Intellectual productivity, the effective generation of cognitive artifacts, is crucial in collaborative settings like brainstorming sessions and hackathons. While various methods aim to enhance this productivity, optimal results aren't always achieved. Quantifying intellectual productivity is challenging, but this study proposes using nonverbal synchrony (NVS) – unconscious behavioral matching between individuals – as a potential indicator. Previous research demonstrated a link between NVS and creativity in dyadic interactions. This study extends this research by investigating NVS in larger groups (six participants) engaged in an idea generation task (alternative uses for a brick). The SenseChair, a non-invasive sensing device embedded in office chairs, was used to measure participants' seated swaying as a proxy for NVS. The researchers hypothesized that higher NVS scores would correlate positively with intellectual productivity, and compared the efficacy of global (all participants) versus pairwise (individual pairs) NVS measures.

Existing research highlights advancements in systems supporting group intellectual production, such as interfaces for visualizing meeting minutes and real-time thematic overviews of conversations. However, quantifying group intellectual productivity remains a challenge. Previous work identified a 'collective intelligence factor' correlated with social sensitivity, statement variability, and group composition (proportion of women). While useful, this factor doesn't directly measure the discussion's content. Nonverbal cues, including NVS, have emerged as potential indicators. Several studies have used cameras and microphones to assess NVS in conversation, but these methods can be intrusive and limit their application in settings with long periods of silence. The SenseChair offers a less obtrusive method to measure subtle body movements during group discussions, focusing on seated swaying.

The study employed data from a previous experiment involving 13 groups of six male participants (ages 20-40) engaged in a 15-minute alternative-use task ('uses for a brick'). Participants were seated in SenseChairs, allowing for non-invasive measurement of seated swaying via pressure sensors under the seat surface. The SenseChair measures center of gravity and weight changes, providing data for calculating NVS. Data was collected at 20 Hz. Time-series sway data was analyzed using a Hanning window and Fourier transformation to obtain time-series amplitude spectra, representing movement fluctuations. Data was normalized to account for individual differences in movement magnitude. Global NVS was calculated as the product of individual normalized spectral power across all six participants. Pairwise NVS was calculated for each of the 15 possible pairs within a group. Task performance was measured by the number of ideas generated, the number of idea categories, and a uniqueness score reflecting originality (higher scores for less common ideas). Spearman's rank correlation was used to assess the relationship between NVS (both global and pairwise mean and variance) and task performance. A resampling analysis was conducted to validate the results.

Analysis revealed no significant correlations between global NVS and task performance. This is likely due to the lower variability of the global NVS measure. However, a marginally significant positive correlation (p = 0.055) was found between the mean pairwise NVS and the number of ideas generated. This suggests that greater synchronization between pairs of participants leads to more ideas. The resampling analysis confirmed the significance of this finding. The probability of observing a correlation coefficient of 0.543 (original analysis result) between idea count and average pair NVS through random shuffling was less than or equal to 0.044, strengthening the evidence of a genuine association. Further analysis examining the rank order of pairwise NVS revealed a negative correlation between the highest-ranked pairwise NVS values and the uniqueness of ideas, and a weak positive correlation between the lowest-ranked pairwise NVS and the number of ideas. This suggests an optimal range for pairwise NVS where both idea generation and originality are maximized.

The findings support the hypothesis that NVS is related to intellectual productivity. The stronger correlation found for pairwise NVS compared to global NVS highlights the importance of considering local interactions within the group. The observation that extremely high pairwise NVS might negatively impact originality suggests that excessive synchronization, potentially reflecting non-task-related interactions, might hinder creative idea generation. Conversely, a minimum threshold of pairwise NVS might be necessary for effective idea generation, indicating the importance of some level of interactional alignment. The study emphasizes the need for more nuanced analyses considering the temporal dynamics and different levels of synchronization within groups.

This study demonstrates a marginally significant positive correlation between pairwise nonverbal synchrony and the number of ideas generated during a group discussion. The use of the SenseChair provides a novel, non-intrusive method for measuring subtle body movements that reflect group interaction dynamics. Future research should focus on larger sample sizes, longitudinal analysis of NVS across different phases of the discussion, and the incorporation of other nonverbal cues to refine the understanding of NVS's role in intellectual productivity. Exploring optimal levels of synchronization and incorporating individual personality traits in the analysis are also promising avenues for future investigation.

The study's limitations include the relatively small sample size, which could affect the generalizability of the results. The focus on male participants might limit the applicability of the findings to other demographics. The originality of ideas was assessed based on overlap with other teams' ideas, which is a limited metric for evaluating idea quality. Finally, the current analysis averaged NVS across the entire task duration, neglecting potential temporal variations in synchrony patterns during the discussion.