Probabilistic functionalism as a limiting condition for robustness

When should we expect behavioural phenomena to be robust? We argue that many phenomena of interest to behavioural scientists, by their very nature, involve manipulations of stimulus characteristics. If there exist contingencies between those stimulus characteristics and outcomes, the former will consequently constitute cues. People may then pick up on whether the cue is guiding them towards their goal or not and adapt their behaviour accordingly. On this view, the robustness of such phenomena is, at least partly, determined by the cue structure in each given setting. In an experiment, we demonstrate that the attraction effect and the default nudge obtain proportionally to how well the manipulated stimulus characteristics predict the superior option. A similar result is found under a more traditional rule learning manipulation. We suggest that the existence of cue-outcome relationships is an (underappreciated) limiting condition for the robustness of behavioural phenomena. We discuss the implications of this perspective for nudging.