Aug 26 2009

By Definition

If you are not aware of Massimo Pigliucci’s excellent blog, Rationally Speaking, take a look. He covers the very important intersection of science and philosophy. His most recent post covers the concept of definitions in science and in general. It’s a great discussion that I wanted to expand upon a bit.

Massimo’s central point is that it is often not necessary, or even possible, to have a precise and unambiguous definition in science, and in fact scientific discovery is often exactly about exploring the nature of a thing.

To understand this we need to understand the nature and role of definitions in science and philosophical thought. At one end of the spectrum there is the fallacy that is called the “Socratic fallacy” (although Massimo points out that Socractes was not guilty of this fallacy) – the assumption that definitions need to be pristine to be useful. At the other end is the fallacy that definitions are hopelessly ambiguous and therefore of no value.

The reality is more complex, as it almost always is.

It has become a cliche of naive argumentation that whipping out a dictionary definition is the mark of a scientific rube. Dictionaries document current consensus – not anything useful in terms of scientific categorization. But they represent one type of “definition” – a meaning agreed upon by convention and useful for unambiguous communication. Dictionary definitions oversimplify and are not useful for scientific discussion.

Science simultaneously deals with two types of categorical definitions – operational definitions and conceptual categories. Massimo aptly characterizes the latter as:

…identified by philosopher of language Ludwig Wittgenstein, and known usually as “family resemblance” or “cluster” concept: it does not admit of a simple definition in terms of a small set of necessary and sufficient conditions. Rather, it is fuzzy, made of a number of conceptual strands that intersect in a complex fashion.

Categories in nature are sloppy – because they are more complex than our attempts at pigeonholing. Massimo uses species and planets as his prime examples, and they are good ones. Is Pluto a planet? That depends on how you define planet, and astronomers recently had to face the fact that there is no clear and unambiguous way to do this. There is a spectrum of objects orbiting our sun, with a variety of origins, fates, compositions, sizes, and relationships to other objects. There is no clean line to draw for planet, or dwarf planet, or planetoid, or whatever.

Species are also fuzzy around the edges – a population blends into a sub-species which blends into a species, which may retain tenuous genetic exchange with other “species”.

I would add an example from my own profession – that of the disease diagnosis. Diseases and disorders are often not discrete entities – they overlap with each other, there often are many subtypes, and there are many features which may or may not be present. This leads to an endless “lumper vs splitter” debate – whether you consider every subtype a different disease or just the spectrum of one disease. Diseases are often fuzzy.

Massimo points out that scientists can explore the origin of planets even without a clear definition – and likewise, physicians can treat a category of disease even without a precise diagnosis (as one may not be possible by definition). This often causes patients undue anxiety, as the common assumption is that treatment is dependent upon a precise label.

Massimo did not discuss, however, the role of an operational definition in science (including medicine). An operational definition is a list of criteria – if a patient has symptom A, B, and C but not D or E then they have syndrome X. Or – if a body has enough gravity to pull itself into a sphere, it orbits the sun, and it gravitationally dominates its orbit, it’s a planet. Operational definitions are partly convention – agreed upon nomenclature. But they do allow us to ask certain kinds of questions.

For example, If I want to know if a drug relieves acute migraine attacks, I need to have an operational definition of an acute migraine attack. This definition will necessarily not be perfect, it will oversimplify the complexity, it will exclude some episodes that may reasonably be considered migraines, but it creates a category that can be broadly and practically applied. It allows me to compare my results to a researcher in Japan, if they follow the same operational definition I spell out. (Often, central bodies will determine such operational definitions that everyone will use.)

Operational definitions are useful, often necessary. In some cases they may even exactly correspond to a real conceptual category. At other times they are an acknowledged approximation and oversimplification.

The trick is to use operational definitions, and nomenclature by convention, as they were meant to be used. Meanwhile, understand the nature of categories in science – they are approximations of a very complex reality. They should not constrain our thought, nor substitute for deep understanding. In fact, there is a useful interplay between our understanding of nature and the way that we categorize it. This is a process of creative destruction, not enshrined icons – like science itself.

16 responses so far