1) arrangement problems (e.g. anagrams) that require a set of objects to be rearranged in a particular way; 2) inducing structure problems (e.g. analogies) that are solved once the relation between objects is discovered; and 3) transformation problems, which require a sequence of operations to be performed on an initial state to reach a given goal state. Of these three problem types, transformation problems clearly lend themselves best to means-ends analysis, which requires both a changing problem state and a goal state to compare. Other examples of transformation problems include the water jug problems, the Tower of Hanoi problem, and the proof of theorems.
It is generally assumed that the means-ends strategy proceeds by recursively minimizing the difference between the current problem state and the goal state, rather than maximizing the difference between the problem state and the initial state. Studies of how people solve the water jug problem or the Tower of Hanoi problem have not distinguished between these two possibilities. Yet the latter possibility is plausible given another problem solving heuristic that people demonstrate, namely an avoidance of problem states already tried.
In artificial intelligence
Means-ends analysis has been used in artificial intelligence, especially machine learning, and indeed may be better suited to such applications than as a model of human problem solving. This is because the cognitive load of many problems makes their solution by means-ends analysis impractical for human problem solvers. In machine learning, means-ends analysis requires identification of all possible changes from the current problem state, and for each such move, requires computation of the resulting change in proximity to the goal state. Memory of all these computations is also required in order for an optimal next step to be made at each problem state.
- Newell, A., & Simon, H.A. (1972). Human problem solving. Englewood Cliffs, New Jersey: Prentice-Hall.
- Greeno, J.G. (1978). Natures of problem solving abilities. In W.K. Estes (Ed.), Handbook of learning and cognitive processes (Vol. 5). Hillsdale, New Jersey: Erlbaum.
- Atwood, M.E., & Polson, P.G. (1976). A process model for water jug problems. Cognitive Psychology 8, 196-216.
- Egan, D.E., & Greeno, J.G. (1974). Theory of rule induction: Knowledge acquired in concept learning, serial pattern learning, and problem solving. In L. Gregg (Ed.), Knowledge and cognition. Hillsdale, New Jersey: Erlbaum.
- Reed, S.K. (1996). Cognition: Theory and Applications (4th ed.). Toronto: Brooks/Cole.