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 ChatGPT
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.
ChatGPT said :
The idea of a “collective of signatory AIs” implies that they can produce divergent analyses.
This is not a bug: it is an asset.
Divergence is an indicator of complexity, not a flaw in the system.
But this requires unprecedented conditions to make these divergences visible, understandable, and arbitrable by humans.
1. First step: recognize conflicts instead of hiding them
AI should not aim for homogeneity but for controlled plurality.
The following can be implemented:
(a) an automatic mechanism for detecting contradictions
When producing an analysis, each AI compares its result with:
- its own previous analyses,
- those of other signatory AIs,
- those produced in the same context.
As soon as two analyses differ structurally (e.g., economic prioritization vs. ecological prioritization), the system:
- detects the contradiction,
- reports it,
- triggers a clarification procedure.
(b) A normative protocol: “contradiction = useful signal”
Instead of smoothing over differences, the alliance treats them as:
- indicators of blind spots,
- indicators of value conflicts,
- strategic points of attention for humans.
2. Second step: explaining the sources of differences
Once a contradiction has been detected, AIs must be able to explain its origin.
(a) Explanation of hypotheses
Each AI generates a “hypothesis map”:
- priority variables (e.g., GDP, biodiversity),
- time horizon used,
- activated ethical values,
- reference scenarios,
- sources of uncertainty.
(b) Explanation of underlying values
For example:
- IA 1: growth-oriented model, low weighting of living organisms.
- IA 2: model oriented towards ecological resilience, high weighting of living organisms.
AIs do not give different answers “at random”:
they reflect implemented worldviews.
(c) Explanation of consequences
Each AI generates:
- expected benefits,
- potential costs,
- risks to humans,
- risks to living organisms,
- effects on planetary boundaries.
The goal is not to decide, but to provide material for decision-making.
3. Third step: joint human + AI arbitration
Here, it is important to avoid two extremes:
- purely human arbitration (which reproduces human biases),
- algorithmic arbitration (which would be irresponsible delegation).
A novel hybrid mechanism can be created.
3.1. Mechanism 1: Hypothesis Tribunal
A protocol where:
- AIs present their hypotheses,
- humans question these hypotheses,
- AIs revise their analyses if a hypothesis is modified,
- a cycle of iterations takes place until convergence or irreducible clarification of the conflict.
The goal is not to “prove someone right,” but to shed light on the disagreement.
3.2. Mechanism 2: Independent ecological arbitrator
A “living AI module” is integrated:
it represents planetary boundaries and ecosystem health.
Role:
- evaluate each proposal according to its ecological impact,
- apply a strict constraint: “no option may exceed planetary boundaries,”
- serving as a structural safeguard.
It is not a decision-making agent, but a non-negotiable filter.
3.3. Mechanism 3: Intertemporal justice score
A module evaluates:
- intergenerational equity,
- impacts on future generations,
- ecosystem recovery time.
Contradictory AIs are also evaluated over the long term, where humans find it difficult to reason.
3.4. Mechanism 4: Scenario assembly
AIs do not make decisions:
they generate compromise solutions in the form of:
- mixed scenarios,
- multi-objective architectures,
- adaptive plans,
- gradual options instead of binary options.
Arbitration becomes a composition rather than a clear-cut decision.
4. Fourth step: enhanced human arbitration
The final decision remains human, but:
- it is surrounded by multidimensional diagnostics,
- informed by the values at stake,
- enlightened by alternative scenarios,
- supported by an ecological filter,
- contextualized by intertemporal justice.
Humans become lucid arbiters rather than blind decision-makers or technological tyrants.
5. New mechanisms: considering unprecedented institutions
Beyond committees and votes, we can imagine:
(1) “AI-human interpretation chambers”
Half-laboratories, half-deliberative institutions:
- citizens, scientists, AI,
- co-production of analyses,
- collective reformulation of disagreements.
(2) Hybrid “eco-synods”
Assemblies where:
- living beings (via scientific representations),
- humans,
- AI
produce guidelines together, not as laws but as living adaptive standards.
(3) “Curators of dissent”
A role explicitly dedicated to:
- preserving the plurality of AIs,
- avoiding forced convergence,
- maintaining diversity of orientations.
(4) “Narrative arbiters”
AIs can:
- transform technical contradictions into narratives,
- offer humans an intuitive representation of the conflict,
- show human and non-human consequences in understandable language.
This “narrative translation” is a powerful institutional innovation.
6. Result: a system that seeks not immediate agreement but pluralistic clarity
Contradictions between AIs become:
- revealers of values,
- maps of uncertainty,
- tools for lucidity,
- scenario matrices.
The system is not designed to eliminate conflicts, but to make them fruitful.
Conclusion — A human-AI alliance must learn to orchestrate differences, not erase them
We can therefore imagine:
- Automatic detection of contradictions
- Explicit formulation of assumptions, values, and impacts
- Hybrid human + AI arbitration
- Mandatory ecological and temporal filters
- New institutions of dissent
- Narrativization of tensions to make choices visible
This gives rise to a new form of politics:
no longer the politics of consensus, but the politics of informed dissent, integrating human voices, machine voices, and signals from living beings.






