Stress is an integral part of human existence. It is the primary survival mechanism with which nature endowed human physiology, and in this sense it is the prerequisite of everything else human. Stress is always present as a tonic component to sustain autonomic functions. What is of singular interest is the phasic component of stress, which represents responses to instantaneous or persistent external challenges. It is assumed that patterns of phasic stress responses deeply affect a person's performance and quality of life. However, such patterns have been inadequately quantified and studied up to now. In this project, the PI will explore a novel methodological framework for measuring and understanding the evolution of stress patterns in humans, and for determining their importance in phenotypes and performance. Physiological stress is effected through blood flow redistribution that results in localized thermoregulatory changes; these changes can be sensed by a thermal imaging sensor and analyzed within a computational framework. The PI argues that there is no better place to monitor stress signs than the face, because it is heavily innervated with sympathetic pathways, is amenable to non-contact (passive) sensing, and is open to cross-verification with traditional observational methods from psychology. He further contends that there is no better experimental framework to perform a stress study than when humans are engaged in challenging tasks with substantial stakes, as exemplified by the paradigm of surgical training. In this project, the PI will develop a novel integrated suite of sensing and computational methods to quantify stress. He will use these tools to correlate underlying physiological sensing with overt visual observation, and also to conduct longitudinal, quantifiable monitoring of realistic behavior with zero interference.
Broader Impacts: In the short term, the findings from the surgical training study will shed light on the role of stress in a very critical profession. In the long term, not only will the technology resulting from this research be directly applicable to diverse fields such as cyber therapy, medical training, medical imaging and cognitive research, but perhaps more importantly this work will likely deepen our understanding of human-computer interaction and create a fundamentally new interface dimension for human-centric computing.
Under a prior NSF award entitled "Interacting with Human Physiology" (2004-2008), we developed methods for unobtrusive monitoring of a subject's physiology, including adrenergic and cholinergic indicators of stress. All these methods were based on facial thermal imaging and computation instead of probes and electronics. They opened the way for quantitative studies of sympathetic responses in natural settings, such as the workplace. In this NSF award entitled "Do Nintendo Surgeons Defy Stress?" (2008-2011), after having refined and thoroughly validated the initial stress measurement methods, we applied them in a longitudinal field study. The study's objective was to uncover the role of stress on performance in a representative example of dexterous human-machine interaction. This example was training of surgeons on inanimate laparoscopic skill boxes. The study arrived at two far-reaching conclusions: (a) Humans, irrespectively of their skill level, attempt to perform dexterous tasks at the same high speed. In novices, high stress levels spawned by the task's challenge, drive this irrational strategy that results in horrendous error rates. In experienced individuals, where the skill has grown to match the neurophysiologically imposed speed level, the challenge has been met and consequently the stress levels have been substantially reduced. (b) There is a distinct correspondence of skill levels to stress levels in dexterous tasking. Hence, physiological indicators of stress can be used to assess competency. While conventional indicators of competency capture apparent performance (i.e., error rate), physiological indicators go beyond that by capturing internalization of skill. It appears that when dexterous interactions with machines are challenging, stress responses define human behavior, in essence interfering with the skill acquisition process. Moreover it seems that uninformed training practices make things worse, by scoring time performance in addition to error rate. This raises the challenge level and consequently elevates stress, reinforcing speedy responses that are mismatching to the trainee's rudimentary skills. The outcome is a positive feedback loop where more stress leads to faster responses and more errors, which in turn create more stress, until neurophysiological latency bottoms out. The trainee spends significant effort and emotional capital in order to break free of this vicious circle and master the art. By pointing to the detrimental role of the time stressor in skill training, the research suggests that by changing the training philosophy, trainees can attain competency in less time and at a minimal emotional burden. The importance of this in an era of skyrocketing educational costs cannot be overestimated. Furthermore, by crafting a grading scheme based on physiological indicators, the research solves a perennial problem in skill certification, that is, when the trainee’s competent performance is a matter of routine and not the product of a successful but agonizing effort.
