Emergence and Levels of Description

This article outlines some standard uses of the word ’emergence’, notes the differences between their respective ontological and epistemological commitments, and comments briefly on the roles which may be played by complementary descriptions of the same item.

Note: more extended comments on both emergence and levels of description appear in the introductory chapter of my book Mind Out of Matter; the chapter can be downloaded free of charge from the page describing the book.

Emergence and Emergentism

Very often within cognitive science, artificial life and related fields, phenomena are referred to as emergent: ‘this behaviour is an emergent property’, ‘these patterns are an emergent phenomenon’, even ‘consciousness is an emergent phenomenon’. Sometimes, an underlying insinuation seems to exist to the effect that everyone really ought to know what the word means and that everyone should just accept that in calling something ’emergent’, we have thereby explained something important.

But the word can take on a wide range of different meanings, and these meanings differ very significantly in their underlying ontological commitments — enough that a use which ‘explains’ for one person merely obfuscates for another. The assertion that ‘x is an emergent property of y‘ might mean any of the following:

  • the appearance of x from y was surprising
  • knowing y, I did not in fact deduce that x would appear
  • knowing only y, I could not possibly have deduced that x would appear
  • knowing only y, deducing that x would appear was not tractable
  • x ‘appears’ when I look at y at a higher level of description
  • laws governing x differ from, and are logically independent of, laws governing y
  • and more…

This last possibility, where the emergence of x from y means that x requires whole new governing laws which do not follow from those of y, I shall unceremoniously banish to the past and thinkers such as Durkheim (1938) and Radcliffe-Brown (1952). This view, shared with the so-called British emergentists (see Alexander 1920, Broad 1925, or Pepper 1926; also see McLaughlin 1992 for recent discussion), has for the most part fallen out of favour in the last few decades. But arguments do continue on questions like, for instance, whether chemical properties bear this sort of relationship to low level physics. In any case, it is important to be aware of this meaning of the word, since it sometimes happens that someone begins stating a position using a weaker version of emergence but finishes with conclusions resting on this far stronger version.

As for the other meanings, all are useful, and it is not my purpose explicitly to defend one against the others. I just want to stress the importance of specifying what exactly is meant when using the word if it is to have any specific meaning whatsoever (rather than a vague cloud of related but distinct meanings). Notice that each meaning above carries with it very specific commitments about the ontological dependence of x on y and about the epistemological features of both x itself and the laws governing x.

In my own work, I personally adopt a convention inspired by the likes of Davidson (1973) and Hellman and Thompson (1975); I opt for what the latter call ‘ontological determination’ — the physical is all there is, and everything that happens physically is wholly governed by the low level laws of physics — coupled with ‘explanatory anti-reductionism’. In other words, nothing ever happens which is not, at the lowest level, entirely due to the laws of physics; yet, in giving intelligible explanations of processes, we may well have to rely on entities constructed at a higher level of description commensurate with that at which we describe the processes themselves

Levels (?) of Description

Closely related to questions about emergence are those about alternative descriptions of the same thing. Usually, people refer to alternative levels of description, suggesting that different kinds of descriptions (for instance, quantum mechanical, chemical, etc.) fit into a neat cascade of explanatory levels emerging from one another and related by supervenience. (See the short introductory notes on supervenience.) But I think this is unfortunate, since more often than not alternative scientific descriptions are actually complementary, ‘horizontally related’, and sometimes largely incommensurable. In other words, alternative descriptions or explanations may simply be different ways of getting at the same thing. (In the guise of a discussion of evolution and Conway’s Game of Life (Poundstone 1985), Dennett (1995c, pp. 166-75) offers the tidiest look at relationships between different ‘levels’ of description I have encountered to date.)

One example of such alternative descriptions with particular relevance for understanding natural cognition and for engineering cognitively sophisticated artificial systems is the difference between computational and dynamical descriptions of cognition. In most cases, I believe both computational and dynamical descriptions have an important role to play. The distinction between them and their particular advantages are presently the subject of heated debates within cognitive science and robotics, but I believe arguing over the relative merits of each is a little like mathematicians battling over reverse Polish notation calculators vs. algebraic calculators while setting the actual mathematics aside. Contra theorists such as Tim Smithers, the well known ‘non-representationalist’ roboticist from the University of the Basque Country, I emphatically do not believe that one should decide in advance to adopt one framework or the other, on pain of tainted science. Whether a good explanation of observed phenomena should be dynamical, computational, or even hybrid in nature is a question properly evaluated in light of empirical evidence and the resulting matches between candidate theories and reality; it is not, to my mind, a matter of evaluating empirical evidence in light of a pre-existing conviction that explanation is to be found in one and only one particular form. (I defend this conciliatory view in an invited commentary for the journal Behavioral and Brain Sciences.)

In any case, some care is due whenever one considers the relationships between alternative descriptions of the same thing; such relationships are usually far from straightforward. I believe a great many confusions in the history of philosophy derive from a failure to appreciate the relationships between different ways of saying the same thing, and the fields of cognitive science and robotics are in danger of repeating these historical mistakes.

This article was originally published by on and was last reviewed or updated by Dr Greg Mulhauser on .

Mulhauser Consulting, Ltd.: Registered in England, no. 4455464. Mulhauser Consulting does not provide investment advice. No warranty or representation, either expressed or implied, is given with respect to the accuracy, completeness, or suitability for purpose of any view or statement expressed on this site.

Copyright © 1999-2020. All Rights Reserved.