Team Situation Awareness, Cohesion, and Autonomic Synchrony 2: Group-level Effects and their Combined Influence on Team Performance

• Speaker: Stephen Guastello, Marquette University
• Abstract: Situation awareness (SA) is a mental state that is instrumental to performance of complex dynamic tasks. It consists of three stages 'knowing the facts about a situation that could be changing rapidly, formulating a mental model of how the situation and facts are operating, and anticipating correctly the outcomes of further action'. The development of SA within groups and teams is thought to be supported by favorable social conditions within the team. Thus the present study followed from earlier work on causal relationships among SA, group cohesion, and autonomic synchrony, the latter being a fundamentally nonlinear process. The present study assessed the combined impact of the three variables on performance in a dynamic decision task. Experimental manipulations were changes in task difficulty, group size, and method of obtaining SA measures. Participants 32 teams of 3-5 undergraduates performing a dynamic decision task that consisted of two matches of a first-person shooter computer game. They also completed self-report measures of cohesion and SA. Synchrony was determined through time series analysis of electrodermal responses using the driverempath framework. ANOVA results showed that cohesion and SA improved over the two matches, and SA was better in smaller groups. Synchrony was stronger in larger groups. Granger regression indicated no causal or reciprocal relationship between SA and cohesion. Synchrony had a small positive effect on cohesion during the first match, which dissipated afterwards. SA had a strong negative impact on synchrony early on, which also dissipated afterwards. The best performing teams were the larger groups, those that were measured for SA without pausing the simulation, were less synchronized, showed better SA, and reported stronger cohesion. The study opens new questions concerning the role of synchrony in situations requiring different amounts of task versus social focus, threat levels imposed by gaming opponents, and role of automated teams operating alongside the humans. Joint work with Laura McGuigan, Henry Vandevelde, Ryan Hagan, Cooper Bednarczyk, and Anthony Peressini, Marquette University.

Identifying Mood Symptom Motifs in Time Series using Dynamic Time Warping

• Speaker: David Kreindler, University of Toronto
• Abstract: Background: If patterns exist in mood dynamics, this would be consequential for the prediction of mood disorders. Dynamic Time Warping (DTW) is a method that enables identification of 'fuzzy' (approximate) matches in time series data. DTW was used to explore whether repeated conserved sub-sequences ('motifs') could be identified in extended, high-accuracy mood symptom time series. Methods: Using a database of self-report questionnaires rating the severity of 11 mood symptoms that were collected every 12h over 1.5 years from 19 subjects with rapidly-cycling bipolar disorder and 19 age- and sex-matched healthy controls, DTW was used to calculate measures of similarity between normalized sub-sequences of 14 or 28 points (i.e., one or two weeks' duration) within individual symptom severity time series. Receiver Operating Characteristic (ROC) curves were created on a subset of the data to determine optimal similarity measure cut-off values, enabling automated motif identification across the entire data set. Results: One or more motifs were identified in most but not all symptom severity time series. However, regardless of sub-sequence length or change to the ' warping window' size, the number of motifs identified from the time series was not significant compared to the motif counts from randomly permuted versions of the original time series. Conclusions: While motif analysis of normalized mood symptom time series using DTW does identify motifs pairs and some motif clusters, the results obtained using randomly permuted time series suggest that, in this analysis, our DTW-based fuzzy matching method may be identifying matches where none in fact exist.

Catastrophe Modelling for Time Series of Reported Cases of COVID-19: Workload Effects in the Health Care System

• Speaker: Stephen Guastello, Marquette University
• Abstract: This project began with the observation of time series trends in the number of COVID-19 positive cases that were reported in each of the 50 US States. At one level of observation there was a weekly cycle of high and low numbers of cases. At the second level that spanned there were visible waves of higher and lower numbers of reports. At the third level, there was a distinctive onset of variability during the peak times of positive reports of cases. The present study examined the possibility that the variability during peak periods was a nonlinear dynamical footprint of cognitive workload in the health care system. Our analysis of time series data from the state of Arizona from March 2020 to October 2021 showed that the swallowtail catastrophe model accounted for the occurrence of waves, and the cusp catastrophe model applied to the residual variance accounted for variability at the peaks. The two functions accounted for 97% of the variance in positive case reports over time. Implications for the management of future pandemics are discussed. Joint work with Noah Osowski, Marquette University.