Linear Ballistic Accumulator Modeling of Attentional Bias Modification Revealed Disturbed Evidence Accumulation of Negative Information by Explicit Instruction

摘要

In recent years, several attentional bias modification (ABM) studies have been conducted. Previous studies have suggested that explicit instruction (i.e., informing participants of the contingency of stimuli) enhances the effect of ABM. However, the specific working mechanism has not been identified. This is partly because reaction time (RT) data are typically reduced to an attention bias score, which is a mere difference of RT between experimental and control conditions. This data reduction causes a loss of information, as RT reflects various cognitive processes at play while making a response or decision. To overcome this issue, the present study applied linear ballistic accumulator (LBA) modeling to the outcomes (RT measures) of explicitly guided (compared to standard) ABM. This computational modeling approach allowed us to dissociate RTs into distinct components that can be relevant for attentional bias, such as efficiency of information processing or prior knowledge of the task; this provides an understanding of the mechanism of action underlying explicitly guided ABM. The analyzed data were RT-observed in the dot-probe task, which was administered before and after three-days of ABM training. Our main focus was on the changes in LBA components that would be induced by the training. Additionally, we analyzed in-session performances over the three days of training. The LBA analysis revealed a significant reduction in processing efficiency (i.e., drift rate) in the congruent condition, where the target probe is presented in the same location as a negative stimulus. This explains the reduction in the overall attentional bias score, suggesting that explicit ABM suppresses processing of negative stimuli. Moreover, the results suggest that explicitly guided ABM may influence prior knowledge of the target location in the training task and make participants prepared to respond to the task. These findings highlight the usefulness of LBA-based analysis to explore the underlying cognitive mechanisms in ABM, and indeed our analyses revealed the differences between the explicit and the standard ABM that could not be identified by traditional RT analysis or attentional bias scores.