Above we discussed how we might design an experimental tool and select participants and what opportunities and challenges digital systems have for the content of experiments. We did not discuss how we assign participants to the experimental conditions. There are two broad categories of how to design experimental studies: between subjects and within subjects.
The between-subjects approach is more common. Each participant is assigned to a single treatment (or control) condition. Code Example 7.1 illustrates how this approach might look through code in the interactive systems. Subjects are randomly assigned into treatment 1, treatment 2 or control conditions. To analyse if the experiment had impact, researchers evaluate if there are differences in the output measures of the research, and evaluations are conducted between experimental subjects. The method allows researchers to examine clear impacts of particular treatments as long as participants in all conditions are similar. Thus, the researcher just needs to measure if participants differed to a large-enough degree between the conditions but were sufficiently similar in background variables.
There is an alternative approach: exposing participants to several conditions. This within-participant research design may be useful for research questions that seek participants to compare and evaluate their experiences. Alternatively, if there is a need to limit the number of participants, the within-participant approach allows testing several treatments for each participant. However, exposing participants to several conditions opens up methodological challenges. How are differences caused by treatments but not other factors like learning or fatigue? We should not expose participants always to conditions A, B and C in the same order but change the order. This is known as the Latin square design process, as illustrated in Code Example 7.2. The Latin square indicates that each of the three treatments (A, B and C) is once first, second and third in the ordering. The participants are distributed across three different conditions, corresponding to different orderings of the three treatments. This means that the same number of participants were exposed to treatment A as the first, as the second or as the third option, enabling similar statistical analysis as in the previous condition.
With this section, the exposure approaches are exactly similar when working with interactive systems as when working with traditional experimental approaches. The biggest difference is that you may need to use code to explicate the assignment criteria by programming them into the interactive system rather than using other means. The common approach is to run between-subjects, and if platforms decrease the cost and time required for massive studies, this may become even more common. However, the within-subjects research approach may provide benefits in special cases and allow more imagination for experimental design.