Marr, D. (1971). The Philosophy and the Approach. In Vision. Retrieved from http://web.stanford.edu/class/psych209a/ReadingsByDate/01_07/Marr82Philosophy.pdf

Summary

Three levels at which we can analyze cognitive systems:

1. Implementation: how the system is physically realized
2. Algorithmic/Representation: what representation the system uses for the input and output, and what the algorithm is that transforms from input to output
3. Computational: what the goal of the system is (what is being computed, and why) and how that goal can be achieved in a more abstract sense

Example 1: understanding bird flight by studying feathers

Example 2: understanding why retinal ganglion cells and lateral geniculate neurons have the receptive fields that they do by studying their physiology

Distinction between what and how. Just because we do not know how to compute something does not mean that analyzing what is being computed is invalid.

Methods

n/a

Algorithm

n/a

Takeaways

Systems can be analyzed at different levels of abstraction, each of which constrains the other (though according to Marr, only loosely). Focusing on the computational level is important because it allows us to approach understanding the mind from a purposeful stance: what problems does the mind face, and how should those problems be solved?