Abrupt strategy change underlies gradual performance change: Bayesian hierarchical models of component and aggregate strategy use.

While researchers have often sought to understand the learning curve in terms of multiple component processes, few studies have measured and mathematically modeled these processes on a complex task. In particular, there remains a need to reconcile how abrupt changes in strategy use can co-occur with gradual changes in task completion time. Thus, the current study aimed to assess the degree to which strategy change was abrupt or gradual, and whether strategy aggregation could partially explain gradual performance change. It also aimed to show how Bayesian methods could be used to model the effect of practice on strategy use. To achieve these aims, 162 participants completed 15 blocks of practice on a complex computer-based task-the Wynton-Anglim booking (WAB) task. The task allowed for multiple component strategies (i.e., memory retrieval, information reduction, and insight) that could also be aggregated to a global measure of strategy use. Bayesian hierarchical models were used to compare abrupt and gradual functions of component and aggregate strategy use. Task completion time was well-modeled by a power function, and global strategy use explained substantial variance in performance. Change in component strategy use tended to be abrupt, whereas change in global strategy use was gradual and well-modeled by a power function. Thus, differential timing of component strategy shifts leads to gradual changes in overall strategy efficiency, and this provides one reason for why smooth learning curves can co-occur with abrupt changes in strategy use. (PsycINFO Database Record

Workplace accommodations for employees with disabilities: A multilevel model of employer decision-making.

PURPOSE: Existing research suggests that the decision to grant or deny workplace accommodations for people with disabilities is influenced by a range of legal and nonlegal factors. However, less is known about how these factors operate at the within-person level. Thus, we proposed and tested a multilevel model of the accommodation decision-making process, which we applied to better understand why people with psychological disabilities often experience greater challenges in obtaining accommodations. METHOD: A sample of 159 Australian adults, composed mostly of managers and HR professionals, read 12 vignettes involving requests for accommodations from existing employees. The requests differed in whether they were for psychological or physical disabilities. For each vignette, participants rated their empathy with the employee, the legitimacy of the employee's disability, the necessity for productivity, the perceived cost, and the reasonableness, and indicated whether they would grant the accommodation. RESULTS: Multilevel modeling indicated that greater empathy, legitimacy, and necessity, and lower perceived cost predicted perceptions of greater reasonableness and greater granting. Accommodation requests from employees with psychological disabilities were seen as less reasonable and were less likely to be granted; much of this effect seemed to be driven by perceptions that such accommodations were less necessary for productivity. Ratings on accommodations were influenced both by general between-person tendencies and within-person appraisals of particular scenarios. CONCLUSIONS: The study points to a need for organizations to more clearly establish guidelines for how decision-makers should fairly evaluate accommodation requests for employees with psychological disabilities and disability more broadly. (PsycINFO Database Record

Hierarchical Bayesian models of subtask learning.

The current study used Bayesian hierarchical methods to challenge and extend previous work on subtask learning consistency. A general model of individual-level subtask learning was proposed focusing on power and exponential functions with constraints to test for inconsistency. To study subtask learning, we developed a novel computer-based booking task, which logged participant actions, enabling measurement of strategy use and subtask performance. Model comparison was performed using deviance information criterion (DIC), posterior predictive checks, plots of model fits, and model recovery simulations. Results showed that although learning tended to be monotonically decreasing and decelerating, and approaching an asymptote for all subtasks, there was substantial inconsistency in learning curves both at the group- and individual-levels. This inconsistency was most apparent when constraining both the rate and the ratio of learning to asymptote to be equal across subtasks, thereby giving learning curves only 1 parameter for scaling. The inclusion of 6 strategy covariates provided improved prediction of subtask performance capturing different subtask learning processes and subtask trade-offs. In addition, strategy use partially explained the inconsistency in subtask learning. Overall, the model provided a more nuanced representation of how complex tasks can be decomposed in terms of simpler learning mechanisms.