🚀 awarelabs.dev — Main developer hub for Aware AI & Computational Watchdog Systems.Explore Interactive Demo →
AWARE ARCHITECTURE SPECIFICATION

Who Watches the Watchmen?

In a media ecosystem where international wires dictate the African narrative, Aware AI deploys an autonomous multi-agent swarm with tiered reasoning to audit media bias with verifiable mathematical rigor.

The 5 Specialized Forensic Agents

Tiered Apodex 1.1 Mini LLM
Node 01Core Fact Extractor

Ground Truth Agent

Cross-examines 1,000+ monitored African and international publishers to build a verified chronological event timeline stripped of editorial spin and emotional adjectives.

Fact Timeline Index (Fc)
Mandate Steps:
  • Ingest raw report
  • Entity & chronological extraction
  • Cross-verify with Pan-African wire databases
Node 02Sentiment & Spin Scanner

Lexical Auditor

Performs dense sentence-by-sentence NLP evaluation, calculating adjective-to-noun ratios, passive voice obfuscation, loaded adjectives, and narrative fluff coefficients.

Spin Coefficient (Sj) & Fluff Index (Nf)
Mandate Steps:
  • Adjective density scan
  • Passive voice attribution check
  • Compute loaded phrasing score
Node 03Omission & Perspective Detector

Context Analyst

Identifies absent historical context, missing economic statistics, and neglected African local stakeholder quotes that fundamentally slant reporting.

Omission Penalty (Om)
Mandate Steps:
  • Compare with historical archives
  • Check local community quotes
  • Flag unverified foreign government claims
Node 04Wire Syndication Auditor

Parrot Detector

Measures string distance and semantic parroting against Western syndication wires (Reuters, AFP, AP). Highlights articles that copy wire narrative without independent local inquiry.

Parrot Similarity Index (Spp)
Mandate Steps:
  • String distance vector analysis
  • Syndication threshold test (>85%)
  • Original commentary scoring
Node 05Final Story Score Synthesizer

Chief Editor Consensus

Aggregates the individual agent evaluations through a weighted mathematical consensus equation to assign the definitive Story Score between 0.0 and 10.0.

Score = w1·Fc + w2·(1-Sj) + w3·(1-Om) + w4·(1-Spp)
Mandate Steps:
  • Weighted Bayesian aggregation
  • Confidence interval computation
  • Publishing to public newsroom & API
Mathematical Transparency

The Story Score Equation

Unlike proprietary black-box media ratings, every parameter in the Aware Story Score is published openly for academic and public scrutiny.

S = w₁·F_c + w₂·(1 - S_j) + w₃·(1 - O_m) + w₄·(1 - S_pp)

Where F_c = Fact Timeline, S_j = Spin Coefficient, O_m = Omission Penalty, S_pp = Parrot Wire Dependency.

9.0 – 10.0: Exemplary Journalism (Zero spin, all perspectives)
7.0 – 8.9: Verified Reporting (Minor narrative framing)
5.0 – 6.9: Moderate Bias (Noticeable omissions)
< 5.0: Heavy Spin / Propaganda (Severe distortions)

Read the Technical Whitepapers

Published under Creative Commons CC BY 4.0 in our open documentation portal.