CGSC 270 Introduction to Cognitive Science
Summary of discussion of explanation in cognitive science
Summary prepared by William Frawley


We discussed the criteria for calling something intelligent or attributing it a mind. We did this through an exercise (thought experiment) in which we considered what kind of "behavior" we would expect a car radio to exhibit, given the knobs, dials, etc. that indicate the radio's "internal state" and given our broader knowledge of what radios generally do. For example, would it "do" anything at all? Does it play music or (just) receive certain frequencies?

We focused principally on intelligence because intelligence and mind are not necessarily equivalent. Among the criteria considered were:

1. ability to learn/change
2. awareness of own states as meaningful
3. ability to engage in complex operations

We tried to develop a more precise vocabulary when proposing such criteria. For example, we compared human learning with the kind of updating that a computer program might do. We also considered change at various levels: whether change in the genetic code is relevant here. We saw that arguments could be given for applying these criteria both to things that seem readily to exhibit intelligence and things that do not. This underscored the need for a precise explanatory vocabulary.

Using the first chapter of Green as a basis, we discussed three metatheoretical strategies: scientific explanation, level of explanation, and method. The details of these are outlined below:

A. Scientific Explanation
precise, rigorous theory-laden vocabulary
predictions
causes (what plays a causal role in behavior?)
B. Level of Explanation
1. Ecological, behavioral, or task level: e.g., what is the entity doing externally?

2. Internal process: e.g., how does the entity produce this behavior?
mental, cognitive, intentional
computational, algorithmic
NOTE: the mental does not have to be computational (Cf. Marr): the computational is one way of making the processes precise. But cognitive science assumes the equivalence of mind and computation.
3. Hardware: e.g., what kind of machine/brain could implement this process leading to this behavior?
Neurological account (wetware)
Machine implementation
Sub-levels: DNA ---> Neuro [PROBLEM: how do non-intentional things, like neurons, become intentional, like thoughts?]
NOTE: we agreed that the most satisfactory account would correlate all three levels, though in practice theorists often choose only one: behaviorism and the ecological, neural reductionism and the hardware.
C. Method
experimentation (theory-laden manipulation)
intuition (self)
observation (non-self: direct/indirect)
breakdown
deduction from models
simulation
PROBLEM: even if you had a complete, rigorous, predictive account of an entity's intelligence at all three levels, would you still have accounted for mind?