Eye-Tracking Evidence for TACOM-Based Assessment of Procedural Task Complexity in a Nuclear Power Plant Full-Scope...

Emergency and operating procedures in nuclear power plants usually require operators to search for information, judge system states, and make control decisions across several linked interfaces. Conventional TACOM assessment quantifies the structural complexity of such procedure-guided tasks, but it does not directly show how operators visually process the task during execution. This study examined whether eye-tracking features can provide preliminary process-level evidence for TACOM-based assessment, rather than replace the TACOM framework. We extracted 25 eye-tracking features from 21 nuclear engineering graduate students while they completed 17 SGTR/SLOCA procedure fragments in an M310 full-scope simulator. A partial least-squares regression model predicted task-level TACOM scores, with RMSE = 0.357, MAE = 0.300, and R² = 0.538 under leave-one-task-out cross-validation. The RMSE corresponded to 18.3% of the observed TACOM range (2.070-4.017), and prediction was more strongly associated with observed task ranking (Spearman's rho = 0.775) than with exact linear calibration (Pearson's r = 0.750). Feature analyses suggested that fixation-duration variability, fixation dwell, pupil response, gaze dynamics, and spatial sampling jointly carried TACOM-related information. Exploratory subdimension analyses further indicated that task scope was most consistently associated with fixation dwell and fixation-time proportion, whereas task uncertainty was more closely associated with pupil variability and spatial entropy. These findings suggest that eye tracking may complement TACOM by describing execution-process demands, although the evidence remains correlational and limited by the small task set, graduate student sample, and task interface variability. Future studies should validate the signatures with licensed operators, larger multi-scenario task sets, independent TACOM scoring, and step-level AOI analyses.