## Biased epistemology

Posted by Arjendu on April 9, 2008

Or seeing what you want to see.

For the 1st lab in my intro mechanics class, I gave them an assignment I have stolen shamelessly from my talented junior colleague Melissa Eblen-Zayas: The students are given a sham theoretical paper (no jokes about redundant terms please) about a simple phenomenon and asked to construct, perform, and analyze an experiment to see if it’s true [I am eliding details because while I have Melissa’s permission to use/adapt her lab, I certainly don’t have that permission to distribute it].

There’s a lot of coaxing involved to help them figure out how to take data, to consider how much data you might take to account for random error, what amounts to enough variation in parameters to test a theory, why it makes sense to try to plot linear graphs rather than quadratic graphs, etc. The whole process is an excellent exercise in helping them understand how and when we accept our models of the material universe and all that went well — particularly with some of these discoveries coming after the fact, after the groups had come to some sort of conclusion about how the theory was supported by their data or not.

What was really compelling yesterday was to see how many groups (8 out of 10) found that the theory was right, even when the data in 4 of those cases was clearly telling them that it was wrong. And after I told them that the theory was wrong in general even though it worked all right in a limited range of experimental parameters, it was remarkable how none of the groups were unable to see that their data supported the theory (since they’d stayed in that limited range).

Not exactly a newsflash, but students are so used to ‘verifying the theory’ that they really have a hard time seeing anything other than what they expect to see. I am hoping their experience helps them get a little more wary of this sort of bias.

Of course this bias holds true of most of life for most of us so it’s a battle against ‘natural’ human behavior.

Update (courtesy Peter Morgan): This is similar to some ideas developed by ZapperZ.

## Leave a Reply