The Trap of Expecting an Expected Outcome

http://thesheetnews.com/2011/09/02/lord-of-the-flies/

The best we can ever do is keep asking questions. To continually look for the novel or the overlooked. To not expect the answer we were expecting. These principals are just a few that come to mind when I am thinking about systems, either environmental, or political. As an ecologist, I have been charged with the responsibility to poke-holes in ideas, find flaws with hypotheses, and only stop researching when my Write in the Rain notebook has been pulled from my cold, dead hands.

Systems are complex beasts that can never be controlled; though they can be observed. The longer one observes the more complex the system inevitably becomes – think factorial. Specifically, I studied interactions of a micro-ecosystem on sage brush in the Eastern Sierra. The research was focused on one fruit fly, Eutreta diana. This all sounds simple enough, but pull back by one level of the hierarchy and we have to account for the sage brush itself, the water it drinks, the nutrients it eats, and the very specific phenols it produces. Add another level and we find ourselves looking at the mycorrhiza, and competition with nearby plants. Add a few more levels still: we can now see multiple sage brush ecosystems interaction with each other; with various populations of fruit flies interacting, equalizing the number of flies per system.  This hierarchy will expand indefinitely… where does it end?

We need limits to contain our systems. Without them chaos ensues, making it nearly impossible to see the interactions within the system! By breaking down this ecosystem into a handful of systems – the fly, the sagebrush, and the other insects – I was able to control the variables much easier. In one experiment I controlled the amount of water and nutrients a number of sagebrush received (some more, some less, some were status quo) to see how and if it effected the number of flies on each plant. We weren’t concerned with other plant species, other insects, or other populations, only the sage brush in a 100meter radius. After approximately 9 months controlling the sage brush nutrient and water intake we found that (drumroll, please) no matter how we controlled these variables, there was no significant change in the number of flies that were inhabiting the plants.

This was a total shock to me. Even more shocking was that an experiment that looked at the different phenols found that each specific phenol-mix released by a given plant had no effect on attracting the flies, regardless of how the choice assay was run, or which phenol was present. It seemed that none of our experiments that we had conducted over five field seasons yielded no significant response… except for one: flies seemed to be attracted to sage brush which other flies had already oviposited (laid eggs) on. Why?

Why do we fall into the trap of expecting an expected outcome? How do we create a more accurate model of our world?

Comments

[Mallika]

Your anecdote raises an interesting point about the challenge of limiting your model of a system so that you can understand it, and ensuring that you see factors that influence the outcome of your model. I won’t pretend to be an expert on fruit flies and sage bushes in the Eastern Sierra, but based on your post, it sounds like your original model was missing some interconnection between fruit flies ovipositing on a plant and future fruit fly behavior. I think this example demonstrates the importance of checking models against observations or experiments to ensure the model’s underlying assumptions are correct. As we saw in your case, this practice may alert you to an interconnection you hadn’t considered previously. I think it’s a good example of why modeling should be viewed as a trial and error process.

[Alec Brazeau]

While I agree with your claim that our systems need defined limits in order to be observed meaningfully, I think that placing limits on systems is limiting (sorry). In order to truly see the interactions and interconnections within a system, the system needs bounds, as you state. However, I think that bounding and limiting a system can cut off important parts of a system and leave them out of our examination of said system. For example, if you were to look at the U.S. Government as a system, it makes sense to bound the system strictly to the well-defined functions and actors in the government. However, this leaves out the media, the public, the international community, and all of the interactions that occur between these actors. So therefore, while I agree bounding a system allows for better analysis, I would urge care when setting up these limits, because it is very easy to leave out important actors/elements of a system.