Validation

Validator classes, constraint enforcement mechanisms, and error detection

Overview

Timepoint Pro uses a plugin-based validation framework to enforce constraints at multiple levels:

• Temporal causality (M1.2) - Entities can’t know things before they learn them
• Information conservation - Knowledge requires exposure events (M3)
• Energy budgets - Interaction costs with circadian adjustments (M14)
• Biological constraints (M4) - Age-dependent capabilities, resource limits
• Dialog realism - Physical/emotional states affect conversation ability
• Relationship consistency - Dialog tone matches trust levels
• Environmental constraints (M16) - Animistic entities impose limits

Validators run automatically during simulation and can be invoked manually for debugging.

Validator Base Class

All validators extend the plugin-based Validator class:

from validation import Validator

class Validator(ABC):
    """Base validator with plugin registry"""
    
    _validators = {}
    
    @classmethod
    def register(cls, name: str, severity: str = "ERROR"):
        def decorator(func: Callable):
            cls._validators[name] = {"func": func, "severity": severity}
            return func
        return decorator
    
    @classmethod
    def validate_all(cls, entity: Entity, context: dict) -> list[dict]:
        violations = []
        for name, validator in cls._validators.items():
            result = validator["func"](entity, context)
            if not result["valid"]:
                violations.append({
                    "validator": name,
                    "severity": validator["severity"],
                    "message": result["message"]
                })
        return violations

Core Validators

Information Conservation

Mechanism: M3 (Exposure Events)
Severity: ERROR

Validates that entity knowledge is a subset of their exposure history:

@Validator.register("information_conservation", "ERROR")
def validate_information_conservation(entity: Entity, context: dict, store=None) -> dict:
    """Validate knowledge ⊆ exposure history"""
    
    # Query exposure events from database
    exposure_events = store.get_exposure_events(entity.entity_id)
    exposure = set(event.information for event in exposure_events)
    
    # Get entity's knowledge state
    knowledge = set(entity.entity_metadata.get("knowledge_state", []))
    
    # Check for anachronisms
    unknown = knowledge - exposure
    if unknown:
        return {
            "valid": False,
            "message": f"Entity knows about {unknown} without exposure"
        }
    
    return {"valid": True, "message": "Information conservation satisfied"}

Example violation:

{
  "validator": "information_conservation",
  "severity": "ERROR",
  "message": "Entity Webb knows about {'O2 fault reading'} without exposure"
}

Fix: Add exposure event:

from workflows import create_exposure_event

create_exposure_event(
    entity_id="Webb",
    information="O2 fault reading",
    source="Chen",
    event_type="told",
    timestamp=timepoint.timestamp,
    confidence=0.9,
    store=store
)

Temporal Causality

Mechanism: M1.2 (Fourth Wall Knowledge Filtering)
Severity: ERROR

Validates that entity knowledge follows causal temporal constraints:

@Validator.register("temporal_causality", "ERROR")
def validate_temporal_causality(entity: Entity, context: dict) -> dict:
    """Validate that entity knowledge follows causal temporal constraints"""
    
    store = context.get("store")
    timepoint_id = context.get("timepoint_id")
    
    knowledge_state = entity.entity_metadata.get("knowledge_state", [])
    invalid_items = []
    
    for knowledge_item in knowledge_state:
        # Check when entity learned this knowledge
        validation = validate_temporal_reference(
            entity.entity_id, knowledge_item, timepoint_id, store
        )
        if not validation["valid"]:
            invalid_items.append(knowledge_item)
    
    if invalid_items:
        return {
            "valid": False,
            "message": f"Entity has knowledge {invalid_items} that violates temporal causality"
        }
    
    return {"valid": True, "message": "Temporal causality satisfied"}

Example violation (FORWARD mode):

{
  "validator": "temporal_causality",
  "severity": "ERROR",
  "message": "Entity Webb has knowledge ['Mission outcome'] that violates temporal causality"
}

In FORWARD mode, entities can’t know about future events. This validator ensures strict forward causality.

Portal mode exception:

In PORTAL mode, backward-inferred knowledge is allowed:

if mode == "portal" and knowledge_item.inferred_backward:
    # Allow knowledge inferred from future state
    continue

Energy Budget

Mechanism: M14 (Circadian Patterns)
Severity: WARNING

Validates interaction costs ≤ capacity with circadian adjustments:

@Validator.register("energy_budget", "WARNING")
def validate_energy_budget(entity: Entity, context: dict) -> dict:
    """Validate interaction costs ≤ capacity with circadian adjustments"""
    
    budget = entity.entity_metadata.get("energy_budget", 100)
    current_knowledge = set(entity.entity_metadata.get("knowledge_state", []))
    previous_knowledge = set(context.get("previous_knowledge", []))
    
    new_knowledge_count = len(current_knowledge - previous_knowledge)
    base_expenditure = new_knowledge_count * 5
    
    # Apply circadian adjustments
    timepoint = context.get("timepoint")
    circadian_config = context.get("circadian_config", {})
    
    if timepoint and circadian_config:
        activity_type = context.get("activity_type", "work")
        expenditure = compute_energy_cost_with_circadian(
            activity_type, timepoint.timestamp.hour, base_expenditure, circadian_config
        )
    else:
        expenditure = base_expenditure
    
    if expenditure > budget * 1.2:  # Allow 20% temporary excess
        return {
            "valid": False,
            "message": f"Energy expenditure {expenditure:.1f} exceeds budget {budget}"
        }
    
    return {"valid": True, "message": "Energy budget satisfied"}

Circadian penalty example:

# Nighttime work (22:00-06:00)
hour = 23
night_penalty = 1.5  # 50% more draining
fatigue_factor = 1.3  # 17 hours awake

total_cost = base_cost * night_penalty * fatigue_factor
# 50 * 1.5 * 1.3 = 97.5

Biological Constraints

Mechanism: M4 (Constraint Enforcement)
Severity: ERROR

Validates age-dependent capabilities and biological plausibility:

@Validator.register("biological_constraints", "ERROR")
@track_mechanism("M4", "constraint_enforcement")
def validate_biological_constraints(entity: Entity, context: dict) -> dict:
    """Validate age-dependent capabilities and biological plausibility"""
    
    age = entity.entity_metadata.get("age", 0)
    action = context.get("action", "")
    violations = []
    
    # Age-based constraint checks
    if age > 100 and "physical_labor" in action:
        violations.append(f"age {age} incompatible with physical labor")
    
    if (age < 18 or age > 50) and "childbirth" in action:
        violations.append(f"age {age} outside plausible childbirth range (18-50)")
    
    if age < 5 and any(a in action for a in ["negotiate", "strategic_planning"]):
        violations.append(f"age {age} incompatible with {action}")
    
    if age > 80 and any(a in action for a in ["sprint", "heavy_lifting", "combat"]):
        violations.append(f"age {age} incompatible with {action}")
    
    if violations:
        return {
            "valid": False,
            "message": f"Constraint violations: {'; '.join(violations)}"
        }
    
    return {"valid": True, "message": "Biological constraints satisfied"}

Example violation:

{
  "validator": "biological_constraints",
  "severity": "ERROR",
  "message": "Constraint violations: age 85 incompatible with sprint"
}

Behavioral Inertia

Mechanism: M1 (Heterogeneous Fidelity)
Severity: WARNING

Validates that personality drift is gradual:

@Validator.register("behavioral_inertia", "WARNING")
def validate_behavioral_inertia(entity: Entity, context: dict) -> dict:
    """Validate personality drift is gradual"""
    
    if "previous_personality" not in context:
        return {"valid": True, "message": "No previous state to compare"}
    
    current = np.array(entity.entity_metadata.get("personality_traits", []))
    previous = np.array(context["previous_personality"])
    
    if len(current) == 0 or len(previous) == 0:
        return {"valid": True, "message": "Personality data not available"}
    
    drift = np.linalg.norm(current - previous)
    if drift > 1.0:  # Threshold for significant personality change
        return {
            "valid": False,
            "message": f"Personality drift {drift:.2f} exceeds threshold 1.0"
        }
    
    return {"valid": True, "message": "Behavioral inertia satisfied"}

Dialog Validators

Dialog Realism

Mechanism: M8 (Embodied States), M11 (Dialog Synthesis)
Severity: WARNING

Checks if dialog respects physical/emotional constraints:

@Validator.register("dialog_realism", severity="WARNING")
def validate_dialog_realism(entity: Entity, context: dict = None) -> dict:
    """Check if dialog respects physical/emotional constraints"""
    
    dialog_data = entity.entity_metadata.get("dialog_data", {})
    turns = json.loads(dialog_data.get("turns", "[]"))
    entities = context.get("entities", [])
    
    validation_issues = []
    
    for i, turn in enumerate(turns):
        speaker_id = turn.get("speaker")
        content = turn.get("content", "")
        
        speaker = next((e for e in entities if e.entity_id == speaker_id), None)
        if not speaker:
            continue
        
        # Apply body-mind coupling
        physical = speaker.physical_tensor
        cognitive = speaker.cognitive_tensor
        coupled_cognitive = couple_pain_to_cognition(physical, cognitive)
        coupled_cognitive = couple_illness_to_cognition(physical, coupled_cognitive)
        
        # Check turn length vs. energy
        content_length = len(content)
        if coupled_cognitive.energy_budget < 30 and content_length > 200:
            validation_issues.append(
                f"{speaker_id} too low energy ({coupled_cognitive.energy_budget:.1f}) "
                f"for long response ({content_length} chars)"
            )
        
        # Check tone vs. emotional state
        emotional_tone = turn.get("emotional_tone", "").lower()
        valence = coupled_cognitive.emotional_valence
        
        if valence < -0.5 and not any(neg in emotional_tone for neg in ["sad", "angry", "frustrated"]):
            validation_issues.append(
                f"{speaker_id} should have negative tone given emotional valence {valence:.2f}"
            )
    
    if validation_issues:
        return {
            "valid": False,
            "message": f"Dialog realism issues: {'; '.join(validation_issues)}"
        }
    
    return {"valid": True, "message": "Dialog respects physical/emotional constraints"}

Example violation:

{
  "validator": "dialog_realism",
  "severity": "WARNING",
  "message": "Dialog realism issues: Webb too low energy (22.5) for long response (342 chars)"
}

Dialog Knowledge Consistency

Mechanism: M3 (Exposure Events), M11 (Dialog Synthesis)
Severity: ERROR

Checks if dialog speakers only reference knowledge they actually have:

@Validator.register("dialog_knowledge_consistency", severity="ERROR")
def validate_dialog_knowledge_consistency(entity: Entity, context: dict = None) -> dict:
    """Check if dialog speakers only reference knowledge they actually have"""
    
    dialog_data = entity.entity_metadata.get("dialog_data", {})
    turns = json.loads(dialog_data.get("turns", "[]"))
    entities = context.get("entities", [])
    
    knowledge_violations = []
    
    for turn in turns:
        speaker_id = turn.get("speaker")
        knowledge_refs = turn.get("knowledge_references", [])
        
        speaker = next((e for e in entities if e.entity_id == speaker_id), None)
        if not speaker:
            continue
        
        speaker_knowledge = set(speaker.entity_metadata.get("knowledge_state", []))
        
        for ref in knowledge_refs:
            if ref not in speaker_knowledge:
                knowledge_violations.append(f"{speaker_id} references unknown knowledge: '{ref}'")
    
    if knowledge_violations:
        return {
            "valid": False,
            "message": f"Knowledge consistency violations: {'; '.join(knowledge_violations)}"
        }
    
    return {"valid": True, "message": "Dialog respects knowledge constraints"}

Example violation:

{
  "validator": "dialog_knowledge_consistency",
  "severity": "ERROR",
  "message": "Knowledge consistency violations: Webb references unknown knowledge: 'Sensor calibration date'"
}

Fix: Add exposure event before dialog:

create_exposure_event(
    entity_id="Webb",
    information="Sensor calibration date",
    source="Chen",
    event_type="told",
    timestamp=timepoint.timestamp,
    store=store
)

Mechanism-Specific Validators

Circadian Plausibility (M14)

Severity: WARNING

Checks if activity is plausible at the given time of day:

@Validator.register("circadian_plausibility", severity="WARNING")
def validate_circadian_activity(entity: Entity, context: dict = None) -> dict:
    """Check if activity is plausible at the given time of day"""
    
    activity = context.get("activity")
    timepoint = context.get("timepoint")
    circadian_config = context.get("circadian_config", {})
    
    hour = timepoint.timestamp.hour
    probability = get_activity_probability(hour, activity, circadian_config)
    
    if probability < 0.05:  # Critical threshold
        return {
            "valid": False,
            "message": f"Activity '{activity}' at {hour:02d}:00 is highly implausible (probability: {probability:.3f})"
        }
    
    return {"valid": True, "message": f"Activity '{activity}' at {hour:02d}:00 is plausible"}

Example:

# Validate "board_meeting" at 3 AM
result = validate_circadian_activity(
    entity,
    context={
        "activity": "board_meeting",
        "timepoint": Timepoint(timestamp=datetime(2026, 3, 6, 3, 0)),
        "circadian_config": {...}
    }
)
# Result: {"valid": False, "message": "Activity 'board_meeting' at 03:00 is highly implausible (probability: 0.02)"}

Prospection Consistency (M15)

Severity: WARNING

Validates that prospective expectations are consistent and realistic:

@Validator.register("prospection_consistency", severity="WARNING")
def validate_prospection_consistency(entity: Entity, context: dict = None) -> dict:
    """Validate that prospective expectations are consistent and realistic"""
    
    prospective_state = context.get("prospective_state")
    expectations = prospective_state.expectations
    
    issues = []
    total_probability = 0
    
    for exp in expectations:
        total_probability += exp.subjective_probability
        
        # Flag unrealistic probabilities
        if exp.subjective_probability > 0.95:
            issues.append(f"Expectation '{exp.predicted_event}' has very high probability ({exp.subjective_probability:.3f})")
        
        # Check confidence vs probability alignment
        if exp.confidence > 0.8 and exp.subjective_probability < 0.2:
            issues.append(f"High confidence ({exp.confidence:.2f}) but low probability ({exp.subjective_probability:.3f})")
    
    # Check total probability isn't unrealistic
    if total_probability > 1.5:
        issues.append(f"Total expectation probabilities ({total_probability:.2f}) suggest unrealistic optimism")
    
    if issues:
        return {"valid": False, "message": f"Prospection consistency issues: {'; '.join(issues)}"}
    
    return {"valid": True, "message": "Prospective expectations appear consistent"}

Environmental Constraints (M16)

Severity: ERROR

Validates that actions respect constraints imposed by animistic entities:

@Validator.register("environmental_constraints", severity="ERROR")
def validate_environmental_constraints(entity: Entity, context: dict = None) -> dict:
    """Validate that actions respect constraints imposed by animistic entities"""
    
    action = context.get("action")
    environment_entities = context.get("environment_entities", [])
    issues = []
    
    for env_entity in environment_entities:
        if env_entity.entity_type == "building":
            building = BuildingEntity(**env_entity.entity_metadata)
            participant_count = action.get("participant_count", 0)
            
            if participant_count > building.capacity:
                issues.append(
                    f"Building {env_entity.entity_id} capacity {building.capacity} "
                    f"exceeded by {participant_count} participants"
                )
            
            if building.structural_integrity < 0.5:
                issues.append(
                    f"Building {env_entity.entity_id} structural integrity too low "
                    f"({building.structural_integrity:.2f}) for use"
                )
    
    if issues:
        return {"valid": False, "message": f"Environmental constraint violations: {'; '.join(issues)}"}
    
    return {"valid": True, "message": "All environmental constraints satisfied"}

Error Severity Levels

ERROR (Fatal)

Blocks simulation from continuing:

• Information conservation violations
• Temporal causality violations
• Biological constraint violations
• Dialog knowledge consistency violations

Behavior: Simulation aborts, requires fix

WARNING (Non-Fatal)

Logs warning but allows simulation to continue:

• Energy budget exceedances (within 20% tolerance)
• Behavioral inertia drift
• Dialog realism issues
• Circadian plausibility concerns

Behavior: Logged, simulation continues

INFO (Informational)

Informational only, no action taken:

• Timeline divergence analysis
• Adaptive entity behavior observations

Behavior: Logged for analysis

Using Validators Programmatically

Validate Single Entity

from validation import Validator

validator = Validator()
result = validator.validate_entity(
    entity=entity,
    context={
        "timepoint_id": "T2",
        "store": store,
        "previous_knowledge": previous_state["knowledge_state"]
    }
)

if not result["valid"]:
    for violation in result["violations"]:
        print(f"[{violation['severity']}] {violation['validator']}: {violation['message']}")

Validate All Entities

for entity in entities:
    result = validator.validate_entity(entity, context)
    if not result["valid"]:
        print(f"Entity {entity.entity_id} has {len(result['violations'])} violations")

Run Specific Validator

from validation import validate_information_conservation

result = validate_information_conservation(
    entity=entity,
    context={"store": store}
)

if not result["valid"]:
    print(f"Violation: {result['message']}")

Best Practices

Enable Validation in Development

Run with validation enabled:

./run.sh run --validate board_meeting

Disable for production:

./run.sh run --no-validate board_meeting

Log Violations

Validation results are logged to logs/validation.log:

2026-03-06 12:34:56 [ERROR] information_conservation: Entity Webb knows about {'O2 fault'} without exposure
2026-03-06 12:35:12 [WARNING] energy_budget: Energy expenditure 127.3 exceeds budget 100

Fix Common Violations

Information conservation:

• Add exposure events before knowledge references
• Use M3 knowledge extraction to track provenance

Temporal causality:

• Check timepoint ordering
• In Portal mode, verify backward inference logic

Energy budget:

• Reduce dialog turn count
• Lower interaction complexity
• Account for circadian penalties

Dialog realism:

• Adjust energy budgets before long conversations
• Sync emotional state before dialog generation

Next Steps

• Learn about Dialog Synthesis to understand context requirements
• Read Templates to configure validation settings
• Explore mechanism-specific validators in Mechanisms