Back

76 Views

Couples / Close Relationships

Observation of Affective Communication during Dyadic Interaction: Comparing Human versus Computer Coding

(PS10-A7) Observation of Affective Communication During Dyadic Interaction: Comparing Human versus Computer Coding

Saturday, November 19, 2022

3:00 PM – 4:00 PM EST

Location: Westside Ballroom, 5th Floor

Author(s)

AM

Alyssa J. Miville, B.A.

Clinical Science PhD Student
Binghamton University
Binghamton, New York

Co-Author(s)

RM

Richard E. Mattson, M.A., Ph.D.

Associate Professor of Psychology
Binghamton University
Binghamton, New York
Brandon E. Gibb, Ph.D.

Professor
Binghamton University
Binghamton, New York

Being able to effectively observe and anticipate the emotions and behaviors of others is a skill that is necessary for understanding and relating to peers, partners, and other loved ones. The importance of this research has been propelled by the social impacts of the pandemic over the course of the past two years. In the realm of couples’ research, literature suggests that individuals differ in how they communicate with their partners, and the expression of more (or less) positive or negative behaviors ultimately associates with different marital and relational outcomes. Several methods have been devised to tap into couples’ communication patterns, including observational coding systems like the Specific Affect (SPAFF) coding system, which has been hailed as the gold standard for observing affective communication in couples. However, there are limitations to behavioral coding systems, including human proneness to coder drift and error, as well as the time-consuming nature of training individuals to be proficient in coding. Identifying alternative, reliable, and potentially more valid ways of observing couples’ affective communication can thus bolster research in this area. A seemingly viable and comparable alternative to human-coded observational systems are computer programs that code for facial affect. We focused exclusively on the integrative analysis program known as iMotions, which, like SPAFF, is based on the Eckman coding scheme. The current project sought to compare the two methods of observing affect during couples’ interactions in both social support and conflict tasks. We processed archival data (N = 36 couples), which was previously SPAFF coded, in iMotions before running correlations between SPAFF and iMotions codes. Evidence for both convergent and discriminant validity was found. Analyses revealed several correlations between the two methods for specific codes, which varied in size (e.g., sadness and contempt/disgust displayed during husbands’ social support tasks ; r=.610*, p < .001 and r = .305, p = .075; respectively). However, we also found that some iMotions codes (e.g., contempt/disgust) were also correlated with other SPAFF codes (e.g., surprise/joy, r = .442, p = .008; fear, r = .433, p = .008), suggesting areas where iMotions and SPAFF coding both characterized particular behaviors as affective, but systematically classified them under different categories. We also evaluated criterion-related validity using known correlates of affective communication (e.g., relationship quality). These analyses revealed a range of associations between the SPAFF codes, whereas the iMotions variables were not appreciably associated with these criterion. Thus, despite potential advantages of computer programs, our results indicate that human coders, while more difficult to train, may be more valid observers of affective communication between couples, at least in comparison to ratings derived from iMotions, as other programs have been shown to exhibit better psychometrics when compared to SPAFF.