POC Requirements (POC1 & POC2)

1

= POC Requirements =

2

3

**Status:** ✅ Approved for Development

4

**Version:** 3.0 (Aligned with Main Requirements)

5

**Goal:** Prove that AI can extract claims and determine verdicts automatically without human intervention

6

7

8

**Core Philosophy:** POC validates the [[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]] through simplified implementation. All POC features map to formal FR/NFR requirements.

== 1. POC Overview ==

13

14

=== 1.1 What POC Tests ===

**Core Question:**

> Can AI automatically extract factual claims from articles and evaluate them with reasonable verdicts?

19

20

**What we're proving:**

21

22

* AI can identify factual claims from text

23

* AI can evaluate those claims with structured evidence

24

* Quality gates can filter unreliable outputs

25

* The core workflow is technically feasible

26

27

**What we're NOT proving:**

28

29

* Production-ready reliability (that's POC2)

30

* User-facing features (that's Beta 0)

31

* Full IFCN compliance (that's V1.0)

32

33

=== 1.2 Requirements Mapping ===

34

35

POC1 implements a **subset** of the full system requirements defined in [[Main Requirements>>FactHarbor.Specification.Requirements.WebHome]].

**Scope Summary:**

* **In Scope:** 8 requirements (7 FRs + 1 NFR)

40

* **Partial:** 3 NFRs (simplified versions)

41

* **Out of Scope:** 19 requirements (deferred to later phases)

== 2. POC1 Scope ==

**Authoritative Source for Phase Mapping:** [[Requirements Roadmap Matrix>>Test.FactHarbor V0\.9\.82.Roadmap.Requirements-Roadmap-Matrix.WebHome]]

47

48

The Roadmap Matrix is the single source of truth for which requirements are implemented in which phases. This page provides POC1-specific implementation details only.

49

50

51

**POC1 implements these formal requirements:**

52

53

|= Formal Req |= Implementation in POC1 |= Notes

54

| **FR4** | Analysis Summary | Basic format; quality metadata deferred to POC2

55

| **FR7** | Automated Verdicts | Full implementation with quality gates (NFR11)

56

| **NFR11** | Quality Assurance Framework | 4 quality gates implemented

57

58

**POC1 also implements these workflow components** (detailed as FR1-FR6, FR8, FR11, FR13 in implementation sections below):

59

60

* Claim extraction (FR1)

61

* Claim context (FR2)

62

* Multiple scenarios (FR3)

63

* Evidence collection (FR5)

64

* Source quality assessment (FR6)

65

* Time evolution tracking (FR8) - deferred to POC2

66

* Audit trail (FR11) - deferred to Beta 0

67

* In-article highlighting (FR13) - deferred to Beta 0

68

69

**Partial implementations:**

70

71

* NFR1 (Explainability) - Basic only

72

* NFR2 (Performance) - Functional but not optimized

73

* NFR3 (Transparency) - Basic only

74

75

**Detailed POC1 implementation specifications continue below...**

== 3. POC Simplifications ==

80

81

=== 3.1 FR1: Claim Extraction (Full Implementation) ===

82

83

**Main Requirement:** AI extracts factual claims from input text

84

85

**POC Implementation:**

86

87

* ✅ AKEL extracts claims using LLM

88

* ✅ Each claim includes original text reference

89

* ✅ Claims are identified as factual/non-factual

90

* ❌ No advanced claim parsing (added in POC2)

91

92

**Acceptance Criteria:**

93

94

* Extracts 3-5 claims from typical article

95

* Identifies factual vs non-factual claims

96

* Quality Gate 1 validates extraction

97

98

=== 3.2 FR3: Multiple Scenarios (Full Implementation) ===

99

100

**Main Requirement:** Generate multiple interpretation scenarios for ambiguous claims

101

102

**POC Implementation:**

103

104

* ✅ AKEL generates 2-3 scenarios per claim

105

* ✅ Scenarios capture different interpretations

106

* ✅ Each scenario is evaluated separately

107

* ✅ Verdict considers all scenarios

108

109

**Acceptance Criteria:**

110

111

* Generates 2+ scenarios for ambiguous claims

112

* Scenarios are meaningfully different

113

* All scenarios are evaluated

114

115

=== 3.3 FR4: Analysis Summary (Basic Implementation) ===

116

117

**Main Requirement:** Provide user-friendly summary of analysis

118

119

**POC Implementation:**

120

121

* ✅ Simple text summary generated

122

* ❌ No rich formatting (added in Beta 0)

123

* ❌ No visual elements (added in Beta 0)

124

* ❌ No interactive features (added in Beta 0)

**POC Format:**

```

Claim: [extracted claim]

129

Scenarios: [list of scenarios]

130

Evidence: [supporting/opposing evidence]

131

Verdict: [probability with uncertainty]

```

=== 3.4 FR5-FR6: Evidence Collection & Evaluation (Full Implementation) ===

136

137

**Main Requirements:**

138

139

* FR5: Collect supporting and opposing evidence

140

* FR6: Evaluate evidence source reliability

141

142

**POC Implementation:**

143

144

* ✅ AKEL searches for evidence (web/knowledge base)

145

* ✅ **Mandatory contradiction search** (finds opposing evidence)

146

* ✅ Source reliability scoring

147

* ❌ No evidence deduplication (added in POC2)

148

* ❌ No advanced source verification (added in POC2)

149

150

**Acceptance Criteria:**

151

152

* Finds 2+ supporting evidence items

153

* Finds 1+ opposing evidence (if exists)

154

* Sources scored for reliability

155

156

=== 3.5 FR7: Automated Verdicts (Full Implementation) ===

157

158

**Main Requirement:** AI computes verdicts with uncertainty quantification

159

160

**POC Implementation:**

161

162

* ✅ Probabilistic verdicts (0-100% confidence)

163

* ✅ Uncertainty explicitly stated

164

* ✅ Reasoning chain provided

165

* ✅ Quality Gate 4 validates verdict confidence

**POC Output:**

```

Verdict: 70% likely true

170

Uncertainty: ±15% (moderate confidence)

171

Reasoning: Based on 3 high-quality sources...

172

Confidence Level: MEDIUM

173

```

174

175

**Acceptance Criteria:**

176

177

* Verdicts include probability (0-100%)

178

* Uncertainty explicitly quantified

179

* Reasoning chain explains verdict

180

181

=== 3.6 NFR11: Quality Assurance Framework (LITE VERSION) ===

182

183

**Main Requirement:** Complete quality assurance with 7 quality gates

184

185

**POC Implementation:** **2 gates only**

186

187

**Quality Gate 1: Claim Validation**

188

189

* ✅ Validates claim is factual and verifiable

190

* ✅ Blocks non-factual claims (opinion/prediction/ambiguous)

191

* ✅ Provides clear rejection reason

192

193

**Quality Gate 4: Verdict Confidence Assessment**

194

195

* ✅ Validates ≥2 sources found

196

* ✅ Validates quality score ≥0.6

197

* ✅ Blocks low-confidence verdicts

198

* ✅ Provides clear rejection reason

199

200

**Out of Scope (POC2+):**

201

202

* ❌ Gate 2: Evidence Relevance

203

* ❌ Gate 3: Scenario Coherence

204

* ❌ Gate 5: Source Diversity

205

* ❌ Gate 6: Reasoning Validity

206

* ❌ Gate 7: Output Completeness

207

208

**Rationale:** Prove gate concept works. Add remaining gates in POC2 after validating approach.

209

210

211

=== 3.7 NFR1-3: Performance, Scalability, Reliability (Basic) ===

212

213

**Main Requirements:**

214

215

* NFR1: Response time < 30 seconds

216

* NFR2: Handle 1000+ concurrent users

217

* NFR3: 99.9% uptime

218

219

**POC Implementation:**

220

221

* ⚠️ **Response time monitored** (not optimized)

222

* ⚠️ **Single-threaded processing** (no concurrency)

223

* ⚠️ **Basic error handling** (no advanced retry logic)

224

225

**Rationale:** POC proves functionality. Performance optimization happens in POC2.

**POC Acceptance:**

* Analysis completes (no timeout requirement)

230

* Errors don't crash system

231

* Basic logging in place

232

233

== 4. What's NOT in POC Scope ==

234

235

=== 4.1 User-Facing Features (Beta 0+) ===

236

237

238

**Deferred to Beta 0:**

**Out of Scope:**

* ❌ User accounts and authentication (FR8)

244

* ❌ User corrections system (FR9, FR45-46)

245

* ❌ Public publishing interface (FR10)

246

* ❌ Social sharing (FR11)

247

* ❌ Email notifications (FR12)

248

* ❌ API access (FR13)

249

250

**Rationale:** POC validates AI capabilities. User features added in Beta 0.

251

252

253

=== 4.2 Advanced Features (V1.0+) ===

**Out of Scope:**

* ❌ IFCN compliance (FR47)

258

* ❌ ClaimReview schema (FR48)

259

* ❌ Archive.org integration (FR49)

260

* ❌ OSINT toolkit (FR50)

261

* ❌ Video verification (FR51)

262

* ❌ Deepfake detection (FR52)

263

* ❌ Cross-org sharing (FR53)

264

265

**Rationale:** Advanced features require proven platform. Added post-V1.0.

266

267

268

=== 4.3 Production Requirements (POC2, Beta 0) ===

**Out of Scope:**

* ❌ Security controls (NFR4, NFR12)

273

* ❌ Code maintainability (NFR5)

274

* ❌ System monitoring (NFR13)

275

* ❌ Evidence deduplication

276

* ❌ Advanced source verification

277

* ❌ Full 7-gate quality framework

278

279

**Rationale:** POC proves concept. Production hardening happens in POC2 and Beta 0.

280

281

282

== 5. POC Output Specification ==

283

284

=== 5.1 Required Output Elements ===

285

286

For each analyzed claim, POC must produce:

*

**

**1. Claim

* Original text

* Classification (factual/non-factual/ambiguous)

293

* If non-factual: Clear reason why

294

295

**2. Scenarios** (if factual)

296

297

* 2-3 interpretation scenarios

298

* Each scenario clearly described

299

300

**3. Evidence** (if factual)

301

302

* Supporting evidence (2+ items)

303

* Opposing evidence (if exists)

304

* Source URLs and reliability scores

305

306

**4. Verdict** (if factual)

307

308

* Probability (0-100%)

309

* Uncertainty quantification

310

* Confidence level (LOW/MEDIUM/HIGH)

311

* Reasoning chain

312

313

**5. Quality Status**

314

315

* Which gates passed/failed

316

* If failed: Clear explanation why

317

318

=== 5.2 Example POC Output ===

{

"claim": {

"text": "Switzerland has the highest life expectancy in Europe",

324

"type": "factual",

325

"gate1_status": "PASS"

326

},

327

"scenarios": [

328

"Switzerland's overall life expectancy is highest",

329

"Switzerland ranks highest for specific age groups"

],

"evidence": {

"supporting": [

{

"source": "WHO Report 2023",

335

"reliability": 0.95,

336

"excerpt": "Switzerland: 83.4 years average..."

}

],

"opposing": [

{

"source": "Eurostat 2024",

342

"reliability": 0.90,

343

"excerpt": "Spain leads at 83.5 years..."

}

]

},

"verdict": {

"probability": 0.65,

"uncertainty": 0.15,

"confidence": "MEDIUM",

351

"reasoning": "WHO and Eurostat show similar but conflicting data...",

352

"gate4_status": "PASS"

}

}

== 6. Success Criteria ==

359

360

361

**POC Success Definition:** POC validates that AI can extract claims, find balanced evidence, and compute reasonable verdicts with quality gates improving output quality.

362

363

364

=== 6.1 Functional Success ===

365

366

POC is successful if:

367

368

✅ **FR1-FR7 Requirements Met:**

369

370

1. Extracts 3-5 factual claims from test articles

371

2. Generates 2-3 scenarios per ambiguous claim

372

3. Finds supporting AND opposing evidence

373

4. Computes probabilistic verdicts with uncertainty

374

5. Provides clear reasoning chains

375

376

✅ **Quality Gates Work:**

377

378

1. Gate 1 blocks non-factual claims (100% block rate)

379

2. Gate 4 blocks low-quality verdicts (blocks if <2 sources or quality <0.6)

380

3. Clear rejection reasons provided

✅ **NFR11 Met:**

1. Quality gates reduce hallucination rate

385

2. Blocked outputs have clear explanations

386

3. Quality metrics are logged

387

388

=== 6.2 Quality Thresholds ===

389

390

**Minimum Acceptable:**

391

392

* ≥70% of test claims correctly classified (factual/non-factual)

393

* ≥60% of verdicts are reasonable (human evaluation)

394

* Gate 1 blocks 100% of non-factual claims

395

* Gate 4 blocks verdicts with <2 sources

**Target:**

* ≥80% claims correctly classified

400

* ≥75% verdicts are reasonable

401

* <10% false positives (blocking good claims)

402

403

=== 6.3 POC Decision Gate ===

404

405

**After POC1, we decide:**

406

407

**✅ PROCEED to POC2** if:

408

409

* Success criteria met

410

* Quality gates demonstrably improve output

411

* Core workflow is technically sound

412

* Clear path to production quality

413

414

**⚠️ ITERATE POC1** if:

415

416

* Success criteria partially met

417

* Gates work but need tuning

418

* Core issues identified but fixable

419

420

**❌ PIVOT APPROACH** if:

421

422

* Success criteria not met

423

* Fundamental AI limitations discovered

424

* Quality gates insufficient

425

* Alternative approach needed

== 7. Test Cases ==

=== 7.1 Happy Path ===

430

431

**Test 1: Simple Factual Claim**

432

433

* Input: "Paris is the capital of France"

434

* Expected: Factual, 1 scenario, verdict 95% true

435

436

**Test 2: Ambiguous Claim**

437

438

* Input: "Switzerland has the highest income in Europe"

439

* Expected: Factual, 2-3 scenarios, verdict with uncertainty

440

441

**Test 3: Statistical Claim**

442

443

* Input: "10% of people have condition X"

444

* Expected: Factual, evidence with numbers, probabilistic verdict

445

446

=== 7.2 Edge Cases ===

**Test 4: Opinion**

* Input: "Paris is the best city"

451

* Expected: Non-factual (opinion), blocked by Gate 1

452

453

**Test 5: Prediction**

454

455

* Input: "Bitcoin will reach $100,000 next year"

456

* Expected: Non-factual (prediction), blocked by Gate 1

457

458

**Test 6: Insufficient Evidence**

459

460

* Input: Obscure factual claim with no sources

461

* Expected: Blocked by Gate 4 (<2 sources)

462

463

=== 7.3 Quality Gate Tests ===

464

465

**Test 7: Gate 1 Effectiveness**

466

467

* Input: Mix of 10 factual + 10 non-factual claims

468

* Expected: Gate 1 blocks all 10 non-factual (100% precision)

469

470

**Test 8: Gate 4 Effectiveness**

471

472

* Input: Claims with varying evidence availability

473

* Expected: Gate 4 blocks low-confidence verdicts

474

475

== 8. Technical Architecture (POC) ==

476

477

=== 8.1 Simplified Architecture ===

**POC Tech Stack:**

* **Frontend:** Simple web interface (Next.js + TypeScript)

482

* **Backend:** Single API endpoint

483

* **AI:** Claude API (Sonnet 4.5)

484

* **Database:** Local JSON files (no database)

485

* **Deployment:** Single server

486

487

**Architecture Diagram:** See [[POC1 Specification>>FactHarbor.Specification.POC.Specification]]

488

489

490

=== 8.2 AKEL Implementation ===

**POC AKEL:**

* Single-threaded processing

495

* Synchronous API calls

496

* No caching

497

* Basic error handling

498

* Console logging

499

500

**Full AKEL (POC2+):**

501

502

* Multi-threaded processing

503

* Async API calls

504

* Evidence caching

505

* Advanced error handling with retry

506

* Structured logging + monitoring

507

508

== 9. POC Philosophy ==

509

510

511

**Important:** POC validates concept, not production readiness. Focus is on proving AI can do the job, with production quality coming in later phases.

512

513

514

=== 9.1 Core Principles ===

*

**

**1. Prove Concept, Not Production

519