From cc2715643ab803a72e84d2fb7b7ad5b9d68507de Mon Sep 17 00:00:00 2001 From: Meta Vens Date: Fri, 1 May 2026 19:29:49 +0100 Subject: [PATCH] Use confidence for perceived attribution strength --- .../agent-state-distributions-and-prompt-projection.md | 2 +- docs/architecture/agent-state-variable-glossary.md | 9 ++++----- docs/architecture/canonical-agent-state-schema.md | 2 +- examples/agent-state.call-of-the-void.json | 6 ++---- schemas/agent-state.schema.json | 8 +------- state/evidence.jsonl | 2 +- state/map.yaml | 4 ++-- tools/validate_agent_state.py | 5 +++-- 8 files changed, 15 insertions(+), 23 deletions(-) diff --git a/docs/architecture/agent-state-distributions-and-prompt-projection.md b/docs/architecture/agent-state-distributions-and-prompt-projection.md index c458a37..2f7539b 100644 --- a/docs/architecture/agent-state-distributions-and-prompt-projection.md +++ b/docs/architecture/agent-state-distributions-and-prompt-projection.md @@ -195,7 +195,7 @@ Canonical authored state should not include epistemic uncertainty about itself. If a value is only an observer's read, store it as perceived state rather than canonical state. Perceived dimensions should not expose canonical `mean` or `plasticity`; they should track observed activation, attributed disposition, -confidence, and correction resistance. This lets the model represent +and confidence. This lets the model represent fundamental attribution bias: someone can behave a certain way under pressure and be perceived as simply being that way. diff --git a/docs/architecture/agent-state-variable-glossary.md b/docs/architecture/agent-state-variable-glossary.md index 4566527..cefb390 100644 --- a/docs/architecture/agent-state-variable-glossary.md +++ b/docs/architecture/agent-state-variable-glossary.md @@ -47,8 +47,7 @@ evidence of what the target is like. { "observed_activation": 0.9, "attributed_disposition": 0.7, - "confidence": 0.4, - "correction_resistance": 0.6 + "confidence": 0.4 } ``` @@ -56,13 +55,13 @@ evidence of what the target is like. | --- | --- | | `observed_activation` | How strongly the observer thinks the target expressed the dimension in the observed event or scene. | | `attributed_disposition` | How much the observer generalizes from the observed behavior into "this is what the target is like." | -| `confidence` | How strongly the observer trusts this read. | -| `correction_resistance` | How much counter-evidence is likely to be needed before the observer revises the attribution. | +| `confidence` | How strongly the observer trusts this read. This is where evidence strength, emotional investment, prejudice, relationship stakes, prior memory, and social reinforcement show up. | This models fundamental attribution bias directly. A character can see someone behave coldly during a crisis and conclude "they are cold," not "they were activated toward distance under exceptional pressure." Correcting that -impression takes work. +impression takes work because the observer's confidence has causes, not because +the perceived variable gets a second stickiness dial. ## Label Set Policy diff --git a/docs/architecture/canonical-agent-state-schema.md b/docs/architecture/canonical-agent-state-schema.md index 0744a46..a6b6ff3 100644 --- a/docs/architecture/canonical-agent-state-schema.md +++ b/docs/architecture/canonical-agent-state-schema.md @@ -69,7 +69,7 @@ their whole soul because someone coughed near a status wound. Canonical variables do not include confidence or certainty. Perceived variables do not expose canonical `mean` or `plasticity`; they store observed activation, -attributed disposition, confidence, and correction resistance. +attributed disposition, and confidence. For full meanings of canonical `mean`, `plasticity`, `current_activation`, perceived attribution fields, and every required v0 label, see diff --git a/examples/agent-state.call-of-the-void.json b/examples/agent-state.call-of-the-void.json index abca1f0..a551746 100644 --- a/examples/agent-state.call-of-the-void.json +++ b/examples/agent-state.call-of-the-void.json @@ -746,14 +746,12 @@ "warmth": { "observed_activation": 0.4, "attributed_disposition": 0.44, - "confidence": 0.34, - "correction_resistance": 0.42 + "confidence": 0.34 }, "suspicion": { "observed_activation": 0.9, "attributed_disposition": 0.83, - "confidence": 0.71, - "correction_resistance": 0.64 + "confidence": 0.71 } }, "beliefs": [ diff --git a/schemas/agent-state.schema.json b/schemas/agent-state.schema.json index 5629079..efec087 100644 --- a/schemas/agent-state.schema.json +++ b/schemas/agent-state.schema.json @@ -654,8 +654,7 @@ "required": [ "observed_activation", "attributed_disposition", - "confidence", - "correction_resistance" + "confidence" ], "additionalProperties": false, "properties": { @@ -674,11 +673,6 @@ "minimum": 0, "maximum": 1 }, - "correction_resistance": { - "type": "number", - "minimum": 0, - "maximum": 1 - }, "evidence": { "type": "array", "items": { diff --git a/state/evidence.jsonl b/state/evidence.jsonl index 46dacac..dc909c9 100644 --- a/state/evidence.jsonl +++ b/state/evidence.jsonl @@ -9,4 +9,4 @@ {"ts": "2026-05-01T18:10:28+00:00", "type": "design", "status": "accepted", "note": "Split performed self from voice style in the v0 schema: presentation_strategy now includes abrasive_boundary and ironic_distance, while voice_style carries sarcasm, dryness, sincerity evasion, emotional evasion, and banter-as-boundary handles."} {"ts": "2026-05-01T18:14:18+00:00", "type": "documentation", "status": "ok", "note": "Added agent-state variable glossary defining numeric components, current v0 labels, label-set policy, and why canonical fixtures include full required vectors rather than only strongly applicable labels."} {"ts": "2026-05-01T18:21:07+00:00", "type": "design", "status": "accepted", "note": "Removed certainty from canonical authored state variables. Canonical character state now stores mean, plasticity, and current_activation; perceived/inferred dimensions use confidence when a read can be wrong."} -{"ts": "2026-05-01T18:26:40+00:00", "type": "design", "status": "accepted", "note": "Reworked perceived dimensions so observer reads do not know canonical mean or plasticity. Perceived state now tracks observed_activation, attributed_disposition, confidence, and correction_resistance to model fundamental attribution bias."} +{"ts": "2026-05-01T18:26:40+00:00", "type": "design", "status": "accepted", "note": "Reworked perceived dimensions so observer reads do not know canonical mean or plasticity. Perceived state now tracks observed_activation, attributed_disposition, and confidence to model fundamental attribution bias; confidence is driven by evidence, emotion, stakes, memory, prejudice, and social reinforcement."} diff --git a/state/map.yaml b/state/map.yaml index fa941ef..c1e0bdc 100644 --- a/state/map.yaml +++ b/state/map.yaml @@ -74,8 +74,8 @@ accepted_design: - Ghostlight should build drama and narrative pressure, then project character-local dialogue contexts so characters can breathe and act from their own partial state. - Call of the Void pressure should come from Sol's whole unresolved inheritance being trapped inside Elysium, not from a lone expedition discovering there is no way home. - The first schema contract lives in schemas/agent-state.schema.json with required first-class dimensions in schemas/agent-state.required-fields.json and a validator in tools/validate_agent_state.py. - - Canonical authored character variables do not carry certainty; perceived dimensions do not expose canonical mean or plasticity, and instead model observed activation, attributed disposition, confidence, and correction resistance. - - Fundamental attribution bias is first-class in perceived state: observers can mistake exceptional behavior for stable character, and correction should require counter-evidence. + - Canonical authored character variables do not carry certainty; perceived dimensions do not expose canonical mean or plasticity, and instead model observed activation, attributed disposition, and confidence. + - Fundamental attribution bias is first-class in perceived state: observers can mistake exceptional behavior for stable character, with stubborn reads emerging from confidence drivers such as memory, stakes, emotion, prejudice, and social reinforcement. current_status: summary: Ghostlight now has the Epiphany-style persistence spine, the first architecture payload, and the initial canonical agent-state schema seam for Call of the Void-style authoring examples. The schema is still v0 and should be hardened before building prompt projection on top of it. diff --git a/tools/validate_agent_state.py b/tools/validate_agent_state.py index 29cd197..b2b2288 100644 --- a/tools/validate_agent_state.py +++ b/tools/validate_agent_state.py @@ -53,14 +53,15 @@ def validate_inferred_state_variable(obj: Any, path: str) -> None: require(isinstance(obj, dict), f"{path} must be an object") require_keys( obj, - ["observed_activation", "attributed_disposition", "confidence", "correction_resistance"], + ["observed_activation", "attributed_disposition", "confidence"], path, ) require("mean" not in obj, f"{path}.mean belongs in canonical state, not perceived state") require("plasticity" not in obj, f"{path}.plasticity belongs in canonical state, not perceived state") require("current_activation" not in obj, f"{path}.current_activation should be observed_activation in perceived state") require("certainty" not in obj, f"{path}.certainty should be named confidence in perceived/inferred state") - for key in ["observed_activation", "attributed_disposition", "confidence", "correction_resistance"]: + require("correction_resistance" not in obj, f"{path}.correction_resistance should be represented through confidence drivers") + for key in ["observed_activation", "attributed_disposition", "confidence"]: require_0_1(obj[key], f"{path}.{key}")