StayTeam: A Structured, Multi‑Agent System for Collaborative Learning and Decision‑Making, Problem‑Solving, Conflict Management, and Team Consensus‑Building

White Paper & Public Disclosure (v2.1)

Purpose: This public‑minimal version establishes a citable record of concepts while intentionally withholding implementation details.

Prepared: v1 on 6 June 2025; v2 on 5 September 2025; this v2.1 on 5 September 2025.

Status: Public, citable disclosure intended to establish prior art. Applications may be in preparation. This document does not grant any license and is not legal advice.

Author: Sabah Farshad, StayTeam.ai

Contact: sabah.fasrhad@Skolech.ru

Keywords: multi‑agent orchestration, collaborative decision‑making, problem‑solving, conflict management, consensus, stages and rounds, privacy‑by‑default, prior art.

Abstract

StayTeam guides groups from debate to decision using a multi‑agent orchestration designed for equitable decision-making and collaborative learning. In this consensus‑gated, stage‑and‑round workflow, participants draft privately to safeguard diverse perspectives; a neutral mediator synthesizes committed inputs alongside process-oriented feedback. Contribution visualization and rewarding mechanisms incentivize quality as a consent‑based vote advances the process when a configured consensus threshold is met. Privacy is enforced by a share‑to‑see model. Progress is coordinated by an explicit governance layer using deterministic state flags. This public version summarizes concepts, roles, and safeguards; implementation specifics are withheld.

1. Problem, Context, Outcomes

Conflict is normal in teams. Unmanaged, it derails projects; well‑channeled, it produces better ideas. StayTeam structures discussion so teams move from viewpoints → options → decisions with fairness, auditability, and momentum.

2. Core Innovations (Concept‑Level)

  1. Multi‑agent role separation. Personal drafting coach (private); neutral mediator (group synthesis); configurable policy layer (governance parameters).
  2. Consensus‑gated stages & rounds. Work proceeds through named stages with repeatable rounds; advancement requires a policy‑set consensus threshold; misalignment triggers another round.
  3. Privacy‑by‑default collaboration. Share‑to‑see hides others’ opinions until a participant contributes, reducing dominance/anchoring.
  4. Deterministic governance. An explicit, auditable state‑flag mechanism coordinates collection, synthesis, voting, and advancement.
  5. Real‑time collaboration. Live state across devices via modern real‑time channel technology, with resilience patterns.

3. High‑Level System Overview (Generalized)

  • Stages & rounds. Projects progress through a defined set of stages (e.g., problem framing, option generation, evaluation, decision, planning, review). Each stage may repeat rounds until the configured consensus threshold is met.
  • Round flow. Private drafting → share or skip → neutral synthesis → approve or request changes (consent‑based voting).
  • Governance & orchestration. A policy layer sets thresholds and rules; agents operate within those constraints; state flags ensure deterministic transitions without exposing drafts prematurely.

4. Data & Privacy (Generalized)

  • Records. The system maintains decision and discussion records per phase in a way that supports traceability without disclosing private drafts before contribution.
  • Privacy model. Share‑to‑see gating, role‑aware visibility, and auditability; privacy and security controls are applied commensurate with risk.
  • Providers. The Service may use major cloud and third‑party model providers; appropriate contractual and technical safeguards apply. Provider/region specifics are intentionally omitted from this public version.

5. Use Cases (Illustrative)

Teams under pressure or in conflict; stalled problem‑solving; facilitation at scale; committees and communities; friends, couples, and families (via tailored packs; non‑therapeutic; not legal advice).

6. Evaluation Philosophy (Non‑Exhaustive)

Time‑to‑decision, participation balance, perceived fairness, reliability, and (optionally) engagement metrics. Methods may include controlled studies and anonymized field telemetry.

7. Technology Variants (Non‑Limiting)

Functionally equivalent implementations may use different agent runtimes, workflow engines, model providers (cloud or local), consensus rules (e.g., approval/Condorcet/weighted/quorum), privacy controls (policy engines, commit‑reveal, client‑side encryption), real‑time transports, and data stores. Listing is illustrative and does not indicate preference.

8. Disclosure Boundaries (Public vs. Proprietary)

This public‑minimal version does not disclose: prompts/system messages; decision thresholds; retry/backoff and routing heuristics; internal schemas/migrations; names or sequences of governance states; provider/region specifics; detailed domain templates for personal/relationship use. These may be provided under NDA or in future filings.

9. Background & Research History (2019–2025)

This work builds on six years of research in collaborative engineering design, team cognition, and AI‑mediated feedback. Selected references are provided to ground the concepts without revealing implementation details.

  • Farshad, S., & Fortin, C. (2023). A Novel Method for Measuring, Visualizing, and Monitoring E‑Collaboration. International Journal of e‑Collaboration, 19(1), 1–21. https://doi.org/10.4018/IJeC.317223
  • Farshad, S., & Fortin, C. (2023). Active Engagement in Collaborative Engineering Design: How to Measure and Use It in a Feedback System? Proceedings of the Design Society: ICED23, 455–464. https://doi.org/10.1017/pds.2023.46
  • Farshad, S., Zorin, E., Amangeldiuly, N., & Fortin, C. (2023). Engagement assessment in project‑based education: a machine learning approach in team chat analysis. Education and Information Technologies. https://doi.org/10.1007/s10639-023-12381-5
  • Farshad, S., Brovar, Y., & Fortin, C. (2024). Enhancing Collaborative Design Through Process Feedback with Motivational Interviewing: Can AI Play a Role? IFIP AICT 702 (PLM 2023), 244–253. https://doi.org/10.1007/978-3-031-62582-4_22
  • Farshad, S., & Fortin, C. (2025). AI‑Driven Feedback for Improving Teamwork and Learning in Collaborative Engineering Design. Proceedings of the Design Society: ICED25. https://doi.org/10.1017/pds.2025.10060
  • Farshad, S., & Fortin, C. (2025). AI Versus Human‑Delivered Feedback in Project‑Based Education: Comparing Explicit and Implicit Attitudes in a Mixed‑Methods Study. Manuscript under consideration in Education and Information Technologies.
  • Farshad, S., & Fortin, C. (under review). How Teams Make Collaboration Work: Lessons from 1000+ Engineering Teams in Collaborative and Concurrent Design (1991–2021). Manuscript submitted to Research in Engineering Design.

Suggested Citation

StayTeam. (2025, September 5). A structured, multi‑agent system for collaborative decision‑making, problem‑solving, conflict management, and team consensus‑building (Public‑minimal v2.1). https://[your‑url]

← Back