The starting point of an experiment as a severe test is to set up a situation that tests an idea and that idea must be causal: one thing should influence another. In HCI-style experiments, this is also expressed as seeing the effect of the independent variable on the dependent variable.