Tighten agent state schema from evaluation

This commit is contained in:
Meta Vens 2026-05-01 20:57:34 +01:00
commit 5f4e7c6bb2
14 changed files with 615 additions and 48 deletions

View file

@ -120,11 +120,6 @@
}
},
"behavioral_dimensions": {
"warmth": {
"mean": 0.54,
"plasticity": 0.42,
"current_activation": 0.46
},
"drive": {
"mean": 0.76,
"plasticity": 0.21,
@ -184,6 +179,11 @@
"mean": 0.69,
"plasticity": 0.34,
"current_activation": 0.78
},
"interpersonal_warmth": {
"mean": 0.54,
"plasticity": 0.42,
"current_activation": 0.46
}
},
"presentation_strategy": {
@ -309,11 +309,6 @@
"plasticity": 0.18,
"current_activation": 0.91
},
"warmth": {
"mean": 0.31,
"plasticity": 0.45,
"current_activation": 0.26
},
"formality": {
"mean": 0.22,
"plasticity": 0.52,
@ -443,6 +438,11 @@
"mean": 0.52,
"plasticity": 0.38,
"current_activation": 0.57
},
"verbal_warmth": {
"mean": 0.31,
"plasticity": 0.45,
"current_activation": 0.26
}
}
},
@ -480,9 +480,70 @@
"linked_event_ids": []
}
],
"relationship_summaries": []
"relationship_summaries": [
{
"memory_id": "memory-rel-cat-oz-current-read",
"summary": "Oz is useful, evasive, and probably a mask for someone more vulnerable than the drone.",
"salience": 0.74,
"confidence": 0.61,
"linked_event_ids": [
"event-drone-lie"
],
"linked_relationship_id": "rel-cat-oz"
}
]
},
"perceived_state_overlays": []
"perceived_state_overlays": [
{
"observer_agent_id": "cat_marrigan",
"target_agent_id": "oz_operator",
"perceived_dimensions": {
"strategic_opacity": {
"observed_activation": 0.86,
"attributed_disposition": 0.72,
"confidence": 0.64,
"evidence": [
"event-drone-lie"
]
},
"cultivated_harmlessness": {
"observed_activation": 0.78,
"attributed_disposition": 0.51,
"confidence": 0.48,
"evidence": [
"event-drone-lie"
]
},
"interpersonal_warmth": {
"observed_activation": 0.52,
"attributed_disposition": 0.39,
"confidence": 0.35,
"evidence": [
"event-drone-lie"
]
}
},
"beliefs": [
{
"belief_id": "belief-cat-oz-is-a-mask",
"subject_id": "oz_operator",
"claim_type": "identity_inference",
"claim": "Oz is probably a mask for someone or something more vulnerable than the drone.",
"confidence": 0.57,
"evidence_event_ids": [
"event-drone-lie"
],
"visibility": "private",
"emotional_charge": 0.71
}
],
"distortions": [
"threat_amplification",
"institutional_suspicion_transfer",
"mask_overfocus"
]
}
]
},
{
"agent_id": "oz_operator",
@ -584,11 +645,6 @@
}
},
"behavioral_dimensions": {
"warmth": {
"mean": 0.67,
"plasticity": 0.4,
"current_activation": 0.74
},
"drive": {
"mean": 0.52,
"plasticity": 0.45,
@ -648,6 +704,11 @@
"mean": 0.64,
"plasticity": 0.44,
"current_activation": 0.72
},
"interpersonal_warmth": {
"mean": 0.67,
"plasticity": 0.4,
"current_activation": 0.74
}
},
"presentation_strategy": {
@ -773,11 +834,6 @@
"plasticity": 0.49,
"current_activation": 0.35
},
"warmth": {
"mean": 0.48,
"plasticity": 0.44,
"current_activation": 0.52
},
"formality": {
"mean": 0.58,
"plasticity": 0.38,
@ -907,6 +963,11 @@
"mean": 0.05,
"plasticity": 0.39,
"current_activation": 0.03
},
"verbal_warmth": {
"mean": 0.48,
"plasticity": 0.44,
"current_activation": 0.52
}
}
},
@ -936,22 +997,33 @@
"linked_event_ids": []
}
],
"relationship_summaries": []
"relationship_summaries": [
{
"memory_id": "memory-rel-oz-cat-current-read",
"summary": "Cat is suspicious enough to be dangerous and principled enough that the danger may have rules.",
"salience": 0.76,
"confidence": 0.58,
"linked_event_ids": [
"event-drone-lie"
],
"linked_relationship_id": "rel-oz-cat"
}
]
},
"perceived_state_overlays": [
{
"observer_agent_id": "oz_operator",
"target_agent_id": "cat_marrigan",
"perceived_dimensions": {
"warmth": {
"observed_activation": 0.4,
"attributed_disposition": 0.44,
"confidence": 0.34
},
"suspicion": {
"observed_activation": 0.9,
"attributed_disposition": 0.83,
"confidence": 0.71
},
"interpersonal_warmth": {
"observed_activation": 0.4,
"attributed_disposition": 0.44,
"confidence": 0.34
}
},
"beliefs": [
@ -1116,7 +1188,7 @@
{
"event_id": "event-drone-lie",
"kind": "dialogue",
"summary": "Oz dodged Cat's question about who was controlling the drone.",
"summary": "Cat asked who was controlling the drone; Oz redirected toward the repair instead of answering the identity question.",
"participants": [
"cat_marrigan",
"oz_operator"
@ -1125,6 +1197,104 @@
"mask",
"privacy",
"suspicion"
],
"observed_exchange": [
{
"speaker_id": "cat_marrigan",
"utterance_summary": "Cat pressed on whether Oz was really autonomous or a mask for someone remote.",
"dialogue_function": "probe_identity",
"observable_cues": [
"pointed humor",
"feigned casualness",
"attention on drone hesitation"
]
},
{
"speaker_id": "oz_operator",
"utterance_summary": "Oz answered the practical repair issue and avoided naming the operator behind the drone.",
"dialogue_function": "evasive_redirect",
"observable_cues": [
"technical compression",
"polite deflection",
"too-human timing"
]
}
],
"private_interpretations": [
{
"agent_id": "cat_marrigan",
"interpretation": "Oz is hiding a person-shaped vulnerability behind the drone body.",
"attributed_motive": "Protect the operator by keeping Cat focused on the broken taxi instead of the mask.",
"appraisal_tags": [
"trust_violation",
"curiosity",
"useful_secret",
"possible_leverage"
],
"confidence": 0.62
},
{
"agent_id": "oz_operator",
"interpretation": "Cat noticed the mask and is testing whether Oz will panic or confess.",
"attributed_motive": "Find out what Oz is without making her own dependence on the repair visible.",
"appraisal_tags": [
"exposure_risk",
"status_threat",
"possible_care",
"conversation_hazard"
],
"confidence": 0.68
}
],
"event_effects": [
{
"agent_id": "cat_marrigan",
"summary": "Cat trusts Oz less, becomes more fascinated, and stores the drone as a mask rather than a personless tool.",
"state_deltas": [
{
"path": "relationships.rel-cat-oz.stance.trust.current_activation",
"delta": -0.07
},
{
"path": "relationships.rel-cat-oz.stance.fascination.current_activation",
"delta": 0.13
},
{
"path": "canonical_state.behavioral_dimensions.suspicion.current_activation",
"delta": 0.11
}
],
"new_belief_ids": [
"belief-cat-oz-is-a-mask"
],
"memory_update_ids": [
"memory-rel-cat-oz-current-read"
]
},
{
"agent_id": "oz_operator",
"summary": "Oz reads Cat as sharper and riskier, but not automatically cruel.",
"state_deltas": [
{
"path": "relationships.rel-oz-cat.stance.fear.current_activation",
"delta": 0.08
},
{
"path": "relationships.rel-oz-cat.stance.respect.current_activation",
"delta": 0.07
},
{
"path": "canonical_state.presentation_strategy.strategic_opacity.current_activation",
"delta": 0.06
}
],
"new_belief_ids": [
"belief-cat-will-pry"
],
"memory_update_ids": [
"memory-rel-oz-cat-current-read"
]
}
]
}
],
@ -1195,7 +1365,7 @@
"presentation_constraints": [
"Be helpful, a little funny, and evasive.",
"Let the drone seem more autonomous than it is.",
"Use warmth carefully; Cat may treat open sincerity as bait.",
"Use verbal warmth carefully; Cat may treat open sincerity as bait.",
"Keep the operator location private unless scene pressure explicitly breaks the mask."
]
}