Is simplicity always desirable in an explanation?

We all see things through the lens of our own interests, and one of my interests is in the limitations of formal models for thinking in general and scientific thinking in particular. Consider as a formal model one that is applicable no matter the content area.  The equation 2+2=4 is a formal model, because it is true no matter what one is adding.

Is simplicity a good formal model of inductive reasoning, in the sense that the simpler explanation ought always to be preferred and treated as the better one?  I found myself considering this question as I read through the papers by Zemla, Sloman, Bechlivanidis, & Lagnado, and by Walker, Bonawitz, & Lombrozo from two vantage points: Occam’s razor and Inference to the Best Explanation (IBE) or abduction.

When people argue that simplicity (or parsimony) ought always to triumph over complexity, they often cite Occam’s Razor, which does indeed put a premium on simplicity.  However, the most popular translation of his dictum is that, “All other things being equal, the simpler hypothesis is to be preferred”, and there’s the rub.  Sometimes, all other things are not equal; it actually is necessary to offer complex explanations for an account to be accurate.  For example, explanations for many behaviors must take into account both genetics and environmental factors.  Similarly, explanations for why plants thrive typically invoke all three principal factors: good soil, adequate sun, and adequate water.

A complementary approach that also argues in favor of complex explanations comes from the notion of inference to the best explanation (IBE) (or abduction) as a model of how people generate and evaluate explanations.  According to IBE, which both papers mention, one explanation is preferred to another when it provides the more plausible account of an event that is also consistent with well established background information, or what Quine & Ullian refer to as a web of belief.  This, in turn, pays homage to the fact that explanations do not exist, and are not evaluated, in isolation.  Rather, any explanation is imbedded in and judged with respect to what else we know about the world.  Evolution is imbedded in, for example, information about natural selection, plate tectonics, population genetics, etc.

The approach taken by Zemla and colleagues acknowledges this when they note that, in everyday explanations, we should consider how an explanation fits with our broader knowledge of the world.  As a result, “Everyday explanations are typically more nuanced than experimental stimuli”: people know that certain events have multiple causes.  Put differently, everyday explanations are typically imbedded in a fairly large web of belief.  The results by Zemla and colleagues take account of this.

I have one quibble with their conclusion.  They conclude that people want to invoke as many explanations as possible to make the result “inevitable”.  However, perhaps it is not so much that people are trying to describe inevitability as it is that people in the study have learned that events of the sort used in the studies often have multiple causes.

In the studies by Zemla and colleagues, complex explanations involved multiple causal mechanisms for a single event.  In the studies by Walker and colleagues, the data presented to the children could have (I think reasonably) been interpreted as showing either that there was one cause for both events or that there were two different causes for each of the two events.  The oldest children in the study, six-year-olds, opted to assume one cause for both events. I found myself wondering what would have happened had there been an adult group as well.  Age six might not have been the developmental end point of this sort of reasoning.  By adulthood, people might be more likely to have learned that, even when dealing with similar plants, one plant might suffer because of poor soil and insufficient water and another (similar) plant might suffer because of poor soil and lack of sun. Adults might have been less likely to show a strong preference for a single cause rather than multiple causes, because they might have already learned that, in everyday situations, multiple causes are the norm.  They might have learned that simplicity cannot always accurately describe the real world.

You may also like