Tuesday, September 14, 2010

Empirical Psychologists vs Armchair Philosophers 1

Some weeks ago in this post at his blog, Andrew was complaining about philosophers threatening to be irrelevant with their out-of-this world thought experiments and imaginary scenarios and the like.  So, being the small-minded person that I am, I'm pleased to find Runeson, of whom Andrew appears to have a decent opinion, has a somewhat less hostile take on out-of-this-world thought experiments
In the ecological perspective an interesting empirical problem follows: How would we fare perceptually if we were placed in a world that was less constrained-for instance, one in which some of the above constraints on room shape were relaxed? An Ames' room viewed from noncanonical points would instantiate such a condition. Those (as yet unspecified) motion-independent invariants whose validity is conditional on normal shape constraints would be specifying the wrong spatial layout. To the extent that those invariants remained perceptually relied upon, the perceiver would be in trouble. Could perceptual leaming effect reliance on only the types of information that remain fully valid (motion perspective, binocular disparity, fine texture gradients, etc.)? If it could, how efficiently, comfortably, and confidently the perceiver could function in that environment appears an open question of both theoretical and applied relevance. (Runeson, 1988, p.302).
What if we lived in a world with different laws of physics?  When is that going to happen?  Yet, such a scenario is one that might have theoretical and applied relevance?  Would Andrew have a (spherical) cow thinking about these questions? =)

Maybe the divide here is not between philosophers and psychologists (as Andrew seems to suggest), but those who are and who are not open to hypothetical reasoning.
Runeson, S. The Distorted Room Illusion, Equivalent Configurations,
and the Specificity of Static Optic Arrays. Journal of Experimental Psychology: Human Perception and Performance 1988, Vol. 14, No. 2, 295-304.


  1. Actually my point was that I think philosophy is intrinsically more susceptible to wandering off into no-relevance land by not being connected to any data; the reason I'm a scientist is that I always end up wanting to actually run the study and get some data on the actual system :)

    I actually don't think this is quite a relevant comparison though. Runeson is asking a question about how an actual system might cope with a specific perturbation via learning. This is entirely different from trying to make an argument about how things work by first stripping away all the annoying complexity; Runeson is still firmly connected to the actual system of interest (the human perception/action system).

    The 'ecological laws' paper touches on this topic more generally by emphasising that ecological laws have scope; this scope is critical for understanding how the system actually works. Breaking that scope will then, of course, have consequences.

    (Space travel is where this will really become relevant, and the question will still be 'how do human perception/action system, which have evolved in the context of Earth, adapt to a new context?', not 'imagine there was never any gravity'. But zero-G is indeed a pretty weird place, perceptually; apparently astronauts often lose track of their orientation and the quickest way to regain it is to look out the window at Earth to visually regain a reference frame. "The enemy's gate is down" :) )

  2. I am mostly writing this tongue cheek, but he's talking about a world with different geometry! I was just talking about using a computer that had a long delay before performing a computation. I think he's farther out that I am.

    Now, you might say that if we were to move into region of spacetime with dramatically different gravitational fields, then Runeson's scenario might hold. But, then, would we even be able to survive those fields without being crushed? That, too, sounds more far out than using a somewhat eccentric computer.

  3. I agree with Andrew that these are very different types of examples. Runeson proposes a world with different physical laws, which would result in different types of information specifying the invariant structure. The nature of the perception/action system is assumed to be the same in our world and this hypothetical world, but the details would change. Of course, if this world were so unpredictable that there were no invariants then we wouldn't be able to cope with it. But that's not any more damning than suggesting we couldn't survive in a world without oxygen. It's just a consequence of the type of organism we are. With the delayed computer example, we're asked to imagine ourselves behaving in a quite strange way. What is the behaviour for? Presumably we need to compute prime numbers (or whatever the task), but then, why are we willing to wait so long for the answer? The person executing this task is pretty strange. I think the critical difference is that the first example is talking about typical behaviour in a strange world, while the second example is about weird behaviour in a normal world.

  4. Hi, Sabethg,

    There seem to me to be at least two issues here. 1) What is the methodological stricture on what is of scientific interest?
    2) Why following this methodological stricture?
    Let me focus on just the first for now.

    Let me try to pin down the source of the weirdness. The human being is perfectly normal, right? Pressing the return key is a perfectly normal action, right? Having a computer run a program for computing prime numbers is not weird? Printing out prime numbers with a computer printer is not weird. The task "Print out the first 1,000 prime numbers" is not weird.

    Maybe you might think the time delay is weird. So, would it be ok if this delay were only 1 msec? If so, what is the limit of non-weird delays?

  5. Yes, I think those are relevant questions, but I want to clarify a couple of things.

    1) I never argued that the computer example is not of scientific interest. It's simply a different type of example and the way in which it's different is not trivial.

    2) Arguing for normality on the basis of task decomposition is weak. For one thing, individual actions can seem perfectly normal, but become abnormal or in combination or in specific contexts. Tell me you wouldn't be concerned about a guy who was going to wait a hundred years for a program to compute some prime numbers? Most likely we would assume he was doing some kind of performance art.

    3) It does fundamentally change the nature of the problem to assume a shorter time-scale. If he gets an answer in 1msec then he can use that information. If I am having a discussion with someone face to face, then a delay of a few seconds is no big deal, but a delay of a few minutes becomes problematic (e.g., because of forgetting). And, the discussion ceases to be a discussion if the delay extends to 120 years because one or both of the partners is going to be dead. Change the context slightly (e.g., to a conversation via the internet) and the relevant time scale also changes (i.e., a delay of minutes or days is acceptable). Time matters, but this is neither here nor there.

    4) I would like to hear your opinion of the point I was raising, namely, that what the behaviour is FOR determines what we might conclude about the computer example. This is similar to the point I was making in another post when I suggested that the fact that someone can press a button accidentally or while intoxicated changes the way we should think about the activity. Is a toddler who presses the button because it looks pretty doing the same task as an adult who is using the program to perform a job at work? Without further constraining the example there doesn't seem to be much we can learn from it.