In their famous 1976 paper, Marr and Poggio start out by saying:

Complex systems, like a nervous system or a developing embryo, must be analyzed and understood at several different levels. Of course, there are logical and causal relationships among them, but the important point is that these levels of description are only loosely related. The underlying philosophical issue here is that reductionism does not imply constructionism.

(the levels they mean here are Marr's levels of description of a neural system, computational, algorithmic, etc.)

What does the last sentence mean? I know what reductionism is, of course, but have never heard of constructionism. A brief search seems to point at something used in the context of education, a concept similar to experiential learning. Is this related?

Please note that I'm not an expert in philosophy, so I would appreciate a non-technical answer.

  • I do not think they use "constructionism" as a term, it is rather the plain meaning that reducing higher level to base level means less than constructing higher level notions out of base notions, they are "only loosely related". There is no "constructing" thermodynamic notions from mechanical ones, for example, mechanics does not determine that higher level description must be in those terms we use. Once those notions are introduced from some empirical considerations then we can relate them to base mechanical ones. But there is an element of pragmatism and convention to their choice too.
    – Conifold
    Jan 8 at 18:58

Think of "reductionism does not imply constructionism" as a rough corollary to "the whole is greater than the sum of the parts." Reductionism is often used (usefully) to identify necessary components of some high-level object or event. But identifying necessary components of a high-level object or event does not imply that these components are sufficient for that high-level object or event to exist.

To use the typical analogy, I can easily identify all of the necessary components of an automobile engine. But if I collect all those necessary components and lay them out on a table, they are not in themselves sufficient to being an automobile engine. They must be combined and arranged in certain specific ways, with certain standards and tolerances, and only then can we say we have an automobile engine. In the context of the paper, we can look at the nervous system of a developing embryo at any reductionistic level we like, but nothing we see there is going to tell us whether that embryo will develop into a banker, a politician, an artist, a murderer, a know-nothing nobody, a saint... What that embryo's nervous system will become is more than the mere chemicals and neurons we can examine.

  • I think I get it. Basically, the fact that something is just built upon the laws of physics, atoms, etc. (reductionism) does not imply that we would know how to build it or infer it from scratch using only those ingredients.
    – Martino
    Jan 9 at 12:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.