Monday, May 27, 2013

What is an Experiment?

What is an experiment?
Experimenting with color (source)

People use the term 'experiment' to meant a lot of things. One may say "she experimented with drugs" or "she performs experimental music". Someone might 'experiment with ones hair' or 'perform a thought experiment'.

All of these uses of the word experiment have distinct connotations, but most of them essentially mean 'to try something and see what happens'. In the examples above, most of the phrases also imply that the experiment is something new. If she experiments with her hair, she's probably trying some new style and seeing if she likes it. If she performs experimental music, she's probably not following the conventional rules for the music she is playing.

Are these kinds of 'experiments' different from the experiments that scientists do? Well yes and no.  The basic definition of an experiment as 'trying something [often new] and seeing what happens' is pretty much what scientists do. So what's different? Why isn't someones 'hair experiment' publishable in a scientific journal?

Mythbusters would have you believe that the only difference between science and screwing around is writing it down:

And that is sort of true.

But what really makes a scientific experiment scientific is controls. In our hair example, you can experiment with your hair by dying it black, and seeing if you like it. But that's not the scientific experiment. To be scientific you would have to decide how to measure how much you like your new hair color. You could do this by filling out a survey each day asking you how many times you thought you were pretty or rating your confidence on a scale of 1-10. You could fill this survey out for a week and then dye your hair and fill the survey out for another week. You could then compare the scores and decide if the new black hair had a 'significant' impact on your self-image. 

Lets say it does impact your self image and you report higher self-confidence that second week. But what if you feel different just because you have a new hair color, not because you have black hair?

Well, you would want to do a control experiment, which controls for the newness of the hair color. You could control for novelty by dying your hair yet a different color, and taking the survey for another week. Or you could take the survey two months after you dyed your hair black to see if you still report higher confidence or if your confidence has dropped back down to normal.

This is not a perfect experiment by any means, it's not even a clever or well-designed one, but it is somewhat scientific. And illustrates what I think is the most important difference between experimenting as in trying something new, and experimenting as in trying to find something out:

The control group

In addition, here is a great example of how important the control group is in science. (See the epilogue)

© TheCellularScale

1 comment:

  1. My take on the difference between an experiment and everything else (which I generally call a study) is not controls, but manipulation. Do humans actively intervene in the system? If so, it's an experiment. To me, an uncontrolled experiment (manipulation without controls) is still an experiment. It's just a very bad experiment that you can't make many conclusions about. Could be suggestive, though.

    If not, then it is at best a natural experiment. The qualifier matters.