A. Newell & H. Simon, The Theory of Human Problem Solving; reprinted in Collins & Smith (eds.), Readings in Cognitive Science, section 1.3.

Author of the summary: Patrawadee Prasangsit, 1999, pp@cc.gatech.edu

Cite this paper for:

• For the purpose of problem solving, humans are representable as information processing systems (IPS)
• External Memory - Most cognitive science models have no EM.

Summary

IPS consists of an active processor, input (sensory) and output (motor) systems, and internal LTM and STM, and EM.

How much processing and IPS can accomplish per unit of time depends on three parameters.
  1. The number of processes it can do simultaneously
  2. The time it takes to do each process
  3. The amount of work done by each individual process

Human IPS is a serial system, meaning that it can execute one elementary information process at a time.

Long-Term Memory (LTM)

• Potentially infinite capacity.
• Information is stored in LTM using symbols and relations among them.
• Certain stimuli or patterns of stimuli become recognizable through learning.  These are called chunks.
• Retrieval (read) time is milliseconds to one second (or not much more).
• Storing (write) time is much higher (in the neighborhood of 5-10 seconds per chunk).

• Very small capacity, about 5-7 symbols.
• Immediately and completely available to the IPS processes.
• All processes take their inputs from and leave their outputs in STM.
• Information in STM decays.  Rehearsal processes are necessary to maintain information in STM.

• The immediately available visual field.
• Infinite capacity, accessed by means ranging from linear scanning to random accessing.
• Its existence makes much difference.
• As an example, adding two 10-digit numbers in head versus on paper.

Problem solving in human is goal-directed.  It takes place by search in a problem space i.e., by considering one knowledge state after another until a desired knowledge state is reached.  The search may involve backup (return to old knowledge states).

Other notes:

• This paper was very original at the time it came out.
• Presents a basic architecture of problem solving models.
• A good reading before SOAR.

Summary author's notes:

• none