From Debate to Deliberation: Structured Collective Reasoning with Typed Epistemic Acts
arXiv:2603.11781v1 Announce Type: new 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 syste
arXiv:2603.11781v1 Announce Type: new 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.
Executive Summary
The article introduces Deliberative Collective Intelligence (DCI), a framework for structured collective reasoning with typed epistemic acts. DCI is evaluated on 45 tasks across seven domains, showing significant improvement over unstructured debate on non-routine tasks. The framework excels on hidden-profile tasks but fails on routine decisions, highlighting task-dependence. DCI produces structured decision packets and minority reports, but consumes more resources and is outperformed by single-agent generation on overall quality.
Key Points
- ▸ Introduction of Deliberative Collective Intelligence (DCI) framework
- ▸ Evaluation of DCI on 45 tasks across seven domains
- ▸ Significant improvement over unstructured debate on non-routine tasks
Merits
Structured Decision-Making
DCI provides a phased process for differentiated participants to exchange typed reasoning moves, preserving disagreements and converging on accountable outcomes.
Improved Performance on Non-Routine Tasks
DCI significantly outperforms unstructured debate on non-routine tasks, demonstrating its effectiveness in complex reasoning scenarios.
Demerits
Resource Intensity
DCI consumes significantly more resources than single-agent generation, which may limit its practical applicability.
Limited Performance on Routine Tasks
DCI fails to outperform single-agent generation on routine decisions, highlighting its task-dependent nature.
Expert Commentary
The introduction of DCI marks a significant advancement in the field of collective intelligence, as it provides a structured framework for complex decision-making scenarios. While DCI's resource intensity and limited performance on routine tasks are notable limitations, its ability to produce structured decision packets and minority reports is a substantial contribution to the field. As AI systems become increasingly prevalent in decision-making processes, the development of frameworks like DCI will be crucial in ensuring that these systems are transparent, accountable, and effective.
Recommendations
- ✓ Further research is needed to optimize DCI's resource consumption and improve its performance on routine tasks.
- ✓ The application of DCI should be explored in real-world decision-making scenarios to assess its practical effectiveness and identify areas for improvement.