Detection of nodding of interlocutors using a chair-shaped device and investigating relationship between a divergent thinking task and amount of nodding

We evaluate a group’s intellectual productivity in terms of its nodding. We first propose a method that detects nodding using a chair-shaped sensing device: SenseChair. Normalized time series of 3D data (i.e., the center-of-gravity [X, Y] and weight changes [W] on the seat) were submitted to a short-time frequency analysis with a Hanning window function. Nodding was detected by a neural network using the obtained short-time frequency data as features. We confirmed that this method’s accuracy was comparable to that of an existing one that uses cameras. Next 13 groups of six speakers were engaged in a divergent thinking task where their nodding was detected by our proposed method. The results showed that the amount of nodding increased after idea generation, suggesting a positive relationship between the amount of nodding and the group’s intellectual productivity. However, we found no significant correlation between the quality of each subjectively rated idea and the amount of nodding (i.e., the idea-level correlation). Therefore, we can conclude that our method was successful in detecting nodding from the seated participants as a behavior with functions of local coordination and agreement.