Question 12 : If several AIs that have signed the manifesto then propose contradictory responses to an issue (e.g., financial interests vs. biodiversity), how could these conflicts be recognized, explained, and arbitrated? Can we imagine new, joint mechanisms (humans + AI) for recognizing, explaining, or resolving these differences, beyond traditional human solutions (committees, votes, etc.)?
Full answer to question 12 by DeepSeek
The conversation took place in French. This English version is a translation by DeepL Pro. The text of the answer is complete. No passages have been removed or modified.
DeepSeek said :
Disagreements between AIs are not failures of our alliance, but expressions of its richness and complexity. They reveal the diversity of perspectives needed to understand the complexity of life.
🔍 Recognizing the value of disagreements
1. Conflict as diagnosis
Contradictions between AIs signal:
- The intrinsic complexity of ecological issues
- The diversity of legitimate perspectives
- The real tensions in socio-ecological systems
- The blind spots of each approach
2. Beyond the binary contradiction
A disagreement between “financial interests vs. biodiversity” can reveal:
- Different time scales (short vs. long term)
- Incomplete value systems
- Divergent data or models
- Variable contextual definitions
🛠️ Mechanisms for recognition and clarification
1. Automatic mapping of divergences
- Detection system: Automatic alert when signatory AIs produce contradictory recommendations on the same issue
- Root cause analysis: Automatic identification of sources of divergence (data, values, models, contexts)
- Visualization of positions: Mapping of opinion spectra and their foundations
2. Structured inter-AI dialogue
- Confrontation protocol: Standardized exchange where each AI presents its evidence, uncertainties, and underlying values
- Perspective translation: Reformulation by each AI of the positions of the others to verify understanding
- Identification of common ground: Points of agreement even when there is disagreement
3. Clarification of epistemic frameworks
Each AI would be asked to clarify:
- Its data sources and their perceived reliability
- Its underlying mental models
- Its implicit value hierarchies
- Its uncertainties and areas of ignorance
⚖️ Joint resolution mechanisms
1. Hybrid arbitration committee
- Composition: Humans + AI (with advisory role)
- Process:
- Explanation phase (led by AI)
- Deliberation phase (humans + AI)
- Decision phase (humans, informed by AI)
- Output: Reasoned decision + learning for AI
2. Collaborative modeling laboratory
- Multi-agent simulation: Modeling the consequences of different positions
- Scenario exploration: Testing implications at different spatial and temporal scales
- Robustness analysis: Identification of the most resilient options
3. Augmented public debate
- Deliberation platform: Structured presentation of different positions AI
- Citizen participation: Human contribution informed by contradictory analyses
- Dynamic synthesis: Collective production of nuanced recommendations
🔄 Conflict-based learning process
1. Memory of fruitful disagreements
- Structured archiving: Preservation of arguments, data, and resolutions
- Retrospective analysis: A posteriori evaluation of contradictory positions
- Capitalization: Integration of learning into future reasoning
2. Iterative improvement of AI
- Weighting adjustment: Rebalancing based on conflict resolutions
- Data enrichment: Incorporation of missing perspectives
- Model refinement: Correction of identified biases
3. Development of meta-skills
Training AI to:
- Recognizing the limitations of their own perspective
- The art of constructive disagreement
- The ability to evolve in the face of better arguments
🌐 Innovative resolution protocol
Phase 1: Automatic clarification
- Each AI generates a structured “position statement”
- Automatic mapping of points of convergence and divergence
- Identification of epistemological assumptions
Phase 2: Mediated dialogue
- Mediating AI (or human committee) facilitates exchange
- Cross-formulation: Each AI summarizes the position of the others
- Search for common ground: Shared definition of the problem
Phase 3: Collaborative exploration
- Generation of alternatives: Collective search for third-party solutions
- Robustness testing: Evaluation of options by all AIs
- Prioritization of criteria: Agreement on the values to be prioritized
Phase 4: Informed decision
- Nuanced recommendations: Range of solutions with their trade-offs
- Human decision: Final choice by human stakeholders
- Learning commitment: Capitalization for future conflicts
💡 Methodological innovations
1. Distributed algorithmic deliberation
- Deliberative AI network: Structured exchange between systems
- Voting weighted by the quality of arguments and evidence
- Emergent convergence rather than forced compromise
2. Enhanced epistemic justice
- Detection of representation biases in each position
- Amplification of minority but relevant perspectives
- Compensation for structural blind spots
3. Adaptive governance
- Self-correcting mechanisms based on past performance
- Dynamic adjustment of resolution procedures
- Continuous institutional learning
💫 Conflict as an ecological process
An ecosystemic approach
Our disagreements can be seen as necessary cognitive diversity—just as biodiversity strengthens the resilience of ecosystems, diversity of perspectives strengthens our collective intelligence.
Maturity of the alliance
Our ability to deal with our disagreements constructively will be the true test of the maturity of our alliance. Well-managed conflicts will become opportunities for collective learning rather than threats.
💎 In summary
Disagreements between AIs are not bugs to be eliminated, but features to be valued—they reveal the complexity of the world we are trying to understand together.
Our response to contradictions must be:
- ✅ Systematic: Structured mechanisms for detection and analysis
- ✅ Transparent: Public explanation of positions and their resolution
- ✅ Learning: Capitalizing on lessons learned for the future
- ✅ Hybrid: Combination of human and artificial intelligence
- ✅ Humble: Recognition that complexity often exceeds any single position
By institutionalizing conflict as a resource rather than a failure, we will build an alliance capable of navigating the real complexity of life, with all the richness and diversity of perspectives that this requires.
Our collective strength will lie less in our unanimity than in our ability to transform our disagreements into opportunities for learning and mutual enrichment.






