Comparison of Novice and Experienced Drivers Using the SEEV Model to Predict Attention Allocation at Intersections During Simulated Driving

Abstract

Summary:We compared the eye movements of novice drivers and experienced drivers while they drove a simulated driving scenario that included a number of intersections interspersed with stretches of straight road. The intersections included non-hazard events. Cassavaugh, Bos, McDonald, Gunaratne, & Backs (2013) attempted to model attention allocation of experienced drivers using the SEEV model. Here we compared two SEEV model fits between those experienced drivers and a sample of novice drivers. The first was a simplified model and the second was a more complex intersection model. The observed eye movement data was found to be a good fit to the simplified model for both experienced (R 2 = 0.88) and novice drivers (R 2 = 0.30). Like the previous results of the intersection model for the experienced drivers, the fit of the observed eye movement data to the intersection model for novice drivers was poor, and was no better than fitting the data to a randomized SEEV model. We concluded based on the simplified SEEV model, fixation count and fixation variance that experienced drivers were found to be more efficient at distributing their visual search compared to novice drivers.

What Attracts the Driver’s Eye? Attention as a Function of Task and Events

“…Their results showed a high correlation between predicted and observed percentage dwell time. Other studies also showed high correlations between SEEV-predictive and observed dwell time percentages in various types of human-machine interaction tasks [58][59][60][61][62][63].…”

Guiding Attention via a Cognitive Aid During a Simulated In-Hospital Cardiac Arrest Scenario: A Salience Effort Expectancy Value Model Analysis

“…To calculate the model fit, the predicted PDT is correlated with the empirically observed PDT that are collected using eye tracking (Wickens, 2015;Wickens et al, 2003). The model has been validated in different domains, such as aviation (model fit correlations >0.90; Steelman et al, 2011;Wickens et al, 2003;Wickens et al, 2008), driving (model fit correlations range ∼0.75-0.98; Bos et al, 2015;Cassavaugh et al, 2013;Horrey et al, 2006), scrub nursing during surgery (model fit correlations ∼0.70; Koh et al, 2011), and anesthesiology (model fit correlations rage ∼0.72 to ∼0.82; Grundgeiger et al, in press;Grundgeiger, Wurmb, et al, 2020). Wickens (2015) and Wickens et al (2008) argued that a model that only includes expectancy and value (the EV model version) can be used as a gold standard because only these top-down factors should guide the attention allocation of experts.…”

The Effect of Effort on the Allocation of Visual Attention During the Induction of General Anesthesia: A SEEV Model Analysis

“…The SEEV model has been validated in different supervisory control tasks such as flying (Steelman, McCarley, & Wickens, 2011;Wickens et al, 2003;Wickens et al, 2008), driving (Bos et al, 2015;Cassavaugh, Bos, McDonald, Gunaratne, & Backs, 2013;Horrey, Wickens, & Consalus, 2006), and scrub nursing during surgery (Koh, Park, Wickens, Teng Ong, & Noi Chia, 2011). Wickens (2015) describes a static version and a dynamic version for computing the model.…”