Phase C: Contagion, Timeline Events, and Merged Pass Reasoning

docs/simulation-v2-architecture.md

@@ -34,7 +34,7 @@ Decisions confirmed before implementation. These override any conflicting detail
 | 5 | Timeline merge semantics | Timeline entry overrides base event for that timestep. |
 | 6 | DB schema for new artifacts | Define conversations/posts/action_history tables before Phase D. |
 | 7 | Name data | Local SSA baby names + Census surnames, bundled CSVs (~500KB), US-only. Non-US via country-specific CSVs later behind same interface: `generate_name(gender, ethnicity, birth_decade, country="US")`. |
-| 8 | Conformity/threshold mechanics | Soft prompt signal only (inject local adoption ratio + conformity phrasing). No hard numeric gate. |
+| 8 | Conformity/threshold mechanics | Soft prompt signal only (conformity self-awareness + peer opinions + mood rendering). No explicit ratios or hard numeric gates. |
 | 9 | Backtesting ground-truth | Define one validation dataset schema before Phase G. |
 
 ### Phase-Specific Decisions
@@ -1034,12 +1034,12 @@ The 12 tenets below define what a high-fidelity simulation must satisfy. Tenets
 | 3 | Social hierarchy & influence topology | **Partial** | Structural role edges, degree multipliers in network config, edge weight hierarchy | No explicit power-law degree enforcement; no hub/opinion-leader generation. Need scenario-dependent centrality targets |
 | 4 | Behavioral heterogeneity | **Partial** | Big Five personality, risk tolerance, institutional trust, cognitive attributes vary per agent | Decision-policy heterogeneity relies entirely on LLM interpretation of persona. Need explicit behavioral parameters (conformity threshold, action inertia) as agent attributes |
 | 5 | Temporal dynamics & decay | **Strong** | Conviction decay, temporal prompt awareness, emotional trajectory, memory history, scenario timeline with evolving events | Need intent→action accountability loop (surface prior action_intent, ask about follow-through) |
-| 6 | Social contagion & network effects | **Partial** | Network propagation, share modifiers, conversation system, aggregate mood, peer opinions | No explicit threshold/complex contagion. Need per-agent conformity parameter + local adoption ratio injection |
+| 6 | Social contagion & network effects | **Partial** | Network propagation, share modifiers, conversation system, aggregate mood, peer opinions | No explicit threshold/complex contagion. Need per-agent conformity parameter + conformity-aware prompt phrasing |
 | 7 | Friction & transaction costs | **Weak** | option_friction on outcomes, bounded confidence mechanics | Biggest gap. Need explicit intent→behavior pipeline: surface what agent planned vs what they actually did. Friction emerges from agent constraints but isn't tracked or measured |
 | 8 | Bounded rationality & heuristics | **Strong** | LLM is inherently a bounded rationality engine. Persona attributes (education, digital literacy, neuroticism) shape heuristic use. Agents satisfice, anchor, exhibit status quo bias naturally | Could strengthen with explicit bias nudges in prompts for specific attributes |
 | 9 | Environmental & contextual sensitivity | **Partial** | Scenario timeline handles exogenous shocks. Channel templates adapt to agent demographics | Need ambient macro context in every prompt (economic conditions, cultural moment). Need previous-timestep macro summary injection |
 | 10 | Identity & group membership | **Partial** | race_ethnicity, political_orientation, religious_affiliation in persona. Social role edges create in-group connections | Need identity-threat framing: when the scenario threatens a group identity, persona rendering should explicitly flag it as identity-relevant |
-| 11 | Preference interdependence | **Partial** | Aggregate mood rendering ("most people I know are doing X"), peer opinions, social posts. Bandwagon/FOMO effects emerge from context | Need explicit local adoption ratio in prompt: "X% of people you know have already done Y." Makes interdependence concrete, not just vibes |
+| 11 | Preference interdependence | **Partial** | Aggregate mood rendering ("most people I know are doing X"), peer opinions, social posts. Bandwagon/FOMO effects emerge from context | Named peer opinions + local mood + macro summary provide social pressure without omniscient ratio framing |
 | 12 | Macro-micro feedback loops | **Partial** | Micro→macro works (agent decisions → aggregate stats). Timeline handles exogenous macro shifts | No endogenous macro: agent behavior doesn't produce emergent macro variables that feed back. Need at minimum: inject previous timestep aggregates as ambient context |
 
 ### Concrete Fixes to Close Gaps
@@ -1087,16 +1087,19 @@ These are the minimum changes needed to move every tenet to **Strong**. Listed i
 
 **Soft conformity/threshold behavior:**
 - Add `conformity` as a standard personality attribute (0-1 scale, correlated with agreeableness). Sampled at population creation time.
-- At prompt build time, compute **local adoption ratio**: what fraction of this agent's network has already taken action (changed position, shared, etc.)
-- Inject into prompt: "About 7 out of 10 people you know have already started making changes. You tend to [wait until most people around you have acted / act independently of what others are doing]." (phrasing depends on conformity level)
-- This gives the LLM explicit threshold context without hardcoding a threshold formula. A high-conformity agent seeing 70% adoption will likely act. A low-conformity agent seeing the same might resist specifically because everyone else is doing it (contrarian behavior).
+- Inject conformity self-awareness into prompt: "I tend to go along with what most people around me are doing" (high) or "I tend to form my own opinion regardless of what others think" (low). Mid-range agents get no explicit phrasing.
+- Social pressure is conveyed through **existing mechanisms**, not explicit ratios:
+  - Named peer opinions: "My coworker Darnell thinks X"
+  - Local mood rendering: "Most people around me seem worried"
+  - Macro summary: "The general mood is shifting toward X"
+- **Rationale:** People don't actually know "7 out of 10 contacts did X" — that's omniscient narrator framing. Real social pressure comes from specific conversations and vague impressions, which the peer opinion and mood systems already capture.
 
 **Macro state feedback:**
 - After each timestep, compute macro summary from TimestepSummary data:
   - Position distribution rendered as "Most people are choosing X. A growing minority is doing Y."
   - Sentiment trend: "The general mood is getting worse / stabilizing / improving."
   - Exposure saturation: "Almost everyone has heard about this now."
-  - Action adoption rate: "About X% of people have already taken concrete action."
+  - Action momentum: "More and more people are taking action" / "Most people are still waiting"
 - Inject this into every agent's next-timestep prompt as ambient context, rendered as what the agent would sense from media/social feeds, not raw numbers.
 - This closes the macro→micro loop: agent decisions → aggregate stats → rendered as ambient context → influences next round of agent decisions.
 
@@ -1203,7 +1206,7 @@ Ship this alone. Every simulation immediately feels more human, and the accounta
 
 - Scenario timeline: sequence of events at specified timesteps
 - Timeline injection into agent prompts as "what's happened since last time"
-- Local adoption ratio computed per agent ("7 out of 10 people you know have acted")
+- Named peer opinions + local mood convey social pressure without explicit ratios
 - Conformity-aware prompt rendering ("You tend to wait for others / act independently")
 - Ambient scenario context field (`background_context` in ScenarioSpec)
 - Macro state feedback: timestep aggregates rendered as ambient vibes in next prompt

extropy/cli/commands/extend.py

@@ -233,6 +233,8 @@ def do_hydration():
             attributes=bound_attrs,
             sampling_order=sampling_order,
             sources=sources,
+            household_config=household_config,
+            name_config=name_config,
         )
         merged_spec = base.merge(extension_spec)
 

extropy/cli/commands/scenario.py

@@ -36,6 +36,11 @@ def scenario_command(
         "-o",
         help="Output path (defaults to {population_stem}.scenario.yaml)",
     ),
+    timeline: str = typer.Option(
+        "auto",
+        "--timeline",
+        help="Timeline mode: auto (LLM decides), static (single event), evolving (multi-event)",
+    ),
     yes: bool = typer.Option(False, "--yes", "-y", help="Skip confirmation prompts"),
 ):
     """
@@ -116,6 +121,8 @@ def on_progress(step: str, status: str):
     def run_pipeline():
         nonlocal result_spec, validation_result, pipeline_error
         try:
+            # Convert timeline mode (auto -> None for LLM decision)
+            timeline_mode = None if timeline == "auto" else timeline
             result_spec, validation_result = create_scenario(
                 description=scenario_desc,
                 population_spec_path=population,
@@ -124,6 +131,7 @@ def run_pipeline():
                 network_id=network_id,
                 output_path=None,  # Don't save yet
                 on_progress=on_progress,
+                timeline_mode=timeline_mode,
             )
         except Exception as e:
             pipeline_error = e
@@ -214,6 +222,29 @@ def run_pipeline():
     )
     console.print()
 
+    # Timeline info (Phase C)
+    if result_spec.timeline:
+        console.print(f"[bold]Timeline:[/bold] {len(result_spec.timeline)} events")
+        for te in result_spec.timeline[:3]:
+            desc = te.description or te.event.content[:40]
+            console.print(f"  • t={te.timestep}: {desc}")
+        if len(result_spec.timeline) > 3:
+            console.print(f"  [dim]... and {len(result_spec.timeline) - 3} more[/dim]")
+        console.print()
+    else:
+        console.print("[bold]Timeline:[/bold] static (single event)")
+        console.print()
+
+    # Background context (Phase C)
+    if result_spec.background_context:
+        ctx_preview = (
+            result_spec.background_context[:60] + "..."
+            if len(result_spec.background_context) > 60
+            else result_spec.background_context
+        )
+        console.print(f"[bold]Background:[/bold] {ctx_preview}")
+        console.print()
+
     # Validation Results
     if validation_result.errors:
         console.print(

extropy/cli/commands/simulate.py

@@ -186,6 +186,11 @@ def simulate_command(
         "-p",
         help="PersonaConfig YAML for embodied personas (auto-detected if not specified)",
     ),
+    merged_pass: bool = typer.Option(
+        False,
+        "--merged-pass",
+        help="Use single merged reasoning pass instead of two-pass (experimental)",
+    ),
     quiet: bool = typer.Option(False, "--quiet", "-q", help="Suppress progress output"),
     verbose: bool = typer.Option(False, "--verbose", "-v", help="Show detailed logs"),
     debug: bool = typer.Option(
@@ -316,6 +321,7 @@ def on_progress(timestep: int, max_timesteps: int, status: str):
                 writer_queue_size=writer_queue_size,
                 db_write_batch_size=db_write_batch_size,
                 resource_governor=governor,
+                merged_pass=merged_pass,
             )
             simulation_error = None
         except Exception as e:
@@ -352,6 +358,7 @@ def do_simulation():
                     writer_queue_size=writer_queue_size,
                     db_write_batch_size=db_write_batch_size,
                     resource_governor=governor,
+                    merged_pass=merged_pass,
                 )
             except Exception as e:
                 simulation_error = e

extropy/core/models/__init__.py

@@ -57,6 +57,8 @@
     # Event
     EventType,
     Event,
+    # Timeline
+    TimelineEvent,
     # Exposure
     ExposureChannel,
     ExposureRule,
@@ -160,6 +162,8 @@
     # Scenario - Event
     "EventType",
     "Event",
+    # Scenario - Timeline
+    "TimelineEvent",
     # Scenario - Exposure
     "ExposureChannel",
     "ExposureRule",

extropy/core/models/scenario.py

@@ -114,6 +114,26 @@ class SeedExposure(BaseModel):
     )
 
 
+class TimelineEvent(BaseModel):
+    """A development in the scenario timeline.
+
+    Timeline events represent how a scenario evolves over time. For evolving
+    scenarios (crises, campaigns), multiple events occur at different timesteps.
+    Static scenarios (policy announcements) have no timeline events.
+    """
+
+    timestep: int = Field(ge=0, description="When this development occurs")
+    event: Event = Field(description="The event content at this timestep")
+    exposure_rules: list[ExposureRule] | None = Field(
+        default=None,
+        description="Custom exposure rules; if None, reuses seed_exposure.rules with updated content",
+    )
+    description: str | None = Field(
+        default=None,
+        description="Human-readable context for this development",
+    )
+
+
 # =============================================================================
 # Interaction Model
 # =============================================================================
@@ -285,6 +305,10 @@ class ScenarioSpec(BaseModel):
 
     meta: ScenarioMeta
     event: Event
+    timeline: list[TimelineEvent] | None = Field(
+        default=None,
+        description="Subsequent developments; None or empty = static scenario",
+    )
     seed_exposure: SeedExposure
     interaction: InteractionConfig
     spread: SpreadConfig

extropy/core/models/simulation.py

@@ -314,6 +314,23 @@ class ReasoningContext(BaseModel):
         default_factory=dict,
         description="Mapping of agent_id → first name for resolving peer references",
     )
+    # Phase C additions
+    timeline_recap: list[str] | None = Field(
+        default=None,
+        description="Bullet list of what's happened so far in the scenario",
+    )
+    current_development: str | None = Field(
+        default=None,
+        description="This timestep's new development (if any)",
+    )
+    observable_peer_actions: int | None = Field(
+        default=None,
+        description="Count of neighbors who visibly acted (shared/posted)",
+    )
+    conformity: float | None = Field(
+        default=None,
+        description="Agent's conformity attribute (0-1)",
+    )
 
 
 # =============================================================================
@@ -390,6 +407,10 @@ class SimulationRunConfig(BaseModel):
         default=None,
         description="Max concurrent async reasoning calls (None = auto from RPM)",
     )
+    merged_pass: bool = Field(
+        default=False,
+        description="Use single merged pass instead of two-pass reasoning (experimental)",
+    )
 
     # Backward compat aliases
     @property

extropy/core/providers/anthropic.py

@@ -29,13 +29,27 @@
 def _clean_schema_for_tool(schema: dict) -> dict:
     """Clean a JSON schema for use as a tool input_schema.
 
-    Removes fields that aren't valid in tool input schemas
-    (like 'additionalProperties' in nested objects that Claude
-    doesn't support in tool definitions).
+    Anthropic structured outputs support additionalProperties: false but NOT
+    schema-valued additionalProperties (e.g. {"type": "number"}).
+
+    This function:
+    - Keeps additionalProperties: false (valid and useful)
+    - Strips additionalProperties when it's a dict/schema (not supported)
+    - Logs a warning when stripping schema-valued additionalProperties
     """
     cleaned = {}
     for key, value in schema.items():
         if key == "additionalProperties":
+            if value is False:
+                # Keep additionalProperties: false - it's valid
+                cleaned[key] = value
+            elif isinstance(value, dict):
+                # Schema-valued additionalProperties not supported - strip with warning
+                logger.warning(
+                    "Stripping schema-valued additionalProperties from tool schema "
+                    "(not supported by Anthropic structured outputs)"
+                )
+            # Skip other truthy values (True, etc.)
             continue
         if isinstance(value, dict):
             cleaned[key] = _clean_schema_for_tool(value)