A fundamental question is how firms adapt to environments that present multiple dimensions. Generally, the number of dimensions may exceed the limits of human attention. Subsequently, as organizations try to adapt to such environments they may be constrained to consider only a few dimensions. In fact, selection of dimensions is a process that may be driven by multiple factors. Especially relevant are the validity, the predictive weight of dimensions, and the perceptual salience of features through which the competing dimensions are perceived. When these aspects diverge, their conflict will affect the learning process. In this paper, we explore how salience and validity interact in the process where agents have to learn to evaluate a stream of task inputs with the number of dimensions that overloads the average short-term memory capacity.
We hypothesize that the interplay between the cue salience and the validity affects the ability to assess accurately the cue-outcome relationship, detect changes in task environments, and adapt decision-making strategies accordingly. To test our hypothesis, we conducted three behavioral experiments and then proposed a classification model that accounts for the observed behavioral outcome. We consider a set of sequential binary classification decision tasks to be completed by a participant. We vary whether the difficulty of the task, also vary whether the performance of another is observable to the participants in the experiment. The behavioral data analysis highlight that cue salience is instrumental in driving learning. The simulation results illustrate that our classification model can indeed replicate the human participant data. The best fit was obtained when the single-cue salience bias was present. This suggests that humans tend to favor decision-making strategies that arise from focusing on a single, yet the most salient, cue when assessing cue-outcome relationship in multi-cue environments. Ours is perhaps the first study to demonstrate the effect of salience and validity interaction on choice and learning in multi-dimensional and changing environments in the context of organizational adaptation. Finally, we discuss variations in our linear classification model and suggest extensions of our experiment.