Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Debate to Deliberation: Structured Collective Reasoning with Typed Epistemic Acts

Sunil Prakash

Abstract

Multi-agent LLM systems increasingly tackle complex reasoning, yet their interaction patterns remain limited to voting, unstructured debate, or pipeline orchestration. None model deliberation: a phased process where differentiated participants exchange typed reasoning moves, preserve disagreements, and converge on accountable outcomes. We introduce Deliberative Collective Intelligence (DCI), specifying four reasoning archetypes, 14 typed epistemic acts, a shared workspace, and DCI-CF, a convergent flow algorithm that guarantees termination with a structured decision packet containing the selected option, residual objections, minority report, and reopen conditions. We evaluate on 45 tasks across seven domains using Gemini 2.5 Flash. On non-routine tasks (n=40), DCI significantly improves over unstructured debate (+0.95, 95% CI [+0.41, +1.54]). DCI excels on hidden-profile tasks requiring perspective integration (9.56, highest of any system on any domain) while failing on routine decisions (5.39), confirming task-dependence. DCI produces 100% structured decision packets and 98% minority reports, artifacts absent from all baselines. However, DCI consumes ~62x single-agent tokens, and single-agent generation outperforms DCI on overall quality. DCI's contribution is not that more agents are better, but that consequential decisions benefit from deliberative structure when process accountability justifies the cost.

From Debate to Deliberation: Structured Collective Reasoning with Typed Epistemic Acts

Abstract

Multi-agent LLM systems increasingly tackle complex reasoning, yet their interaction patterns remain limited to voting, unstructured debate, or pipeline orchestration. None model deliberation: a phased process where differentiated participants exchange typed reasoning moves, preserve disagreements, and converge on accountable outcomes. We introduce Deliberative Collective Intelligence (DCI), specifying four reasoning archetypes, 14 typed epistemic acts, a shared workspace, and DCI-CF, a convergent flow algorithm that guarantees termination with a structured decision packet containing the selected option, residual objections, minority report, and reopen conditions. We evaluate on 45 tasks across seven domains using Gemini 2.5 Flash. On non-routine tasks (n=40), DCI significantly improves over unstructured debate (+0.95, 95% CI [+0.41, +1.54]). DCI excels on hidden-profile tasks requiring perspective integration (9.56, highest of any system on any domain) while failing on routine decisions (5.39), confirming task-dependence. DCI produces 100% structured decision packets and 98% minority reports, artifacts absent from all baselines. However, DCI consumes ~62x single-agent tokens, and single-agent generation outperforms DCI on overall quality. DCI's contribution is not that more agents are better, but that consequential decisions benefit from deliberative structure when process accountability justifies the cost.
Paper Structure (74 sections, 2 theorems, 2 equations, 2 figures, 9 tables)

This paper contains 74 sections, 2 theorems, 2 equations, 2 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Contributions.
  3. What is New in DCI?
  4. Background and Related Work
  5. Multi-Agent Debate
  6. Multi-Agent Frameworks
  7. Ensemble Methods and Self-Consistency
  8. Deliberative Democracy and Speech Act Theory
  9. Social Choice and Arrow's Impossibility
  10. Positioning
  11. The DCI Framework
  12. Delegate Model
  13. Session Model
  14. Interaction Grammar
  15. Design Rationale for the 14 Acts
  16. ...and 59 more sections

Key Result

Theorem 1

For any DCI session with finite delegate set $|\Delta| = n$, maximum rounds $R_{\max}$, maximum options $K_{\max}$, and maximum recursion depth $D_{\max}$, DCI-CF terminates in at most rounds, where $B_d$ is the maximum number of sessions at depth $d$.

Figures (2)

  • Figure 1: DCI-CF algorithm flow. Stages 3--6 form a loop bounded by max_rounds. If convergence fails after round exhaustion, Stage 7 provides a deterministic fallback. Every path terminates at Stage 8 with a structured decision packet.
  • Figure 2: Condensed DCI session on an architectural design task. Four archetypes exchange typed epistemic acts across two phases, surfacing tensions that are preserved as first-class objects. The session converges on a pragmatic reframing while retaining the minority position and specifying reopen conditions---artifacts absent from single-agent or debate outputs.

Theorems & Definitions (6)

  • Definition 1: Decision Packet
  • Theorem 1: Termination
  • proof : Proof sketch
  • Remark 1
  • Theorem 2: Termination of DCI-CF
  • proof