The Dissimilarity-Consensus Approach to Agreement Analysis in Gesture Elicitation Studies

We introduce the dissimilarity-consensus method, a new approach to computing objective measures of consensus between users' gesture preferences to support data analysis in end-user gesture elicitation studies. Our method models and quantifies the relationship between users' consensus over gesture articulation and numerical measures of gesture dissimilarity, e.g., Dynamic Time Warping or Hausdorff distances, by employing growth curves and logistic functions. We exemplify our method on 1,312 whole-body gestures elicited from 30 children, ages 3 to 6 years, and we report the first empirical results in the literature on the consensus between whole-body gestures produced by children this young. We provide C# and R software implementations of our method and make our gesture dataset publicly available.