Toward an increased understanding of suboptimal behavior from a goal-directed perspective

People often engage in suboptimal behaviors that do not align with their explicit goals. Examples are action slips, impulsive behaviors, and behavioral reluctances. These types of behavior are especially likely when a behavior has been selected frequently in the past and when individuals face poor operating conditions (e.g., a lack of time, attention, or motivation) due to fatigue, stress, or cognitive load. 

Traditional dual-process models explain suboptimal behavior as caused by stimulus-driven processes, in which a stimulus activates the representation of a stimulus-response association. These are contrasted with goal-directed processes in which the response with the highest expected utility is selected, that is, the response with the highest expectancy to result in an outcome, weighted by the value of this outcome. Moreover, traditional dual-process models endorse a default-interventionist architecture in which stimulus-driven processes are the default determinant of behavior and goal-directed processes can only intervene under ample operating conditions. This builds on the assumptions that only stimulus-driven processes are strengthened via extensive repetition and that only these processes can operate automatically. The same assumptions are typically invoked to explain the increased likelihood of suboptimal behavior after extensive repetitions and under poor operating conditions.

Moors et al. (2017) recently challenged the default-interventionist architecture and proposed an alternative dual-process model with a parallel–competitive architecture. According to this model, goal-directed processes are also strengthened with repetition and can operate automatically as well. As a result, they often operate in parallel and compete with stimulus-driven processes. The competition, moreover, should often be won by the goal-directed process so that this process is the default determinant of behavior, including suboptimal behavior. 

In the current dissertation, the implications of this alternative model are tested and explored across three parts. First, we test explanations of the alternative model for suboptimal behavior under poor operating conditions. Specifically, we test whether frequently and/or recently processed expectancies and/or outcomes may be processed more automatically (in the sense of faster), thereby resulting in faulty or incomplete goal-directed processes under poor operating conditions (when time is scarce). Second, we test the auxiliary assumption of the alternative model that the value and selection of a response correspond to the current expected utility of this response. This assumption is pitted against the alternative assumption that the value and selection of a response may be influenced by past outcomes of the response. Third, we explore the practical implications of the alternative model for behavior and behavior change by applying the model to a specific suboptimal behavior, namely the reluctance of researchers to adopt open-access publishing.