- When two people engaged in conversation consciously or unconsciously mirror each other’s behavior, we say that symmetry exists between them. If this behavior is measured, it may then be possible to quantify the amount of symmetry in this two-person system. Some methods for quantifying symmetry assume stationarity, and thus expect the behavior at any interval of each time series to be representable by a single set of parameters. This assumption is strong, especially given that the presence and amount of symmetry is expected to change from moment to moment. The Pairwise Approximate Spatiotemporal Symmetry (PASS) algorithm (Sjobeck, Boker, Scheidt, & Tschacher, under review) emphasizes the ever-evolving nature of symmetry by first segmenting pairwise time series into moments of symmetry and non-symmetry and then producing symmetry metrics which account for the segments. While most obviously appropriate for interpersonal data, like the movements of dyads in conversation, this algorithm can be used to measure the symmetry in intrapersonal time series, like signals from different regions of the brain. The PASS algorithm will be demonstrated here with both simulated data and the head movements of conversation partners. Research in group dynamics suggests that symmetry might also be present at the group level. This talk introduces efforts made to extend the PASS algorithm to circumstances with three or more coupled time series. The so-called Threeway Approximate Spatiotemporal Symmetry (TASS) algorithm segments three time series into moments of symmetry and non-symmetry. The threeway segmentation of the TASS algorithm can then be compared to the pairwise segmentation of the PASS algorithm, to understand how pairwise relationships contribute to the threeway relationships at the group level. The TASS algorithm will be demonstrated with simulated data and the fMRI signals of associated brain regions. These two algorithms represent efforts taken to understand the segmentability of symmetry relationships between coupled sources, and how these sources use symmetry to communicate affiliation and shared understanding through redundancy.
Time and Location:
12:30pm, Zoom
Date:
Thursday, November 5, 2020
Subtitle:
"Symmetry Segmentation of Bivariate and Trivariate Coupled Time Series." (Zoom link -- https://virginia.zoom.us/j/8368706265#success, Meeting ID: 836 870 6265)