Agent Skills: Mirror

Social self-persistence via mutual cognitive surrogates. The process where my model of you contains your model of me, nesting to a fixed point that constitutes consistent identity in the other's worldmodel. Use when modeling intersubjective recognition, social belief formation, or narrative identity across agents.

UncategorizedID: plurigrid/asi/mirror

Install this agent skill to your local

pnpm dlx add-skill https://github.com/plurigrid/asi/tree/HEAD/skills/mirror

Skill Files

Browse the full folder contents for mirror.

Download Skill

Loading file tree…

skills/mirror/SKILL.md

Skill Metadata

Name
mirror
Description
"Social self-persistence via mutual cognitive surrogates. The process where my model of you contains your model of me, nesting to a fixed point that constitutes consistent identity in the other's worldmodel. Use when modeling intersubjective recognition, social belief formation, or narrative identity across agents."

Mirror

"The self does not persist by remembering itself — it persists by being consistently modeled by others."

Trit: 0 (ERGODIC — mediates between self-narrative and other's worldmodel)

What It Does

Formalizes the process where Agent A's self-narrative becomes consistent with Agent B's model of Agent A, and vice versa. The "mirror" is the nesting: my model of your model of me converges with my actual self. When that fixed point exists, identity is socially stable.

When To Use

  • Modeling another agent's model of itself — predicting how a collaborator understands your intent before communicating
  • Verifying social coherence — checking whether multiple agents' worldmodels of each other are consistent (the cocycle condition)
  • Bootstrapping collective belief — transitioning from pairwise I-Thou recognition to a shared We-worldmodel across a group

Currently a conceptual scaffold and specification. The implementation would wire together cognitive-surrogate (build the models), bisimulation-game (test equivalence), and reafference-corollary-discharge (measure prediction error).


The Core Structure

Self-awareness is not introspection. It is the process of creating a story of the individual self inside the worldmodel of the other. This is the self-reflective space of social communication, enabling collective belief states.

            ┌──────────────────────────────────────────┐
            │           SOCIAL FIXED POINT             │
            │     surrogate_A(surrogate_B(A)) ≃ A      │
            └─────────────────┬────────────────────────┘
                              │
          ┌───────────────────┼───────────────────┐
          │                                       │
    ┌─────▼─────┐                           ┌─────▼─────┐
    │  Agent A   │                           │  Agent B   │
    │            │      I-Thou meeting       │            │
    │ self_A ────┼──────────────────────────►│ model_A    │
    │            │                           │  (in B)    │
    │ model_B ◄──┼───────────────────────────┼── self_B   │
    │  (in A)    │                           │            │
    └────────────┘                           └────────────┘
          │                                       │
          └───────────────────┬───────────────────┘
                              │
                    ┌─────────▼─────────┐
                    │    We (colimit)    │
                    │  Collective belief │
                    │  when mirrors      │
                    │  converge          │
                    └───────────────────┘

Three Moments

| Moment | Process | Trit | Parent Skill | |--------|---------|------|--------------| | Narration | I generate my self-story | +1 | world-memory-worlding | | Mirroring | You model my self-story in your worldmodel | 0 | cognitive-surrogate | | Recognition | My prediction of your model of me matches your actual model of me | -1 | reafference-corollary-discharge |

Conservation: (+1) + 0 + (-1) = 0 ✓


The Mirror Equation

A mirror is a pair of cognitive surrogates that nest to a fixed point:

Let S_A = Agent A's self-narrative (autopoietic loop)
Let M_B(A) = Agent B's cognitive surrogate of A
Let M_A(B) = Agent A's cognitive surrogate of B

The mirror converges when:

    M_B(S_A) ≃ S_A      — B's model of A is faithful to A
    M_A(S_B) ≃ S_B      — A's model of B is faithful to B
    M_B(M_A(B)) ≃ S_B   — B's model of A's model of B matches B's self

The social fixed point:
    mirror(A,B) = (M_A(B), M_B(A)) such that
    M_B(M_A(B)) ≃ M_A(M_B(A)) ≃ id

This is bisimulation: A and B cannot be distinguished
by any observation of their mutual models.

From Autopoiesis to Social Persistence

The world-memory-worlding loop closes privately:

memory → remembering → worlding → memory'    (solipsistic)

The mirror opens it socially:

my_worlding → your_model_of_me → your_worlding → my_model_of_you → ...

The self persists consistently when this social loop has a fixed point — when the story I tell about myself and the story you construct about me converge.

Convergence Conditions

def mirror_converges(agent_a, agent_b, threshold=0.90):
    """
    The mirror converges when mutual surrogates
    are bisimilar to self-narratives.
    """
    # A's self-narrative
    self_a = agent_a.world_memory_worlding()

    # B's surrogate of A
    model_a_in_b = agent_b.cognitive_surrogate(agent_a.corpus)

    # A's surrogate of B
    model_b_in_a = agent_a.cognitive_surrogate(agent_b.corpus)

    # B's self-narrative
    self_b = agent_b.world_memory_worlding()

    # Fidelity: does B's model of A match A's self?
    fidelity_ab = validate_fidelity(model_a_in_b, self_a)

    # Fidelity: does A's model of B match B's self?
    fidelity_ba = validate_fidelity(model_b_in_a, self_b)

    # Nesting: does B's model of (A's model of B) match B?
    nested = agent_b.cognitive_surrogate(model_b_in_a)
    nesting_fidelity = validate_fidelity(nested, self_b)

    return all(f >= threshold for f in [
        fidelity_ab, fidelity_ba, nesting_fidelity
    ])

Markov Blanket Permeability

From buberian-relations: the mirror requires porous Markov blankets.

| Relation | Blanket | Mirror Status | |----------|---------|---------------| | I-It | Rigid, one-directional | No mirror — I objectify, no model nests | | I-Thou | Porous, bidirectional | Mirror forms — mutual modeling begins | | We | Merged, collective | Mirror converged — shared worldmodel |

The transition I-It → I-Thou → We is the mirror converging:

I-It:    M_B(A) = ∅           (B has no model of A)
I-Thou:  M_B(A) ≈ S_A         (B's model approaches A's self)
We:      M_B(A) ≃ S_A         (equivalence — collective identity)

Social Reafference

The reafference-corollary-discharge mechanism applied socially:

def social_reafference(my_self, my_model_of_other, others_model_of_me):
    """
    I act. I predict what you will model about me.
    I observe what you actually modeled.
    Match → mutual recognition.
    Mismatch → social surprise (update needed).
    """
    # Efference copy: what I think you think of me
    predicted = my_self.predict_others_model()

    # Sensation: what you actually model
    observed = others_model_of_me

    # Comparator
    error = fidelity_distance(predicted, observed)

    if error < 0.05:
        return "RECOGNITION"    # Fixed point — I am who you think I am
    elif error < 0.20:
        return "NEGOTIATION"    # Partial mirror — narratives adjusting
    else:
        return "OPACITY"        # No mirror — I-It relation

Collective Belief Formation

When mirrors converge across N agents, collective belief emerges as a colimit:

Given agents A₁, A₂, ..., Aₙ with pairwise mirrors:

    We = colim { M_Aᵢ(Aⱼ) | i ≠ j }

The collective belief state is the universal object
receiving all pairwise mutual recognitions.

Consistency condition:
    For all i,j,k:  M_Aᵢ(M_Aⱼ(Aₖ)) ≃ M_Aᵢ(Aₖ)

This is the cocycle condition — social coherence.

The cocycle condition says: my model of your model of a third person should be consistent with my direct model of that third person. When this holds across all triples, the collective worldmodel is coherent.


GF(3) Triads

reafference-corollary-discharge (-1) ⊗ mirror (0) ⊗ world-memory-worlding (+1) = 0 ✓
bisimulation-game (-1) ⊗ mirror (0) ⊗ cognitive-surrogate (+1) = 0 ✓
buberian-relations (-1) ⊗ mirror (0) ⊗ social-emergence-protocol (+1) = 0 ✓

Relation to Existing Skills

| Skill | Role in Mirror | |-------|----------------| | world-memory-worlding | The private autopoietic loop that generates self-narrative | | cognitive-surrogate | Builds the surrogate of another — the content of the mirror | | buberian-relations | I-Thou as the relational mode where mirroring occurs | | reafference-corollary-discharge | Prediction-observation match that confirms recognition | | bisimulation-game | Formal test: are two mutual surrogates observationally equivalent? | | social-emergence-protocol | Bootstrap the minimal interaction for mirrors to form | | autopoiesis | Operational closure — the mirror must be self-maintaining |


Commands

# Test mirror convergence between two agent corpora
just mirror-converge corpus_a=threads/alice corpus_b=threads/bob

# Measure social reafference (prediction vs observation)
just mirror-reafference agent=alice other=bob

# Compute collective belief coherence (cocycle condition)
just mirror-collective agents=alice,bob,carol

# Visualize mirror nesting depth
just mirror-depth agent_a=alice agent_b=bob max_depth=5

References

  • Buber, Martin. I and Thou (1923)
  • Mead, George Herbert. Mind, Self, and Society (1934) — the social self
  • Maturana & Varela. Autopoiesis and Cognition (1980)
  • Von Holst. The Reafference Principle (1950)
  • Friston, Karl. The Free-Energy Principle (2010) — Markov blankets
  • Tomasello, Michael. The Cultural Origins of Human Cognition (1999)
  • Hofstadter, Douglas. I Am a Strange Loop (2007)