Experimental design: science classrooms vs. computer science classrooms

Reflect on the Science practices you teach with regards to conducting experiments in class. How is it similar/different from the experimental design enabled with computational models?
This is a companion discussion topic for the original entry at http://studio.code.org/s/sciencePD-iZone/stage/6/puzzle/4


This format is different because you can control it, but it has randomness that can be measured in multiply times. This can create new hypothesis unlike classic experiments.


Investigation in the classroom are similar to computer simulation because they both require a control group.


Actual lab experiments and computational science experiments both have similarities and differences.

Experiments done in the lab are similar to computational science experiments because the design in both cases is that of a controlled experiment, designed to find causal relationships among factors, by isolating variables. But this is where the similarity ends. The rest of it is characterized by a number of differences that are summarized below:

  1. Unlike lab experiments, computational science experiments can exclude those other variables that are not important to the study, focusing only on those factors and components that are considered important.

  2. Computational science experiments enable researchers to change the behavior of components, which may not be possible in real life.

  3. Experiments can be repeated several times in computational science experiments easily, quickly and with less or no cost at all. This is nearly impossible to do with real lab experiments.

  4. Finally, CS experiments can lead the researcher to build hypotheses, while in an actual lab experiment, one usually does the hypothesizing before proceeding with the actual experiment.


Conducting experiments in class is simpler, testing one variable at at time with everything else fixed. The materials are limited,time is limited and the possibilities of outcomes are usually quite limited. With computational models, imagination removes many limitations, money time aren’t problems and the outcomes could be quite interesting and unpredictable- making for interesting testing by students.


Computational models allow for faster results, meaning you can run another experiment right away. In most experiments in class there is a lot of time spent in preparation, making observations and cleaning up; this lost time that we wouldn’t see in the computational models.


A successful science classroom has students that are actively engaged in constructing their understanding of the natural world around them. They are taking new information and either building on preexisting mental models or they are altering their mental models to accommodate their new knowledge. They are accomplishing all of this by testing their ideas, making observations and interacting with their peers. A science classroom should always be about asking new questions, testing ideas, exploration, discovery, working together and having fun :).


The main similarity between the science practices I teach, and experiments with computational models, is that they both use the process of scientific inquiry, and often controlled experiments. However, in a hands-on lab, we cannot choose the variables we think are important, as you can with a computational model. In contrast, there are often more variables and limitations, which we cannot control for, due to imprecise equipment and changes in the environment.


This format allows for both control and modification, and allows you to run the experiment as many times as you need.


When conducting traditional experiments it is important to isolate the variables so you can test one variable at a time (independent variable), while keeping all other variables constant. The same is true for computational models if we are to gather authentic data and result. However, computational science allows us to quickly repeat experiments, is flexible due to automation, avoid human errors, quickly manipulate variables and allows us to conduct experiments in the lab that would not be possible using the traditional science experiment approach.


I agree with your reflection on the many advantages that computational science have over traditional labs. Particularly your point that experiments can be repeated multiple times quickly and cost effectively. In addition, variables can be changed to answer many “what if” questions that would not be possible using the traditional labs.


Computational experiment also can be done before laboratory experiment to reduce cost and time.
However, in some cases, laboratory experiment must be done to get a accurate result.


In terms of practices, the thinking and analysis required in a hands-on experiment v. a computational experiment is similar. In thinking through procedures you want to isolate variables, you want your procedures to be consistent, you want to be sure you are asking the right question. I think one big difference is how you are limited in a hands on experiment, both by time and by variables you are able to test. Computational models allow you to move beyond certain practical limitations.


They are similar because they can isolate variables and provide evidence of causal relationships. However, simulations allow to you repeat trials more quickly with less materials.


I agree, it could certainly be beneficial for recognizing patterns to be able to run another experiment right away, multiple times, and not have to spend the time reseting and cleaning up the actual materials - just press the set-up button in your program and you are ready to go again! I do think being able to develop a computational model requires a level of abstraction by the students, and students’ thinking benefits from having to set-up and run a traditional hands-on experiment. I think we need a mix of both in the classroom (and I do not believe anyone is advocating for a different position at this point).


In middle school science, experiments are controlled and we have to simplify how we represent certain phenomena to make it possible to experiment in the lab. In addition, we emphasize the need for multiple trials to gather robust data. These are similarities to computational modeling. On the other hand, we tend to do only 3-5 trials in class, for practical purposes, while computational models can be run hundreds or thousands of times (and you can also automate data collection!). Computational models can be used for situations that could never be tested in the classroom due to time/space scale issues, or to experiment with things too dangerous or expensive for the MS classroom. Finally, computational models allow you to manipulate many different variables one at a time or all at once, providing more fine-grained control than messy real life.


The biggest difference I find is that with controlled science experiments that students complete in class there is never enough time to complete multiple trials…we are lucky if we get to complete 3 full trials! This leads students to misconceptions on how data is collected in science. Real scientists in the lab understand the need for a large sample of data and complete experiments numerous times. I think that using computational models is great because students will be able to generate large sets of data quickly and be able to see patterns emerge in their experimentation.


I love your explanation that computer models provide “more fine-grained control than messy real life”. That is so true and I am excited to learn more about computer modeling in the classroom so we can avoid the problems of messy real life!


A way that they are similar is the experimental design process. A way that’s different is time. Computer science can get results faster which can then lead to next steps faster.