Page 140 - Getting the Picture Modeling and Simulation in Secondary Computer Science Education
P. 140

138
Chapter 6
model’s outputs fall into acceptable range (operational validation). The parameter variability - sensitivity analysis technique was employed by groups G1 and G2. For example, student S2, who modeled potato farming, when asked if their group was playing with the distance between potato plants, answered, “yes, that is the density percentage. How many plants are being planted.” Almost all the groups employed operational validation to make sure their models’ outputs fall into acceptable range. Student S4, modeling the evacuation of people in a burning building, said, “Alarm bell number 5 was in the middle, so then you had alarm bells, four emergency exits, that was I believe almost the best. [...] was around 600 [tics], everything was around 600, that was good. Except, one alarm bell [only], that is alarm bell number one, I believe it was down there in the left corner. Well, then it takes really 900 tics.”
The students we interviewed employed the parameter sweeping techniques not only to validate their models by getting confirmation that the models were good enough, but also to calibrate their models by choosing the appropriate parameter values, and during experiments which some of them saw as additional confirmation of the validity of their models as well. Groups G1 and G5 are examples of groups who employed systematic parameter sweeping while running their models to perform experiments and reported that the results thus obtained additionally convinced them of the validity of their models. So did group G3 too: they tested various parameter values and discovered that their Mars colony was sustainable when it contained 17 people. Group G4 also reported finding parameter values that guaranteed a desired constant output value.
Several students report hard coding some of the values into their models rather than having the user determine these parameter values when running the model. For example, the model made by group G6 has no input parameters at all, but the model’s parameters are determined at random during the execution of the program.
Observing Outcomes
As the outcomes are generated, there are two main ways to observe the outcomes and behavior of a model: visual inspection and making use of quantitative performance measures.
Visual Inspection. We saw that almost all the groups employed visual inspection either by looking at the animation of the whole model, or by tracing



























































































   138   139   140   141   142