Here's the scenario:
Students are in a lab, modeling groundwater flow. They make an observation. They come up with an explanation for their observation. And then, they are asked to describe something, anything, they can do to test their model.
They say to make the same observations that they started with.
So. Repeating the experiment is reasonable to see if their observations were just a fluke, or whether they made some mistake. But they've got to do something different if they really want to test their hypothesis.
If I talk to them while they're working on this part of the lab, I tell them that they need to propose to do something different. But if I don't point out that there's a problem with confirming your model by going back to your original data... they don't see that there's a problem.
It strikes me that this is an important point about science, here. And it's a particularly tricky problem for geology, because we constantly look at things that have happened long ago. (That 1.4 billion-year-old granite whose contact aureole I'm studying? It's done intruding, it's cold, it's eroding. It's not going back to the mid-crust any time soon.) So we constantly have to think about what other things our models might predict. Is there any place we haven't looked, or any kind of data that we haven't collected, that could test our model?
If students don't understand that point, it's easy for them to conclude that geology is just a form of story-telling. Not that there's anything wrong with story-telling - I love a good story - but it's much, much more than that. And that, for me, is the fascinating part of it.
But how could I make that point to the students, better than I am doing it now?
Wednesday, November 7, 2007
Teaching: testing conceptual models
Subscribe to:
Post Comments (Atom)
11 comments:
I think this is a very good point, and to be honest, not one I think enough about myself. I would imagine you already do a pretty good job by actually pointing this difference out to the students.
I often think that the concept of hypotheses in geology works somewhat differently than in other natural science fields (correct me if I'm wrong, I'm thinking out loud right now). But the problem with a geological model or hypothesis is that often we cannot test the particular model itself (in that case I would like to build a very big floodplain and sit around and wait for some 20 million years and see what happens), but test small part of this idea indirectly. It took me a long time as a student to fully realize the consequences of this and in a way I think I'm still struggling with it from time to time. I'm thinking about the way one cannot be sure that it is possible to test a hypothesis in the field (depending on outcrops, what is available) or how research proposals/ programs are often designed more like "We want to find out more about problem x" than to specifically test a hypothesis.
Maybe it is really important as a geology teacher to make sure we use examples of how to test and challenge conceptual models in lectures/ seminars. One practical option could be to bring in other types of data that might support the models the students are developing or not. I don't know how you run your practicals of course, but often excercises work on one type of data a time (like today we look at core data, next week we will work with thin sections). Maybe it is important to show/use multiple sources of data in the same class to make the point.
Maybe 'testing your model' is the wrong way of putting it. Although it's implicit to those of us more familiar with the scientific process, testing in science isn't like testing a fuse, or the fuel economy of your car: what you're really looking for are consequences. Say you can measure X, Y and Z in a system. Your model is based only on measurements of X. Therefore it obviously (you hope!) 'predicts' the behaviour of X - but what does it say about the possible behaviour of the things you haven't yet measured: Y and Z?
For example (and this may not be the best demonstration but I have plate tectonics on the brain at the moment), when we discovered that ocean crust was being created at mid-ocean ridges, it logically followed that unless the Earth was getting bigger, somewhere crust was being consumed at an equal rate. Sea floor stripes predict subduction zones!
By the way, that groundwater model looks really cool...
Geology seems to me, to be a bit like Evolutionary Biology. Of course, you can have a hypothesis about how something happened - but a TESTABLE hypothesis requires more work.
I loved this post of yours - because, as a molecular biologist, I'm often faced with a similar situation where one experiment provides you with support of your theory, helps you form a hypothesis about what may be happening - but a second and (often) third experiment, using different methodologies, is always needed to be conclusive beyond reasonable doubt. So often I come across papers that basically read:
We used technology X and discovered Something New. We therefor conclude that Something New is important.
But they often fail to validate Something New with Technology Y and Z.
Keep up the good blogging. Outside my field of expertise, but a good read everytime. Cheers -- Jonathan.
Saxifraga -
Yes, I agree that the concept of hypotheses works somewhat different in geology. (In fact, I've got an article saved in my teaching stuff about that very issue: Cleland, C., 2001, Historical science, experimental science, and the scientific method: Geology, v. 29, no. 11, p. 987-990.)
I think that "we need more data" can be a valid way to test a model, particularly when the existing data is very limited spatially. (For instance, you've got a few outcrops, and a few drill holes, and maybe some seismic data - does that characterize the geometry of your floodplain deposit in enough detail to support your explanation of a process more than an alternate explanation, or are the limitations in the amount of data great enough to allow many possible interpretations?)
And as for using multiple types of data - some of our labs are focused on techniques, and some are focused on looking at a particular problem. I'm working on developing a group research project for the intro class that involves comparing various kinds of data - maybe when that's actually implemented, it will help.
Chris -
I think I actually used the word "hypothesis" in the lab (because I wanted to make students explicitly aware of the scientific method). And I guess I would like to keep the same language, simply because I want the students to realize that this is what scientists do.
And, yes, plate tectonics is a great example. I guess I would argue that if we understand something really fundamental, then the model should predict other behavior or other observations. Plate tectonics definitely does. So does evolution. (And so, I would argue, does the modeling of temperature changes associated with CO2.) And that's a point that I would like the intro students to take away about the sciences in general - this is why evolution isn't simply a matter of faith in science as opposed to religion, for example.
And, yes, the groundwater model is amazing. Groundwater is such an important resource, and it's so hard for the general public to understand... and these models let people see the water flow. (The link in my post is to one group who sells it. I think Ward's Scientific does, as well.)
Jonathan -
Thanks! And, yes, I agree entirely that geology is much like evolutionary biology. (In fact, paleontology pretty much links the two disciplines.)
So often I come across papers that basically read:
We used technology X and discovered Something New. We therefore conclude that Something New is important.
I think that this is mostly a problem of the Minimum Publishable Unit. Our productivity as academic scientists is measured (in part) in numbers of papers, so we are rewarded for emphasizing the significance of each individual part of our research. Also, sometimes the best test for a new model involves using techniques that one may not have available, either because they involve another subdiscipline, or because they require instruments that your research doesn't have. For instance, I don't have access to instruments that do uranium-lead dating, so if I want to know the age of zircons in my rocks, I need to convince a collaborator that my problem is significant enough for him or her to want to date my rocks.
I tend to find models more convincing when different research groups with different approaches (and different expectations) end up agreeing on them. That's an even stronger confirmation than multiple types of data from the same research group, in my mind.
Thanks for the reference. I will look it up now.
Whadda you mean we don't test hypotheses? My boss hypothesised that there was an economically viable uranium deposit out in the middle of the outback, and I spent the last two months sweating it on a RAB rig in order to test that hypothesis.
Exploration is very scientific methodological. We formulate hypotheses during the wet season and test during the dry.
:D Exploration (both mineral and oil/gas) is the place where structural hypotheses get tested the most.
It's got to be pretty cool to draw a somewhat interpretative cross-section, that's got its start in surface geology, but that uses all kinds of models of tectonics and sedimentary history and fault behavior to fill in the mystery stuff underground... and then have somebody drill it and tell you that you're right.
I'm interested in the referenced paper above... but being that I am at the NIH - we don't generally carry subscriptions to Geology journals. =D Any chance someone could forward me a PDF?
I've got the dead tree edition of that issue of Geology (and my library only subscribes to the dead tree edition - yes, I know that's very 20th century of us, and we're constantly talking to the librarians about getting electronic access). So... if nobody responds, maybe interlibrary loan? (That's the way I read nearly everything.)
Post a Comment