Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beyond Polarization: Opinion Mixing and Social Influence in Deliberation

Mohak Goyal, Lodewijk Gelauff, Naman Gupta, Ashish Goel, Kamesh Munagala

TL;DR

This paper introduces opinion mixing, quantified by Kendall's $\tau$, as a diagnostic complement to traditional polarization metrics in deliberation. By analyzing three large deliberative polling events (MCF-2022, MCF-2023, SOF) and linking transcripts to survey responses, the authors show that deliberation typically reduces rank-order preservation, i.e., increases mixing, while variance changes are heterogeneous and mean reversion is modest. A transcript-based mechanism analysis with LLM annotations reveals that expressed support and justification quality robustly predict opinion shifts, whereas novelty does not, suggesting that well-reasoned argument uptake drives reorganization. The findings imply that deliberation reshapes opinions through selective uptake of high-quality arguments, producing complex patterns of movement that standard polarization metrics may miss, with clear design implications for fostering reasoned discourse online.

Abstract

Deliberative processes are often discussed as increasing or decreasing polarization. This approach misses a different, and arguably more diagnostic, dimension of opinion change: whether deliberation reshuffles who agrees with whom, or simply moves everyone in parallel while preserving the pre-deliberation rank ordering. We introduce \opinion mixing, measured by Kendall's rank correlation (τ) between pre- and post-deliberation responses, as a complement to variance-based polarization metrics. Across two large online deliberative polls spanning 32 countries (MCF-2022: n=6,342; MCF-2023: n=1,529), deliberation increases opinion mixing relative to survey-only controls: treatment groups exhibit lower rank correlation on (97%) and (93%) of opinion questions, respectively. Polarization measures based on variance tell a more heterogeneous story: controls consistently converge, while treated groups sometimes converge and sometimes diverge depending on the issue. To probe mechanisms, we link transcripts and surveys in a third event (SOF: (n=617), 116 groups) and use LLM-assisted coding of 6,232 discussion statements. Expressed support in discussion statements strongly predicts subsequent group-level opinion shifts; this correlation is amplified by justification quality in the statements but not by argument novelty. To our knowledge, we are the first to observe how different notions of argument quality have different associations with the outcome of deliberation. This suggests that opinion change after deliberation is related to selective uptake of well-reasoned arguments, producing complex patterns of opinion reorganization that standard polarization metrics may miss.

Beyond Polarization: Opinion Mixing and Social Influence in Deliberation

TL;DR

This paper introduces opinion mixing, quantified by Kendall's , as a diagnostic complement to traditional polarization metrics in deliberation. By analyzing three large deliberative polling events (MCF-2022, MCF-2023, SOF) and linking transcripts to survey responses, the authors show that deliberation typically reduces rank-order preservation, i.e., increases mixing, while variance changes are heterogeneous and mean reversion is modest. A transcript-based mechanism analysis with LLM annotations reveals that expressed support and justification quality robustly predict opinion shifts, whereas novelty does not, suggesting that well-reasoned argument uptake drives reorganization. The findings imply that deliberation reshapes opinions through selective uptake of high-quality arguments, producing complex patterns of movement that standard polarization metrics may miss, with clear design implications for fostering reasoned discourse online.

Abstract

Deliberative processes are often discussed as increasing or decreasing polarization. This approach misses a different, and arguably more diagnostic, dimension of opinion change: whether deliberation reshuffles who agrees with whom, or simply moves everyone in parallel while preserving the pre-deliberation rank ordering. We introduce \opinion mixing, measured by Kendall's rank correlation (τ) between pre- and post-deliberation responses, as a complement to variance-based polarization metrics. Across two large online deliberative polls spanning 32 countries (MCF-2022: n=6,342; MCF-2023: n=1,529), deliberation increases opinion mixing relative to survey-only controls: treatment groups exhibit lower rank correlation on (97%) and (93%) of opinion questions, respectively. Polarization measures based on variance tell a more heterogeneous story: controls consistently converge, while treated groups sometimes converge and sometimes diverge depending on the issue. To probe mechanisms, we link transcripts and surveys in a third event (SOF: (n=617), 116 groups) and use LLM-assisted coding of 6,232 discussion statements. Expressed support in discussion statements strongly predicts subsequent group-level opinion shifts; this correlation is amplified by justification quality in the statements but not by argument novelty. To our knowledge, we are the first to observe how different notions of argument quality have different associations with the outcome of deliberation. This suggests that opinion change after deliberation is related to selective uptake of well-reasoned arguments, producing complex patterns of opinion reorganization that standard polarization metrics may miss.
Paper Structure (31 sections, 6 equations, 3 figures, 3 tables)

This paper contains 31 sections, 6 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Three Deliberation Events.
  3. Overview of Results
  4. Mechanisms of Opinion Change.
  5. Related Work
  6. Data and Deliberative Polling Setup
  7. Methodology
  8. Notation and Preliminaries
  9. Coding of Statements in Deliberation Transcripts
  10. Human Validation of Stance Annotations
  11. Linking Discussion Features to Opinion Change
  12. Results
  13. Characterizing Opinion Change
  14. Opinion Mixing: Widespread Reordering After Deliberation
  15. Variance Change: Heterogeneity Versus Uniformity
  16. ...and 16 more sections

Figures (3)

  • Figure 1: Treatment vs. control comparisons. Each point represents one item. (a) Kendall's $\tau$ between pre- and post-deliberation opinions; nearly all points lie below the diagonal, indicating greater opinion mixing in the treatment group. (b) Percentage change in opinion variance; control items are in a narrow band of mild decreases, while treatment items show heterogeneous effects spanning both convergence and divergence. (c) Mean reversion; more points lie above the diagonal, indicating modestly greater moderation of extreme opinions in the treatment group.
  • Figure 2: Surface plots illustrating the interaction between expressed support and justification quality on opinion change. The left panel includes all participants; the right panel restricts to participants who did not contribute to the discussion on that topic.
  • Figure 3: Distribution of the fraction of non-contributing participants across room-agenda pairs. The red dashed line indicates the mean fraction (0.284).

Theorems & Definitions (3)

  • Definition 1: Change in Disagreement
  • Definition 2: Stance Softening
  • Definition 3: Kendall's Rank Correlation ($\tau$) abdi2007kendall