Most AI systems are trained on large datasets drawn from stable internet environments.Real-world crises are different: signals are incomplete, events unfold unpredictably, and meaning must be interpreted in real time.Large-scale datasets capturing these environments are extremely rare. Krisis Labs develops multimodal datasets and analytical frameworks designed to help AI systems understand these environments.Krisis Dataset
The project draws on a long-horizon observational archive documenting crisis environments in real time, including extensive raw rushes from frontline documentary work, alongside decades long cultural narratives and media archives that capture how humans interpret events as they unfold.Krisis Labs develops and licenses multimodal datasets and analytical frameworks to research institutions, AI labs, and robotics companies studying decision-making under uncertainty.Research areas
• multimodal reasoning in uncertain environments
• world-model training data
• narrative and cultural context modeling
• early-signal detection in crisis dynamicsICU Framework
The Krisis Labs datasets form the foundation of the ICU Framework (Intelligence in Crisis & Uncertainty), a proprietary research model exploring how humans interpret signals, assign meaning, and make decisions in unstable environments.The framework organizes multimodal data into three complementary layers:• Judgment under uncertainty
• Meaning under uncertainty
• Signals under uncertaintyDataset ArchitectureJudgment Under UncertaintyFrontline Crisis Dataset
A large, pre-social media observational archive of raw rushes and documentary footage captured in active crisis environments. The material records events unfolding in real time, preserving the uncertainty, repetition, and decision-making processes that shape human behaviour under instability. Portions of this archive have been edited into award-winning documentary films.Use cases
• world-model training
• robotics perception
• decision-making systemsMeaning Under UncertaintyCultural Narrative Dataset
A multimodal archive documenting how filmmakers and cultural creators interpret social change and conflict.
Scale
• 3,500+ films
• 4.5M words
• thousands of imagesUse cases
• narrative reasoning
• cultural context models
• interpretation of eventsSignals Under UncertaintyKrisis Observatory (in development)
A research system designed to capture emerging signals of instability and crisis dynamics.Use cases
• crisis forecasting
• geopolitical risk analysis
• real-time AI training environmentsFounded by an award-winning frontline conflict journalist and PhD researcher whose single-operator documentary work in crisis environments forms the basis of a unique observational archive developed through 30 years of research on perception and decision-making under instability.Founder
Dr. Carmela Baranowska, PhD
Frontline conflict journalist and documentary filmmaker, and the only recipient of both the Rory Peck and Walkley Awards for this work. Her observational archives form the foundation of the Krisis datasets.
Krisis Labs, California, USAPortions of this archive have been edited into award-winning documentary films, including Taliban Country and Scenes from an Occupation. These films are scheduled for re-release on digital platforms including Amazon.