LogicWarrior Demand Reason

7Nov/100

Underdetermination and You

Summary: Underdetermination occurs when a group of facts could be explained by multiple theories and a theory that has a rival theory in this case is said to be underdetermined.  Remembering underdetermination is a powerful way to avoid jumping to conclusions in cases where an immediate conclusion is not needed and reminds one of the important of context in forming conclusions.  Check your theories for underdetermination by seeing if another theory could explain present facts and if you find one, find an experiment that could separate the two.

Background on Underdetermination: I first ran into underdetermination when arguing with someone about the usage of science as a way to find out things about the world.  He retorted that most areas of knowledge were underdetermined and, I having no response to this, ceded that he had used argumentum ad dictionarium to win.  Underdetermination is simply the phenomenon whereby a given theory that is otherwise incompatible with another theory both explain a given set of facts.  Scientific examples of this are legion as underdetermination exists until one find the fact that is explained by one theory but not the other and experimentation allows one to sift out the invalid theory.  Underdetermination can also be taken to extremes to question all conclusions made from observation but that is beyond the scope of this piece.

Underdetermination is easy to spot when there are two theories, but much more difficult when there's only one.  How are we to know that the Newtonian model has the monopoly of truth on orbital mechanics?  Quite simply, we don't as there could be an equally valid theory that posited the same as Newtonian mechanics but only for our region of space and we'd have no way to dispel it.  Let's focus on systems that are a bit less abstract.

Underdetermination in Human Interactions:  Figuring out what's going on is hard and I'll use Andrew Wyeth's Christina's World as an example.

Copyright Wyeth Estate, used here for commentary.

The constellation of facts as presented is a woman in a field gazing towards a farmhouse.  When I first saw the piece I thought it was a statement about human aloneness, or the status of women, or rural life or some such thing, only later did I learn that the person depicted had her legs paralyzed from polio and crawling is the way she got around.  All my theories were underdetermined (each could be valid with the available facts) but that extra context ruled them all out.  The same thing happens when we interact with someone a limited number of times.  People are complicated and only over long periods can enough data be collected to create cogent narratives of how someone operates.  So, why not just say "don't jump to conclusions"?  Because this only hints at the depth of the phenomenon, and because sometimes you should.

The Cost of Underdetermination: If you're in a scenario with a high failure cost or one where the maximum amount of information available is low, running with an underdetermined theory is the best you can do.  I'd consider it unreasonable to wait until a realtor has sold a significant enough sample size of houses to draw a conclusion if one is planning on moving at a particular time so their five house track record may be enough but otherwise, waiting is almost always an option.

You meet someone at a party and decide they're a jerk after talking for about 10 minutes about music, but this could be caused by the person in question having had a bad day, you misreading the interaction, the topic of music being one on which you disagree, the other people in the group agitate you, or he indeed being a jerk.  But what's the cost of delaying that assessment?  It's a party so you'll mingle with other groups over time or move onto other topics and probably run into him again unless you make a conscious effort to avoid.  This second chance encounter allows you to potentially rule out topic, mood, and grouping all at a trivial cost.

Tactics: If you do find yourself in a case where two theories could explain something, ask "what would differentiate the two?"  I had a coworker who gave me stern looks when I shouted an expletive upon injuring myself or are test going awry so I assumed he didn't like curses but realized he may not like loud noises.  I tested this by just yelling "GHAAAAA" and I got the same look.  Until that test, there were two reasonable possibilities; not liking yelling, or no liking cursing and both were underdetermined until that test.

Stanford Encyclopedia of Philosophy Entry

Wikipedia Entry on Christina's World