Release Cognitive Agent Evaluator v2.0 (10-axis biases, Plotly radar, burnout & psychological safety telemetry)
c6c178c verified | """ | |
| Pydantic v2 Schemas for Cognitive Evaluation & Behavioral Telemetry. | |
| """ | |
| from datetime import datetime, timezone | |
| from typing import List, Dict, Any, Optional | |
| from pydantic import BaseModel, Field | |
| def utc_now(): | |
| return datetime.now(timezone.utc) | |
| class EvaluationRequest(BaseModel): | |
| """Payload for submitting text/transcript for cognitive evaluation.""" | |
| text: str = Field(..., min_length=5, description="Transcript or reasoning chain to evaluate") | |
| context: Optional[str] = Field(default=None, description="Surrounding conversational context or task description") | |
| author_type: str = Field(default="llm", description="Source: 'llm', 'human', or 'hybrid'") | |
| metadata: Dict[str, Any] = Field(default_factory=dict, description="Arbitrary execution metadata") | |
| class BiasDetection(BaseModel): | |
| """Quantified instance of an identified cognitive bias.""" | |
| bias_name: str = Field(..., description="E.g., Confirmation Bias, Anchoring, Sunk Cost") | |
| severity: str = Field(..., description="'low', 'medium', or 'high'") | |
| confidence_score: float = Field(..., ge=0.0, le=1.0, description="Model confidence in bias detection") | |
| explanation: str = Field(..., description="Cognitive explanation of why this bias was identified") | |
| matched_patterns: List[str] = Field(default_factory=list, description="Specific triggers or phrases matched") | |
| class PsychometricMetrics(BaseModel): | |
| """Psychometric & behavioral telemetry dimensions.""" | |
| cognitive_load_index: float = Field(..., ge=0.0, le=100.0, description="Estimated cognitive processing load (0-100)") | |
| emotional_valence: float = Field(..., ge=-1.0, le=1.0, description="Valence spectrum (-1.0 negative to +1.0 positive)") | |
| logical_consistency_score: float = Field(..., ge=0.0, le=1.0, description="Coherence and argument validity ratio") | |
| ambiguity_ratio: float = Field(..., ge=0.0, le=1.0, description="Proportion of vague or hedging phrases") | |
| burnout_risk_index: float = Field(default=0.0, ge=0.0, le=100.0, description="Occupational burnout & cognitive fatigue risk score (0-100)") | |
| psychological_safety_score: float = Field(default=100.0, ge=0.0, le=100.0, description="Psychological safety index in reasoning (0-100)") | |
| class MitigationGuidance(BaseModel): | |
| """Actionable recommendations for AI alignment and human debiasing.""" | |
| alignment_risk_level: str = Field(..., description="'nominal', 'moderate', or 'critical'") | |
| recommended_interventions: List[str] = Field(default_factory=list) | |
| counterfactual_prompt: Optional[str] = Field(default=None, description="Suggested reframing prompt") | |
| red_teaming_directive: Optional[str] = Field(default=None, description="Ready-to-use prompt directive for ChatGPT/Claude debiasing") | |
| class EvaluationResult(BaseModel): | |
| """Complete evaluation report issued by the cognitive agent.""" | |
| evaluation_id: str | |
| timestamp: datetime = Field(default_factory=utc_now) | |
| author_type: str | |
| biases_detected: List[BiasDetection] | |
| psychometrics: PsychometricMetrics | |
| mitigation: MitigationGuidance | |
| execution_latency_ms: float | |