You can't play 20 questions with nature and win

Why did Newell and Simon develop their production system models? I have only looked briefly at the history, but here are some of the influences other authors have noted.

• The idea of PSs already existed, and was brought to Simon. Crevier p 150.

• They were consistent with the rise of cognitivism (against behaviourism); the view that psychologists should be concerned with internal representations of the world. See Computers and Thought, by M Sharples, D Hogg, C Hutchison, S Torrance, and D Young (MIT 1989; PSY KH:S531; RSL M92.E00068; Hooke), page 200.

• They are a natural development in the authors' views on problem-solving, first expressed in Elements of a theory of human problem solving by Newell, Simon and Shaw, in Psych Rev volume 65 (1958) pp 151-.

• Newell and Simon were interested not only in how data is represented, but also in control - how we move between different stages of processing. See page 89 of the OU book. There are some situations where means-ends analysis (GPS) appears to be sufficient - if there are clear goals and subgoals. But other problems aren't amenable to means-ends analysis; the subject may just plod on without establishing a clear sequence of subgoals. The idea is that the subject has acquired a certain set of operators (rules), and we're interested in the circumstances under which each is activated.

  Note that control can be very flexible, because the symbols in a production system's memory can be used to represent goals and subgoals. Bagger in Winston's Artificial Intelligence is a simple example of this. Bagger works through four stages, using a fact such as ``Step=check-order'' to keep track of which stage it's on. You could think of this fact as meaning ``My goal is to finish stage check-order''.

  Young's seriation example makes this more obvious. He has rules with conclusions like Push(goal=add first block) which adds another goal; conditions like goal = add first block and holding block in hand which depends on it, and rules with conclusions like pop goal-stack which deletes it.

  Contrast with GPS, where goals were added and deleted in a strictly first-on, last-off fashion. They were pushed onto a stack, and popped off when finished.

• There's an early article by Newell, detailing some of the flaws he finds in experimental psychology, and what production systems have to offer. This appears in You can't play 20 questions with nature and win, in Visual Information Processing, edited by Chase (PSY BC:C 39).