The investigators were able to evaluate the high frequency interaction data streamed by the badges. They also had promising experimental evidence indicating that positive and negative affective reactions to specific team member interactions can be predicted from badge data. Specifically, interactions between team members that involved a cognitive stressor were associated with negative affect and performance failures for the stressed team member. HR predicted both affective states. However, negative affect was predicted by a significant interaction between HR and HRV. This was a very promising finding, because it indicated that the psychosocial status of team members can be inferred analytically from interaction metrics combined with HR and HRV data streams that are monitored by the badges.
This study had the following specific aims:
In the phase 4 validation experiment, prior phase 3 validation design was adapted and extended to evaluate the efficacy of the GSR sensing modality as part of the TIS sensor array. The effectiveness of the GSR sensor as a predictor of affective responses to stressed interactions was examined. In particular, the research was primarily focused on its ability to add to the prediction of affect beyond the HR and HRV metrics that were previously validated. The validation experiment was designed to create differential degrees of stress across conditions by using a strong CT (i.e., similar to phase 3 validation which yielded a 78% failure rate) versus a moderate CT (i.e., an easier CT with a target failure rate of approximately 50%). This allowed an evaluation of the sensitivity of the GSR sensor to different degrees of stress. In addition, interactions within conditions were differentially stressed (i.e., resource exchanges with CT versus without CT) which allowed an examination within individuals.
The final results showed that heart rate had more consistent relationships with the key study variables (PA, NA, cognitive testing) than GSR. GSR only related to NA, and its relationship was redundant with information conveyed by HR alone. As previously discussed, HR is indicative of both sympathetic and parasympathetic nervous system activity, whereas GSR is solely an indicator of the sympathetic nervous system. It is interesting and perhaps consistent with this conceptualization that GSR only related to NA; however, given the literature, It was anticipated that GSR would aid in the prediction of NA for this reason. Although HR may simply be a more effective physiological measure, there are alternative reasons why this result may have occurred.
The GSR sensors used may simply be less precise than the more established HR sensors that were utilized. It is also possible that GSR and HR may not be sufficiently “in phase.” That is, if galvanic skin responses occur more slowly than HR responses, the interaction-level analyses would not effectively capture the relevant information from the GSR metrics; recall the failure of GSR SD to relate to any of the key study variables. However, previous research has shown that GSR and HR responses tend to occur on the same timescale. It was also plausible that skewness within physiological data affected the analyses. To address that possibility, all the analyses were replicated using log-transformed data. The relationships between GSR / GSR SD and key study variables were unchanged.
On balance, It is concluded that GSR using the sensors employed in this research does not offer incremental validity for the TIS badge. Although there is sufficient ambiguity to note that these findings are not definitive, the research described in this report does not support a GSR sensor as a strong prospect for improving the TIS technology.
|Mission||Launch/Start Date||Landing/End Date||Duration|