Given that an extensional definition is one that defines the definienda by enumerating (partially or exhaustively) examples that belong to it (i.e. the list defines the concept), and intensional is a definition if it focuses on the principles behind the definienda, i.e. necessary and sufficient conditions, and therefire lists things that comply to these conditions as belonging to the given definition (i.e. rules define the concept), then which of these a Turing-test-like definition belongs to?

For those who are not familiar with the Turing test: it is a phenomenological test to overcome an indistinguishability-problem, by passing the decision to an objective judge. In this way it selects those entities that comply to a certain concept (the definition) in an indirect way. To put it simply, if a human judge cannot distinguish a computer from another human by interacting with them then we can take both the computer and the human as "intelligent" entities. Similarly, if a biological cellullar organism does not distinguish between a natural and an artificial cell (e.g. binds to both), than we might want to call both natural and artificial cells "living" (or not, but this is not the point here).

Irrespective of the fact whether such a definition is useful or not, I want to know what kind of definition is a Turing definition. It seems extensional, as on the one hand it practically distinguishes between entities whether they belong or not to the same set. Though on the other hand it relies on implicit (and unknown) principles, conditions, that control whether an entity belongs to the set or not, which might render it intensional. Furthermore, it is not enumerative in any sense, as it does not define the definienda by listing examples.

One might even think it is a third type of definition.


It's closest to an operational definition. The idea of such definitions is to identify empirical criteria that would provide every reason to believe that we are before an instance of F, and propose such criteria as the definition of F. Other operational definitions are the phenetic concept of biological species, or verificationism about meaning.

In general, operational definitions are discredited in philosophy: they stem from very extremely empiricist positions in epistemology that are nowadays generally considered as misguided. Operational definitions hope to find an empirically salient symptom of the definiendum whose presence coincide perfectly with it. But symptoms are hardly ever necessary, or sufficient. In the case that interests you, it appears both that a judge could mistake a cleverly built (but dumb) automaton for an intelligent agent, and an intelligent (but, say, shy) agent for a dumb automaton.

That doesn't mean they are useless, though: operational definitions might not be great qua definitions, but they provide a good empirical foothold for the investigation of the putative definiendum.

  • Thank you, this seems to overlap with my intuitive understanding! Complex-but-dumb and intelligent-but-hidden were my two main concerns too so far. Jun 19 '12 at 7:38
  • One could argue that an operational definition is a case of a definition by extension: those objects which are classified in one way by the operation share some property. But that property is not necessarily intelligence, in the case of the Turing Test. (If there is no way of reliably reproducing the outcome of the test, this just shows that the test itself is unreliable for measuring what it is imagined to measure, or just that the property which it does measure is extremely changeable with time, further suggesting that what people imagine the test measures is in fact something of a chimera.) Jun 19 '12 at 12:48
  • @NieldeBeaudrap Well, it would not be very informative to define intelligence as "the property that intelligent entities have". In general, you want the definiens to appeal to properties other than the one to be defined.
    – Schiphol
    Jun 19 '12 at 15:01
  • @IstvánZachar Yes, this is a problem of operational definitions in general. I have elaborated about it a bit in the answer now.
    – Schiphol
    Jun 19 '12 at 15:04
  • 1
    @NieldeBeaudrap, I think you mean "intensional": the objects that satisfy the definition are picked up by one of their properties: passing the Turing test. But otherwise I see what you mean now. Thanks for clarifying.
    – Schiphol
    Jun 19 '12 at 15:38

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.