A sequential sampling account of response bias and speed–accuracy tradeoffs in a conflict detection task.

Abstract

Signal Detection Theory (SDT; Green & Swets, 1966) is a popular tool for understanding decision making. However, it does not account for the time taken to make a decision, nor why response bias might change over time. Sequential sampling models provide a way of accounting for speed-accuracy trade-offs and response bias shifts. In this study, we test the validity of a sequential sampling model of conflict detection in a simulated air traffic control task by assessing whether two of its key parameters respond to experimental manipulations in a theoretically consistent way. Through experimental instructions, we manipulated participants' response bias and the relative speed or accuracy of their responses. The sequential sampling model was able to replicate the trends in the conflict responses as well as response time across all conditions. Consistent with our predictions, manipulating response bias was associated primarily with changes in the model's Criterion parameter, whereas manipulating speed-accuracy instructions was associated with changes in the Threshold parameter. The success of the model in replicating the human data suggests we can use the parameters of the model to gain an insight into the underlying response bias and speed-accuracy preferences common to dynamic decision-making tasks.

Cognitive control and capacity for prospective memory in complex dynamic environments.

“…That nonexpert (novice) participants adopted a conflict bias in the current study without an explicit incentive to do so speaks to the face validity of the conflict detection task. That is, participants treat missing an aircraft conflict (i.e., possible passenger fatalities) as worse than making a false alarm (i.e., possibly increasing aircraft travel time; see Neal & Kwantes, 2009; Vuckovic et al, 2013, 2014). This finding supports the idea that people can make deliberate strategic adjustments to threshold bias to avoid failing to detect critical events (e.g., potential conflicts).…”

“…One question of interest is whether PMDC is capable of fitting the performance data from our conflict detection task. Evidence accumulation models have proved useful for understanding conflict detection in the past (Neal & Kwantes, 2009; Vuckovic, Kwantes, Humphreys, & Neal, 2014; Vuckovic, Kwantes, & Neal, 2013), accounting for accuracy and RT, and providing sensible psychological interpretations of the effects of manipulations of bias and speed-accuracy instructions. However, previous models required task-specific inputs such as relative speed and angle of approach, and thus have limited generalizability beyond the specific scenarios on which they are trained.…”

Equal evidence perceptual tasks suggest key role for interactive competition in decision-making

“…For example, a standard paradigm is to present stimuli with varying levels of support for one of multiple (e.g., two) alternatives (Erlick, 1961;Lee & Janke, 1964;Ratcliff, 2006;Ratcliff & Rouder, 1998;Swensson, 1972). Another benchmark involves the data that result from changes in the task instructions, such as emphasizing the speed or accuracy of the decision (Ratcliff & McKoon, 2008;Strayer & Kramer, 1994;Vickers, Burt, Smith, & Brown, 1985;Vickers & Smith, 1989;Vuckovic, Kwantes, Humphreys, & Neal, 2014;Wagenmakers, Ratcliff, Gomez, & McKoon, 2008;Wickelgren, 1977). Somewhat paradoxically, due to our consensus about experimental benchmarks, all serious theoretical contenders have been optimized to pass these benchmarks with ease, ultimately creating a theoretical stalemate until other experimental benchmarks are employed (Brown & Heathcote, 2008;Ratcliff, 1978;Ratcliff & Rouder, 1998;Tsetsos, Usher, & McClelland, 2011;Usher & McClelland, 2001).…”

An evidence accumulation model of perceptual discrimination with naturalistic stimuli.

“…Recent efforts have shown that accumulation models can account for behavior amidst additional task complexity. For example, this class of model has been applied to air traffic control conflict detection tasks (Loft, Bolland, Humphreys, & Neal, 2009;Vuckovic, Kwantes, Humphreys, & Neal, 2014), unmanned aerial vehicle simulation target detection tasks (Palada, Neal, Vuckovic, Martin, Samuels, & Heathcote, 2016), and medical image decision-making tasks (Trueblood et al, 2018). Using naturally varying stimuli (e.g., medical images) comprising covarying visual features, or more controlled stimuli (e.g., simulated targets) involving simpler decision rules, these prior studies have probed the generality of several different evidence accumulation models.…”