Mind · Strategy
PerspectiveEnsemble
Simulating diverse experts
Definition
Mechanism
The Agent instantiates N virtual personas to debate the Problem from distinct, pre-defined viewpoints. The Synthesis emerges from their interaction. Unlike a single Chain-of-Thought, this forces the modeling of conflicting priors. It generates diverse viewpoints using SteelmanCheck on opposing arguments, then resolves the conflict via Aggregate (e.g. Mode).
Preconditions
- A persona set with pre-defined, mutually distinct viewpoints
Invariants
- Semantic Distance, every round: CosineSimilarity(Persona_A_Output, Persona_B_Output) < 0.6 for each round, not only the first. Checked once, it prevents early collapse; checked continuously, it is also what detects a persona drifting off its stance.
- Independence: Personas cannot see peer outputs in Round 1
Risks
Failure modes
Strawman Waltz
Personas agree too easily (Fabricated consensus).
Role Drift
Personas lose their specific stance over long Context windows.
Relationships
Builds on
References
Autonomous Goal-Directed Entity
Infrastructure
Reduce a set to a summary statistic
Infrastructure
Sequential data container (linked list)
Infrastructure
Portable execution environment with inherited constraints
Infrastructure
A state gap with a cost of inaction
Qualify the opposition, then release only if the conclusion remains robust
Combining elements
Used by
References
Composes with
Value-based state evaluation
Reference
Parameters
personasList[String], [3, 7]- Distinct viewpoints
roundsInteger, [1, 5]- Debate turns