Research

Amid rising global polarization, finding ways to disagree constructively is vital. This paper examines whether the medium of disagreement—spoken or written—shapes conversation outcomes. A series of randomized experiments (n = 1,576 conversation partners who had 1,842 conversations; n = 1,432 observers) suggests that spoken conversations with a disagreeing counterpart lead to greater understanding, lower conflict, more favorable impressions of one’s counterpart, and greater attitude alignment than written ones. Across experiments, speech also fostered more conversational receptiveness—cues in language that signal openness to opposing viewpoints—which partly mediates the effects of medium on these constructive disagreement outcomes. The conversation medium further moderated the association between language and outcomes: receptiveness was a stronger predictor of constructive disagreement in writing than in speech, suggesting people use less receptive language in the very medium in which it may be most effective (i.e., the written medium). A final study suggests that people may misjudge the effects of medium, wrongly believing spoken (versus written) disagreement will be less constructive and preferring to write to disagreeing counterparts. Despite people’s erroneous beliefs, spoken conversation offers a promising path to disagreeing constructively.

Organizations are filled with conversations, and how people treat each other can have a big impact on their performance, trust, learning, and satisfaction. Politeness is a critical dimension of those conversations. While most analyses of organizational text focus on content - the topics, ideas, and beliefs being talked about - previous work has suggested that how a topic is talked about can also have a large impact. In particular, elements of sentence structure—how a thought is phrased—can influence how a message is perceived. In this paper, we discuss and update the politeness R package, which applies linguistic theory to extract features of sentence structure from natural language text. While early models of politeness features have been discussed in computer science journals, we integrate that work into the study of organizations and introduce major improvements in the package. We also report new empirical findings showing that the current version of the package outperforms all previous versions (and other common benchmarks) for reproducing human annotations in new data.

Disagreement is often central to high-stakes problem-solving, yet managing the trade-off between asserting one’s position and accommodating others remains difficult. Using dyadic video-call data and post-call evaluations of outcomes from both high-stakes (N = 4,153 speaker turns) and low-stakes contexts (N = 63,416 speaker turns), we examine how critical moments during disagreement shape whether disagreements are experienced as constructive. Dyads that rated their disagreements as constructive exhibited critical moments characterized by question asking, positive framing, direct address, and inclusive language, typically preceded by clearer and less energetically forceful expressions from their partner. In contrast, unconstructive disagreements were marked by abrupt breaks from conversational norms, including sharp shifts toward self-focus behavior, including increased use of direct commands and negation. Together, these findings suggest that problem-solving success depends on engaging differences directly while maintaining mutual regard.

Research consistently shows a partisan well-being gap, with liberals reporting lower well-being than conservatives. We test whether differences in social media use explain this gap. To test this possibility, we test whether liberals’ lower wellbeing is mediated by their greater social media use. Students from UT Austin (N = 1,578) reported their political orientation, wellbeing (i.e., affective wellbeing, life satisfaction, loneliness), social media app use (hours and minutes on Facebook, Instagram, Twitter, TikTok, Snapchat, Reddit and YouTube), at monthly, weekly and momentary levels. Replicating prior research, we found that liberals had lower overall wellbeing, life satisfaction and felt lonelier. Critically, we also found that liberals used more social media (i.e., TikTok and Twitter) than conservatives, which mediated the relationship between liberalism and lower wellbeing at the weekly level. We plan to replicate these findings in a broader young adult sample to assess generalizability.