Analyzing Web pages visual scanpaths: between and within tasks variability

Abstract

In this paper, we propose a new method for comparing scanpaths in a bottom-up approach, and a test of the scanpath theory. To do so, we conducted a laboratory experiment in which 113 participants were invited to accomplish a set of tasks on two different websites. For each site, they had to perform two tasks that had to be repeated ounce. The data were analyzed using a procedure similar to the one used by Duchowski et al. [8]. The first step was to automatically identify, then label, AOIs with the mean-shift clustering procedure [19]. Then, scanpaths were compared two by two with a modified version of the string-edit method, which take into account the order of AOIs visualizations [2]. Our results show that scanpaths variability between tasks but within participants seems to be lower than the variability within task for a given participant. In other words participants seem to be more coherent when they perform different tasks, than when they repeat the same tasks. In addition, participants view more of the same AOI when they perform a different task on the same Web page than when they repeated the same task. These results are quite different from what predicts the scanpath theory.

Comparing scanpaths during scene encoding and recognition: A multi-dimensional approach

“…This supports the original descriptions by Noton and Stark (1971), and extends the findings of with this dataset, but who did not fully compare similarity both within and between participants. Interestingly, the findings are also somewhat different from those reported recently by Drusch and Bastien (2012) in the context of webpage browsing. In that study, within-participant similarity was higher when a participants performed two different tasks than when they performed the same task twice.…”

“…Recent research has focused on deriving methods to quantify the similarity between pairs of scanpaths such as those observed by Noton and Stark (1971). Such methods are useful, both for basic research into perception and for applications such as the viewing of webpages (Drusch & Bastien, 2012;Josephson & Holmes, 2002). used two scanpath comparison algorithms to compare the eye movements made when trying to remember scenes (encoding) and when recognizing them later.…”

Using principal component analysis to characterize eye movement fixation patterns during face viewing

“…A number of data-driven methods to analyze eye movement data have been previously proposed. In particular, clustering is an important alternative to prespecified ROIs (Drusch, Bastien, & Paris, 2014; Göbel & Martin, 2018; Latimer, 1988; Naqshbandi, Gedeon, & Abdulla, 2017; Santella & DeCarlo, 2004). In general terms, these approaches use clustering algorithms, such as k -means to group spatially or temporally proximate fixations.…”

“…In general terms, these approaches use clustering algorithms, such as k -means to group spatially or temporally proximate fixations. Clustering techniques have been used to identify where individuals look when viewing web pages (Drusch et al, 2014) or scenes (Santella & DeCarlo, 2004) as well as to decode the task performed by participants (Naqshbandi et al, 2017). Clustering approaches require additional steps to characterize individual differences.…”

Lifelong Learning from Sustainable Education: An Analysis with Eye Tracking and Data Mining Techniques

“…One example of this type of measurement is the scan-path metric that describes both the spatial and the temporal sequence of the fixations that the participant has completed during the completion of a task. Those indicators are thought to represent evidence that the learner has performed task-resolution processes [4].…”