“…Here, we use network theory methods to extract patterns of similar response of users to each particular profile of the Bot. Creating a correlation matrix is a first step to infer such information from multichannel data; it has been practiced in the analysis of large throughput gene expression experiments [36,37], classifying similar force-distance curves in repeated force-spectroscopy measurements of molecular forces [38,39], finding relevant correlations in fluctuations of different shares in stock market [40] and identifying patterns in traffic jamming [41,42]. Here, we construct Pearson's coefficient C i j of the valence time series between all pairs of users, i, j = 1, 2 The time series of valence {v i (t k )} for each user i = 1, 2, • • • 91 are available in the experimental dataset.…”