0

The Turing test seems often to be regarded as an operational definition of human-like intelligence (eg in Russel and Norvig, AIAMA). What is an operational definition and how does the Turing test regarded as an operational definition differ from the Turing test regarded as an empirical test which tells us something about the world?

Roddus
  • 629
  • 3
  • 14
  • 1
    "An [operational definition](https://en.wikipedia.org/wiki/Operational_definition) specifies concrete, replicable procedures designed to represent a construct." It is an empirical test of "construct"'s presence, in this case of "human-like intelligence", especially useful when full theoretical understanding of it is lacking and empirical inquiry is aimed at developing it. – Conifold Jan 31 '23 at 09:27
  • So the concrete replicable procedure is the Turing test as Turing (1950) describes it. What is the (presumably theoretical?) construct the procedure is designed to represent? And by virtue of what, exactly, does the procedure *represent* the construct? How is this type of representation to be understood? Many thanks. – Roddus Feb 03 '23 at 00:18
  • The mechanism of representation in operational definitions is subject to future study, just as the supposed construct itself, it may well be coincidental. All that matters is a good correlation between what the procedures detect and the intended object of study (mastery of language vs intelligence). In the course of that study, it may well turn out that the expected correlation was not that good after all, and the original Turing test has been much reinterpreted in the light of criticism and experience with chatbots, see [SEP](https://plato.stanford.edu/entries/turing-test/#Tur195ImiGam). – Conifold Feb 03 '23 at 02:54

1 Answers1

2

The first time I heard of operational definitions was a text, title long forgotten, on time. The fourth dimension is notoriously difficult to pin down. So, as I was told, scientists simply define time as that which a clock measures. This is an operational definition i.e. an operation needs to be performed (herein look at the hands of an analog clock) to find out what time (it) is. As far as I can tell, this is done to avoid what Robert M. Pirsig in his book Zen and the Art of Motorcycle Maintenance calls getting stuck.

Operational definitions looks like are primitive definitions for example water pre-chemistry would've been defined as e.g. that which quenches thirst or that which puts out fire. Conifold mentions quite clearly this aspect in his comment: "theoretical understanding (of it) is lacking".

Agent Smith
  • 2,488
  • 6
  • 22
  • 1
    Thumbs-up, but it could be slightly better. There's a tiny bit more to the definition. Different clocks will be more or less desirable for various purposes. The notion of a *good* clock is based on the purposes. A *good* clock is one that gives a version of time such that motion looks simple. https://van.physics.illinois.edu/ask/listing/1361#:~:text=To%20quote%20Misner%2C%20Thorne%2C%20and,in%20excited%20states%20of%20atoms. – Boba Fit Jan 31 '23 at 21:42
  • @BobaFit Could you give an example of what you mean by that? Like it's not really about a real clock but just about "what a clock would read in that situation", doesn't matter what kind of clock. Just that digital clocks would introduce a incremental progress with delays so something sufficiently continuous would be good, but that would be good in any context. So what do you mean by "motion looks simple"? – haxor789 Feb 01 '23 at 07:48
  • Imagine using a pendulum clock. Imagine using the position of the pendulum at any given spot in its swing to indicate time. If you had the small units of time evenly spaced along the swing, then it could give smaller units. But if you used this to time the motion of some object you would see that object apparently speeding up and slowing down. That's because the pendulum is speeding up and slowing down. So the pendulum can be a clock this way, but not a *good* clock. Only the full swing works properly this way. To make a good clock you would need to space the divisions very carefully. – Boba Fit Feb 01 '23 at 13:05
  • "Operational definitions looks like are primitive definitions for example water pre-chemistry would've been defined as e.g. that which quenches thirst or that which puts out fire", but many things quench thirst, and many things put out fire. So "puts out fire" couldn't be a definition of what we usually call water since dirt also puts out fire. I'm interested in the idea of definition as a linguistic object which object says nothing new about the world, compared to empirical testing which does. Is an operational definition merely a common-or-garden definition (linguistic object)? – Roddus Feb 03 '23 at 00:33
  • I was comparing water to a primitive human, its essential qualities as manifested to him via its utility to him, to water to a chemist. The utility-based definition is basically what an operational definition is. – Agent Smith Feb 03 '23 at 02:24