Scan pattern similarity in repetitive user interface examination

Sequential visual information sampling helps understand human information accumulation strategies in context of optimal decision making. In this study we investigated whether participants develop systematic sequential scanning behaviour when repeatedly looking at multiple information sources in a scenario resembling a semi-automated control room. Ten participants engaged in a road traffic management task. They were asked to make a series of binary decisions as accurately and as quickly as possible. In order to make the decisions, four information sources were provided via sub-windows in a user interface. These sources contained conflicting information for 75% of the total 32 trials. Eye movements were recorded at 30 Hz using Tobii Glasses v1 and mapped to the four regions of interest (ROIs). Similarity in scan patterns across trials and/or participants was calculated in Matlab using string-edit/bioinformatics methods (Levenshtein distance, sequence alignment, kmer detection) and network analysis. Results for the different analysis methods showed the general trend for scan patterns to not become more similar with elapsed trial. Scan pattern similarity also remained low when accounting for the four experimental conditions. Study results hence suggest that participants did not converge on a specific visual scanning strategy; this finding may change in the presence of feedback.