NamicGreen – Science. AI. Journalism.

AIJIM: Evidence-First Newsroom AI — Real-Time, Explainable & Auditable (FTC 2025)

AIJIM (Artificial Intelligence Journalism Integration Model) shows how newsrooms can integrate real-time AI that is model-agnostic, explainable, and auditable. This FTC 2025 talk walks through the motivation, a six-stage workflow, human-in-the-loop validation, and AIJIM’s vision for evidence-first journalism. What you’ll learn Workflow (6 stages): Data → Real-time analysis → Auto-report → Human-in-the-Loop → QA & Explainability → Publication with an Evidence Chain Explainability by default: fast CAM heatmaps; audit-level LIME/SHAP; research directions incl. CASTLE/PASTLE HITL & Crowd verification: 252 validators; fairness, consensus, and IAA Pilot (Mallorca, 2024): 1,000 images → 50 new waste sites identified; 85.4% detection accuracy; ~90% validator agreement; −40% reporting latency R&D roadmap: agentic systems, multimodal LLMs, robustness & adversarial defense; GovTech integration and EU AI Act alignment Resources • Website & demos: https://www.aijim.org • Paper / Proceedings: Future Technologies Conference (FTC) 2025 · Springer LNNS https://link.springer.com/chapter/10.1007/978-3-032-07989-3_26 • Contact & pilots: [email protected] Citation (example) Tiltack, T. (2025). AIJIM: A Scalable Model for Real-Time AI in Environmental Journalism. In: Future Technologies Conference (FTC) 2025, Springer LNNS. Transparency This is the presenter’s own publication of the session; the conference may provide its own recording/edits. Third-party materials are credited or licensed. Hashtags #AIJIM #NewsroomAI #ExplainableAI #EnvironmentalJournalism #EvidenceFirst #ECAMX #AEIDelta #TCA #Reliability #EUAIAct