< Back to previous page

Publication

The impact of expertise on gaze behavior in perceptual-cognitive tasks

Book - Dissertation

In many professional domains, errors in decision making resulting from insufficient attention to visual cues can lead to disastrous consequences. For example, if a radiologist fails to detect a tumor, the patient would not receive proper treatment on time, or if a pilot does not detect failures in the cockpit instruments, the airplane might crash. It is thus important to obtain knowledge about the origin of these errors and adapt training protocols accordingly. Over the past decade, perceptual-cognitive skills have been highlighted of utmost important for successful decision making in most professional domains. Indeed, perceptual failures are one the biggest causes for airplane accidents or misdiagnoses in radiology. In order to enhance and train perceptual-cognitive skills in specific professional domains, clear insights in underlying processes of such skills is essential. A broad systematic literature search, reported in Chapter 2, has shown a clear link between visual search behavior and perceptual-cognitive skills. Bruno, Walker and Abujudeh (2015) also reported a relation between inefficient visual search and perceptual errors. Furthermore, specific visual search behavior, measured by eye-movement characteristics, might explain underlying processes that result in superior task performance. These processes are identified according to three theories: (i) The long-term working memory theory, proposing that with experience more information is available for rapid extraction in the expert's long-term working memory; (ii) The information-reduction hypothesis, proposing that with experience attention is efficiently directed to relevant information and irrelevant information is ignored; (iii) The holistic model of image perception, proposing that visual span extends with experience, which provides the opportunity to first capture the global picture and direct search efficiently afterwards. Evidence for the last two theories can also be obtained with a measure of domain-generic skills. In this PhD project three generic tests were used: (i) the Leuven Embedded Figures test (L-EFT), i.e. a test in which figures have to be differentiated from a noisy background (second theory), (ii) the Coherent motion task, i.e. a task in which a coherent motion direction of moving dots has to be identified on a chaotic moving background (second theory), and (iii) the Navon selective attention task, i.e. a task analyzes participants' capability to detect figures either at a global (large figure) or local level (smaller figure) (third theory). However, results obtained in this PhD project only provided evidence for a relation between performance, experience and holistic domain-generic skills (Navon selective attention task) (Chapter 3 and 4). These results suggest that experts indeed use a global-local information processing strategy, as described by the holistic model of image perception, in both the specific task as well as in other generic settings. In this PhD project, two expertise domains were studied in more detail, aviation and radiology. The reasoning for selecting these two domains was threefold: (i) the different types of tasks completed in these domains; (ii) the importance of "efficient" visual search behavior for successful task completion in both domains; and (iii) difference in information provided during training about the importance of "efficient" visual search in these domains. Specifically, pilots mainly complete dynamic instrument monitoring tasks. In cases of instrument failures, fast and accurate responses are essential. In contrary, on a daily base, radiologists complete static medical image interpretation tasks, with almost no time pressure. While completing medical image interpretation tasks, high accuracy levels are utmost important. In literature, a relation between experience, visual search behavior and high task performance was extensively described in both domains. However, only during pilots' education visual search strategies are trained. In contrary, radiologists never received such information during training. Complete eye-tracking datasets describing visual search in both domains have shown that all three theories, described above, operate complementary as a result of experience. Also, a relationship between specific visual search behavior as described by the theories and enhanced specific task performance was observed (Chapter 3 and 4). For this reason, we propose a unified theory of superior perceptual-cognitive skills that evolve with experience: the long-term working memory - guided search theory (Figure 2, Discussion and Conclusion). This theory proposes that with experience an enormous amount of information becomes available in long-term memory. Information from long-term memory is then activated in working memory, which is then becomes easily accessible if required for task completion. This information will help the expert to efficiently direct his or her attention towards essential information. The expert can differentiate this information from irrelevant information, because of previous stored knowledge about relevant and irrelevant information in long-term memory. In addition, the expert will be able to capture and process the global picture with one short first glimpse. This short glimpse is sufficient for global information processing, due to rapid information availability in working memory. Finally, the capability to globally process the scene is assumed to result in an efficient local search. This local search results in superior perceptual-cognitive skills and, consequently, improved decision making. Chapter 2, 3 and 4 describe the importance of visual search behavior and its link with expertise. Therefore, a training intervention based on an expert's visual search behavior, as eye movement modelling example (EMME), was tested (Chapter 6). Specifically, during training participants watched video clips of an expert radiologist's EMME while completing the same task, i.e. a focal lung pathology-detection task on chest X-rays. Video clips were shown with either a blurred background and bright moving circles resembling the expert's fixations (spotlight group) or with a bright background and moving circles representing the fixations (circle group). All groups, including a control group, received feedback about the correct location of the pathology. Evidence was observed in favor of the circle training. Participants from the circle group were better in detecting pathologies during a transfer-test. Furthermore, measures of sensitivity and specificity indicated that all participants benefited from just receiving information about the correct location of the pathology. This observation suggests that participants developed an enhanced pattern recognition skill due to repetitive exposure to patterns of chest X-rays and pathology. No differences in visual search strategy were observed between groups, suggesting that visual search most likely evolves as an implicit automatic process, as a consequence of enhanced pattern recognition skills. Interestingly, this skill was already obtained after a very short intervention, which might suggest that this is one of the first perceptual-cognitive skills that evolve with experience (Chapter 6). Because a link between enhanced pattern recognition and visual search behavior is observed, it is interesting to monitor this behavior while executing professional tasks or during training. In Chapter 5, a method providing scores of visual search behavior in the cockpit related to specific flight maneuvers is developed. This method can be used as the baseline for further evolution of human-machine interaction monitoring systems. To conclude, the knowledge collected in this project might provide important support for instructors or practitioners in professional domains which require an extensive use of vision such as aviation and radiology. Whether our findings can be generalized to other professional settings where time-constraint decision making is 'key' for successful performance, remains to be seen. Implementing this knowledge in training programs and monitor visual search behavior is expected to reduce human-related errors in such professional domains significantly, which would eventually enhance human healthcare and flight safety.
Publication year:2020
Accessibility:Closed