In the Shodor scholars program workshop today, the students learned about computational modeling and population growth. To do this, each student had around 100 M&Ms and spilled 5 of them on a plate. Then they counted which M&Ms landed with the M up. Then they took half of the number of M&Ms on the plate and added that many M&Ms. This showed 50% population growth. Then they repeated the process until they got up to 100 M&Ms. Then they got together as a class and recorded the information as a whole.
After that, Jeff Krause created a line graph of the information. Later the class opened up agent sheets and made a model that simulated the M&M experiment. This basically dumped the M&Ms on the plane and counted how many M&Ms were face up. Then it would reproduce the M&Ms and do the experiment again, just like the students did themselves.
In the afternoon, the students took what they had learned from their M&M population model in the morning and made an AgentSheets model of that same experiment. They made agents for each of the M&Ms, and a "counter" agent that would make sure all of the M&Ms flipped at the same time as well as counting and recording the total number of M&Ms each time step. They then used AgentSheets' graphing feature to create a quick graph of their data. After running the model a few times, the students decided to compare the results their AgentSheets model gave them to their real-world data. They ran 3 experiments using AgentSheets, then exported their data to an Excel document and graphed the data from their real-world experiment alongside the data from the model. Although both sets of data followed an exponential curve, they were noticeably different, which indicated a systematic error in their model; in other words, their model was consistently not behaving exactly like their real-world experiment. The students then thought about what differences there were between the model and real life. The students quickly realized that in their model, they always started with 3 M&Ms facing up and 2 facing down, whereas in their experiment each of the starting M&Ms had a 50% chance of being face up or face down. They also found that in their model they had given M&Ms a higher chance of flipping to face up, which also changed their results. After making these changes to the model, they graphed their results again and found that the data from their model was indistinguishable from the data from their experiment that morning.
After succeeding in making a model from a real-life scenario, the students got a chance to make their own changes to their models. Students added things like students who eat the M&Ms, bombs which destroy M&Ms, and grass for the M&Ms to eat. At the end of the class, students shared their modifications with other members of the class.