Modeling Performance at the Trial Level Within a Diffusion Framework
Abstract
<p>Human performance, when examined in a novel task context, typically shows evidence of efficiency gains over time. How is this efficiency achieved? One notion that seems very reasonable is the idea that participants dynamically alter their behaviour in response to errors (e.g., Brewer & Smith, 1984 ; Laming, 1979 ; Rabbitt, 1966a , 1978 ; Rabbitt, Cumming, & Vyas, 1978 ; Rabbitt & Rodgers, 1977 ). For such corrections to occur, the errors must first be noted, which can be challenging in contexts where no external feedback is provided (e.g., Rabbitt, 1966b ). In the present article, we focus on this issue and describe simple yet powerful error detection and correction processes that can be added to diffusion models of binary decision to enhance performance efficiency in a dynamic manner at the trial level.</p> <p>As psychological research focuses more on complex data patterns, there has been a corresponding increase in the reliance on computational theories over more traditional verbal theories. Computational models allow researchers to specify the basic processes that they view as relevant to performance, and then to explicitly show via simulation that the interaction of these processes can indeed produce data patterns consistent with those emitted