ThoughtMCP

/** * Unit tests for UncertaintyQuantifier */ import { beforeEach, describe, expect, it } from "vitest"; import { UncertaintyQuantifier } from "../../cognitive/UncertaintyQuantifier.js"; import { DataPoint, KnowledgeState, ReasoningStep, } from "../../interfaces/probabilistic-reasoning.js"; describe("UncertaintyQuantifier", () => { let quantifier: UncertaintyQuantifier; beforeEach(() => { quantifier = new UncertaintyQuantifier(); }); describe("assessEpistemicUncertainty", () => { it("should assess epistemic uncertainty for complete knowledge", () => { const knowledge: KnowledgeState = { known_facts: ["fact1", "fact2", "fact3"], uncertain_beliefs: new Map([["belief1", 0.8]]), knowledge_gaps: [], confidence_in_knowledge: 0.9, last_updated: Date.now(), }; const uncertainty = quantifier.assessEpistemicUncertainty(knowledge); expect(uncertainty).toBeGreaterThan(0); expect(uncertainty).toBeLessThan(1); expect(uncertainty).toBeLessThan(0.5); // Should be low for complete knowledge }); it("should assess high epistemic uncertainty for incomplete knowledge", () => { const knowledge: KnowledgeState = { known_facts: ["fact1"], uncertain_beliefs: new Map([ ["belief1", 0.3], ["belief2", 0.4], ]), knowledge_gaps: ["gap1", "gap2", "gap3"], confidence_in_knowledge: 0.3, last_updated: Date.now() - 24 * 60 * 60 * 1000, // 24 hours ago }; const uncertainty = quantifier.assessEpistemicUncertainty(knowledge); expect(uncertainty).toBeGreaterThan(0.5); // Should be high for incomplete knowledge expect(uncertainty).toBeLessThan(1); }); it("should handle empty knowledge state", () => { const knowledge: KnowledgeState = { known_facts: [], uncertain_beliefs: new Map(), knowledge_gaps: [], confidence_in_knowledge: 0.1, last_updated: Date.now(), }; const uncertainty = quantifier.assessEpistemicUncertainty(knowledge); expect(uncertainty).toBeGreaterThan(0.5); // Should be high for empty knowledge expect(uncertainty).toBeLessThan(1); }); }); describe("assessAleatoricUncertainty", () => { it("should assess aleatoric uncertainty for consistent data", () => { const data: DataPoint[] = [ { value: 0.8, uncertainty: 0.1, source: "source1", timestamp: Date.now(), reliability: 0.9, }, { value: 0.82, uncertainty: 0.12, source: "source2", timestamp: Date.now(), reliability: 0.85, }, { value: 0.78, uncertainty: 0.08, source: "source3", timestamp: Date.now(), reliability: 0.95, }, ]; const uncertainty = quantifier.assessAleatoricUncertainty(data); expect(uncertainty).toBeGreaterThan(0); expect(uncertainty).toBeLessThan(1); expect(uncertainty).toBeLessThan(0.5); // Should be low for consistent data }); it("should assess high aleatoric uncertainty for variable data", () => { const data: DataPoint[] = [ { value: 0.1, uncertainty: 0.4, source: "source1", timestamp: Date.now(), reliability: 0.3, }, { value: 0.9, uncertainty: 0.6, source: "source2", timestamp: Date.now(), reliability: 0.4, }, { value: 0.5, uncertainty: 0.8, source: "source3", timestamp: Date.now(), reliability: 0.2, }, ]; const uncertainty = quantifier.assessAleatoricUncertainty(data); expect(uncertainty).toBeGreaterThan(0.3); // Should be higher for variable data expect(uncertainty).toBeLessThan(1); }); it("should handle empty data array", () => { const uncertainty = quantifier.assessAleatoricUncertainty([]); expect(uncertainty).toBe(0.5); // Default uncertainty for no data }); it("should handle single data point", () => { const data: DataPoint[] = [ { value: 0.7, uncertainty: 0.2, source: "source1", timestamp: Date.now(), reliability: 0.8, }, ]; const uncertainty = quantifier.assessAleatoricUncertainty(data); expect(uncertainty).toBeGreaterThan(0); expect(uncertainty).toBeLessThan(1); }); }); describe("combineUncertainties", () => { it("should combine epistemic and aleatoric uncertainties", () => { const epistemic = 0.3; const aleatoric = 0.4; const combined = quantifier.combineUncertainties(epistemic, aleatoric); expect(combined).toBeDefined(); expect(combined.total_uncertainty).toBeGreaterThan(0); expect(combined.total_uncertainty).toBeLessThan(1); expect(combined.epistemic_component).toBe(epistemic); expect(combined.aleatoric_component).toBe(aleatoric); expect(combined.interaction_effects).toBeGreaterThanOrEqual(0); expect(combined.confidence_bounds).toHaveLength(2); expect(combined.confidence_bounds[0]).toBeLessThan( combined.confidence_bounds[1] ); }); it("should handle zero uncertainties", () => { const combined = quantifier.combineUncertainties(0, 0); expect(combined.total_uncertainty).toBeGreaterThanOrEqual(0); // Should have minimum uncertainty expect(combined.interaction_effects).toBe(0); }); it("should handle maximum uncertainties", () => { const combined = quantifier.combineUncertainties(0.95, 0.95); expect(combined.total_uncertainty).toBeLessThan(1); // Should not exceed 1 expect(combined.interaction_effects).toBeGreaterThan(0); }); }); describe("trackConfidenceEvolution", () => { it("should track confidence evolution through reasoning steps", () => { const reasoning: ReasoningStep[] = [ { id: "step1", type: "premise", content: "Initial premise", confidence: 0.8, uncertainty: 0.2, evidence_basis: [], logical_form: "premise", alternatives: [], }, { id: "step2", type: "inference", content: "First inference", confidence: 0.7, uncertainty: 0.3, evidence_basis: [], logical_form: "inference", alternatives: [], }, { id: "step3", type: "inference", content: "Second inference", confidence: 0.6, uncertainty: 0.4, evidence_basis: [], logical_form: "inference", alternatives: [], }, { id: "step4", type: "conclusion", content: "Final conclusion", confidence: 0.65, uncertainty: 0.35, evidence_basis: [], logical_form: "conclusion", alternatives: [], }, ]; const evolution = quantifier.trackConfidenceEvolution(reasoning); expect(evolution).toBeDefined(); expect(evolution.initial_confidence).toBe(0.8); expect(evolution.final_confidence).toBe(0.65); expect(evolution.confidence_trajectory).toHaveLength(4); expect(evolution.confidence_trajectory).toEqual([0.8, 0.7, 0.6, 0.65]); expect(evolution.confidence_changes).toBeDefined(); expect(Array.isArray(evolution.confidence_changes)).toBe(true); expect(evolution.stability_measure).toBeGreaterThan(0); expect(evolution.stability_measure).toBeLessThanOrEqual(1); }); it("should handle empty reasoning steps", () => { const evolution = quantifier.trackConfidenceEvolution([]); expect(evolution.initial_confidence).toBe(0.5); expect(evolution.final_confidence).toBe(0.5); expect(evolution.confidence_trajectory).toEqual([0.5]); expect(evolution.confidence_changes).toHaveLength(0); expect(evolution.stability_measure).toBe(1.0); }); it("should identify significant confidence changes", () => { const reasoning: ReasoningStep[] = [ { id: "step1", type: "premise", content: "Initial premise", confidence: 0.5, uncertainty: 0.5, evidence_basis: [], logical_form: "premise", alternatives: [], }, { id: "step2", type: "inference", content: "Strong evidence found", confidence: 0.9, // Significant increase uncertainty: 0.1, evidence_basis: [ { id: "e1", content: "Strong evidence", type: "observational", reliability: 0.9, relevance: 0.9, timestamp: Date.now(), source: "test", weight: 1.0, }, ], logical_form: "inference", alternatives: [], }, ]; const evolution = quantifier.trackConfidenceEvolution(reasoning); expect(evolution.confidence_changes.length).toBeGreaterThan(0); const change = evolution.confidence_changes[0]; expect(change.old_confidence).toBe(0.5); expect(change.new_confidence).toBe(0.9); expect(change.reason).toContain("evidence"); }); it("should calculate stability measure correctly", () => { // Stable confidence trajectory const stableReasoning: ReasoningStep[] = Array.from( { length: 10 }, (_, i) => ({ id: `step${i + 1}`, type: "inference" as const, content: `Step ${i + 1}`, confidence: 0.8 + (Math.random() - 0.5) * 0.02, // Small variations around 0.8 uncertainty: 0.2, evidence_basis: [], logical_form: "inference", alternatives: [], }) ); const stableEvolution = quantifier.trackConfidenceEvolution(stableReasoning); expect(stableEvolution.stability_measure).toBeGreaterThan(0.8); // Unstable confidence trajectory const unstableReasoning: ReasoningStep[] = Array.from( { length: 10 }, (_, i) => ({ id: `step${i + 1}`, type: "inference" as const, content: `Step ${i + 1}`, confidence: Math.random(), // Random variations uncertainty: 0.2, evidence_basis: [], logical_form: "inference", alternatives: [], }) ); const unstableEvolution = quantifier.trackConfidenceEvolution(unstableReasoning); expect(unstableEvolution.stability_measure).toBeLessThan(0.9); }); }); describe("edge cases", () => { it("should handle extreme values gracefully", () => { const extremeKnowledge: KnowledgeState = { known_facts: Array.from({ length: 1000 }, (_, i) => `fact${i}`), uncertain_beliefs: new Map(), knowledge_gaps: Array.from({ length: 1000 }, (_, i) => `gap${i}`), confidence_in_knowledge: 0, last_updated: 0, }; const uncertainty = quantifier.assessEpistemicUncertainty(extremeKnowledge); expect(uncertainty).toBeGreaterThan(0); expect(uncertainty).toBeLessThan(1); }); it("should handle very recent knowledge updates", () => { const recentKnowledge: KnowledgeState = { known_facts: ["fact1"], uncertain_beliefs: new Map(), knowledge_gaps: [], confidence_in_knowledge: 0.9, last_updated: Date.now(), // Very recent }; const uncertainty = quantifier.assessEpistemicUncertainty(recentKnowledge); expect(uncertainty).toBeLessThan(0.5); // Should be lower for recent knowledge }); it("should handle very old knowledge updates", () => { const oldKnowledge: KnowledgeState = { known_facts: ["fact1"], uncertain_beliefs: new Map(), knowledge_gaps: [], confidence_in_knowledge: 0.9, last_updated: Date.now() - 365 * 24 * 60 * 60 * 1000, // 1 year ago }; const uncertainty = quantifier.assessEpistemicUncertainty(oldKnowledge); expect(uncertainty).toBeGreaterThan(0.2); // Should be higher for old knowledge }); }); });