This article investigates the causal antecedents of conflictual language and the geometry of interaction in online threaded conversations related to climate change. We employ three annotation dimensions, inferred through LLM prompting and averaging, to capture complementary aspects of discursive conflict (such as stance: agreement vs disagreement; tone: attacking vs respectful; and emotional versus factual framing) and use data from a threaded online forum to examine how these dimensions respond to temporal, conversational, and arborescent structural features of discussions. We show that, as suggested by the literature, longer delays between successive posts in a thread are associated with replies that are, on average, more respectful, whereas longer delays relative to the parent post are associated with slightly less disagreement but more emotional (less factual) language. Second, we characterize alignment with the local conversational environment and find strong convergence both toward the average stance, tone and emotional framing of older sibling posts replying to the same parent and toward those of the parent post itself, with parent post effects generally stronger than sibling effects. We further show that early branch-level responses condition these alignment dynamics, such that parent-child stance alignment is amplified or attenuated depending on whether a branch is initiated in agreement or disagreement with the discussion's root message. These influences are largely additive for civility-related dimensions (attacking vs respectful, disagree vs agree), whereas for emotional versus factual framing there is a significant interaction: alignment with the parent's emotionality is amplified when older siblings are similarly aligned.
Online platforms have become central arenas for public debate on politically salient and scientifically complex issues, including climate change. While these environments promise broader participation and exposure to diverse viewpoints, a growing body of research shows that they are also prone to conflictual dynamics that can undermine deliberation, discourage participation, and distort the circulation of information and ideas (Bennett & Pfetsch, 2018;Baughan et al., 2021;Altenburger et al., 2024). Understanding how such conflict emerges, evolves, and propagates within online discussions is therefore critical for assessing the democratic implications of social media-mediated communication.
Existing research has shown that conflict in online conversations is not solely a function of issue-based disagreement, but is deeply shaped by how disagreement is expressed. More specifically, linguistic and pragmatic studies distinguish between issue-based disagreement and interpersonal conflict, emphasizing the role of tone, politeness, emotional framing, and respectfulness in determining whether exchanges remain productive or escalate into hostility (Brown & Levinson, 1987;Somasundaran & Wiebe, 2009;Danescu-Niculescu-Mizil et al., 2013). Computational analyses further demonstrate that conversational conflict is dynamic, often evolving from rational argumentation into personal attack through identifiable shifts in sentiment, emotionality, and rhetorical style (Habernal et al., 2018;Wang & Cardie, 2016;Zhang et al., 2018).
More recent work has moved beyond static classification of conflictual language to investigate its temporal and interactional antecedents. Studies of conversational derailment, trolling, and toxic escalation highlight the importance of situational context, interaction structure, and timing, showing that exposure to prior conflict, irregular conversational rhythms, and early interactional cues significantly shape subsequent behavior (Cheng et al., 2017;Roos et al., 2020;D’Costa et al., 2024). These findings suggest that conflict cannot be fully understood by treating messages as independent observations, but instead emerges from structured processes unfolding over time and across interactional positions.
At the same time, much of the existing literature focuses either on pairwise interactions or on aggregate conversational outcomes, leaving the role of local conversational structure underexplored. Threaded discussions are inherently hierarchical: replies are embedded within branching trees, where messages respond not only to a single parent post but also to an evolving local context shaped by earlier sibling replies and by the initial tone of branch-initiating messages. Prior research indicates that early exchanges can establish interactional norms that condition later participation and tone, but systematic evidence on how such norms propagate within discussion branches remains limited (Danescu-Niculescu-Mizil et al., 2013;Zhang et al., 2018).
In this paper, we address this gap by analyzing the geometry of online conversations, which we define as the temporal, sequential, and structural organization of threaded discussions, and its relationship to conflictual language. Using climate change-related discussions from the Internet Argument Corpus, we study how stance (agreement vs disagreement) tone (attacking vs respectful) and emotional versus factual message framing respond to (i) conversational timing, (ii) alignment with parent and sibling posts, and (iii) early branch-level signals. By combining LLM-based annotations with regression models that explicitly account for discussion structure, we provide a compact account of how conflictual discourse emerges and propagates within online conversations.
Our contribution is threefold. First, we show how reply timing and conversational rhythm are systematically associated with shifts in civility, disagreement, and emotional framing. Second, we document strong alignment effects with both parent posts and local sibling contexts, highlighting the importance of meso-level conversational environments. Third, we demonstrate that early branch-level responses moderate subsequent stance alignment, revealing how initial interactional cues shape the trajectory of agreement and disagreement in discussion branches. Together, these findings advance our understanding of conflictual language as a dynamic, structured process and offer new insights into the micro-level and meso-level conversational mechanisms through which online discourse may shape democratic debate.
The remainder of the article is structured as follows. In Section 2 we review existing work on conflictual language, conversational dynamics, and the role of temporal and structural factors in online discussions. In the Methods and Data section (Section 3) we explain the construction of the climate change discussion corpus, the LLM-based annotation procedure, and the representation of discussions as reply t
This content is AI-processed based on open access ArXiv data.