ABSTRACT
Plans are an ubiquitous part of human activity. A plan can be defined as a structured event series that generally contains one or more goals. Plans range from the short term and motoric, such as planning a sequence of key presses (Pascual-Leone et al., 1993) to the long term and cognitive, such as deciding on the steps required for air traffic controllers to land a specific airplane (Suchman, 1987). How plans are developed and executed has been the focus of study in artificial intelligence (AI) (Allen, Kautz, Pelavin, & Tenenberg, 1991; Hammond, 1994), cognitive science (Friedman & Scholnick, 1997; Hoc, 1988), and neuropsychology (Owen, 1997). In their prescient book on planning, Miller, Galanter, and Pribram (1960) revealed the difficulty that neuropsychology might have with identifying which brain structures would be concerned with planning as defined by contemporary computer science terminology, eventually admitting: “The relation between computers and the brain was a battle the authors fought with one another until the exasperation became unbearable.” The responsibility for this difficulty may partly lie in the different methods used to investigate planning by each discipline. Besides their differences, each discipline’s methods has particular weaknesses. For example, Langley and Drummond have recently decried the non-experimental basis of much of the AI literature on planning (Langley & Drummond, 1990). They have argued for the development of testable hypotheses that can be experimentally addressed such as: “What are the resources required to
generate a plan?” “In reacting to an unexpected event, how much sampling of the environment is done?” “What is the ratio of deliberation to execution (and what does that ratio depend upon)?” “How do subjects modify stored plans versus constructing entirely new plans?”
