Adaptive design

Experiments are usually designed “statically”, before they are conducted. While this allows a large degree of experimenter control, in order to design an experiment optimally, we usually need to know quite a bit about the expected behaviour of participants in an experiment. Moreover, even if we succeed in designing an experiment optimally for an average participant, when there are large individual differences, the design will be suboptimal for a large proportion of participants. In adaptive design, an experiment is designed “on the fly”, while it is running. Using computational models, we can determine relevant characteristics of a participant from their earlier behaviour, and tailor the design to the participant. This can have substantial benefits when conducting studies to e.g. compare different theories.