What is it? How does it work?
Experimental research is the process of resolving a research problem through the use of experimentation.
First, a hypothesis is formed to bring the research problem down to a testable statement or question. This could be something so simple as “What soft drink creates the largest geyser when mentos are added?”. This provides us with what exactly needs to be measured and manipulated to find useful data and eventually conclusions. These measurements or variables are categorised into 3 different groups. Independent, dependent and controlled. There are also confounding variables which are usually not identified till after a first set of experiments are conducted and analysed.
- The dependent variables are the results that are recorded during or after the experiment. From our example this would be things such as geyser height and width.
- An independent variable is the variable that will be manipulated or altered by the researcher to affect the dependent variables. From our example this would be the soft drink that was used in each test.
- Controlled variables are variables which are recorded and kept the same across multiple experiments as a means to avoid random results (confounding variables). From our example this could be the amount of liquid in the bottle, the amount of times it is shaken or how many mentos are added.
- Confounding variables are variables that most-likely weren’t considered before testing that ‘pollute’ or affect the results of the experiment . They come about as a result of poor variable control and can greatly affect the validity of the research. From our example it could be something like not having the different soft drinks cooled to the same temperature as this would affect the rate of reaction.
From this testing we gain raw data and from here we look to analyse what we have recorded and form conclusions to answer, prove or falsify our original hypothesis. Although if the analysed data is random and doesn’t provide useful conclusions further testing may be required to find what could be causing this.
This is experimental research in a general sense but it can be done in a great deal of ways, I could go on for quite a bit longer about group sampling and the different experimental designs but you can read about those on this interesting website I found here.
https://explorable.com/true-experimental-design – For scientific experiments.
https://explorable.com/quasi-experimental-design – For social-scientific experiments.
What kinds of questions/problems might it be useful for?
To put it simply, any question or problem that testing can support, prove or disprove. Beyond this it depends what kind of information is needed to resolve the question or problem. This is where the way the research is conducted might be different and very much ties into the different experimental designs that I have linked above.
For example using true experimental design is useful for scientific research where variables that can be easily manipulated and controlled (such as chemistry or physics). Though inversely this means it would be a bad fit for sociology or psychology as it can be difficult to control or even account for certain variables. That’s where quasi-experimental design comes into play, where it it is known that you cannot control every possible variable but you acknowledge that and produce results that are as good as they are going to get.
In a way, it could also help to reform better questions as having a research problem or question that isn’t testable might not be ideal.
What are the strengths and weaknesses of the approach?
Once again in varies in what experimental designs are used so it can vary.
It’s greatest strength is the fact that if the entire process is done right and the controlled and confounding (whether you know it or not) variables are the same you will always produce the same results.
Obviously if the nature of the research problem is that it is not testable by any means then the approach isn’t going to work.
Convenient control, compared with other research methods, experimental research on the object, environmental conditions have a higher degree of control.
The shortcomings of Experimental Research are mainly man-made, because there are human subjective factors exist, people will perceive the objective facts as they wish to see them. Such as the same user may be in accordance with their usual habits to get completely different results
How could it be used in IT research? Any Examples?
A difficult question.
In terms of quasi-experimentation, maybe recording user feedback on changes that have been implemented in a system over time to see whether the end user feels the system has improved.
In terms of true experimentation perhaps on a smaller and less formal scale debugging and testing could count. As you have a problem that needs to be resolved and you are changing and testing singular things at a time to see whether it has a positive effect. Also from that experience it would help you create more accurate predictions of other problems you are trying to resolve in the future.