Next weekend, groups of college students from around the world will take part in a large (but rarely publicized)
competition.
These students typically are in mathematical, engineering, or hard-sciences curricula at their school. They attempt to work in teams to analyze, model, and solve complex problems drawn from the real world.
The problems range from efficient construction of
stunt sets in Hollywood to using computers to help medical doctors use sophisticated cancer-fighting tools (at the same link, further down the page).
The students are challenged to do research, create a model of the problem, and use the model to produce a solution. Before the end of the competition, the team must produce a paper which presents their results.
Out of the hundreds of teams which take part each year, a large number are included in the Meritorious category, and a small number achieve the Outstanding category. Outstanding papers are published by the
agency that oversees the competition.
Some years, the Outstanding teams get an opportunity to present their results before government or industry leaders who are interested in the problem. In 2001, for example, a large number of teams studied the
evacuation of South Carolina coast under the threat of a hurricane landfall. The authors of the
Outstanding papers were invited to present their results to the government agencies of South Carolina who oversee such things.
I mention this partly because I owe a great deal to this competition. Without it (and without the professor who encouraged me to become involved), I would have not thought of pursuing my interests in mathematics.
It also taught me the strengths and weaknesses of computer models. A good modeling team can create a powerful model, but if the model ignores (or assumes to be unimportant) an important real-world variable, then the model has only limited use.
One example may help this: in a problem dealing with computation of
lawful capacity of a public building, the modeling team must take several variables into account. Among these variables are the floor space in the building, the pathways to the emergency exits in the building, the type of the exit (single door, double door, sliding door, etc.), and the distribution of the exits around the edge of the building. However, the modelers will also need good information on how to model crowds of people moving through a restricted passage. Each of these factors may need to be modified by the kind of emergency which occurs.
If the modelers ignore any of these factors, then the usefulness of the model decreases significantly. However, if they pay attention to all pertinent factors, the model can be used as a good resource. At this level of certainty, the model is often said to have proved a particular result.
Of course, there are also computer modeling attempts which are much more
open-ended. When studying events like asteroid impacts, the number of possible variables scales upwards into the hundreds. There are wind patterns, dust clouds, the short-term (and long-term) effects of asteroid impact in the middle of a glacier. There is also the possible alteration in reflected and absorbed solar energy, due to the aforementioned dust clouds. Such alterations will affect different regions of the planet in different ways; they may also affect the severity and quantity of storm systems over the oceans. The number of variables involved in each of these calculations is large. The number of possible interactions between the variables is also large. Lastly, the interactions aren't all well-understood.
In these situations, the model may produce a large number of predictions--but it is much harder to say what exactly the model proves. That question is rather open-ended, especially if empirical data is lacking on the subject.
In the case of large asteroid impacts on Earth, I suspect that few people want to actually see such an event happen, even though it will generate great amounts of empirical data.
Computer models are powerful tools. But like most results of academic study, the results of the model are as dependable as the sources used to create the model; the results are also as rigorous as the analysis done by the model architect. Lastly, the usefulness of the modeling results are heavily dependent upon the assumptions incorporated into the model.
Labels: math, programming