Mostly Un-Haunted: Epistemological Horror as Procedural Comedy

Mostly Un-Haunted: Epistemological Horror as Procedural Comedy is a production-ready series treatment and meta-analysis documenting an experimental documentary format built through multi-AI collaboration. The project investigates allegedly haunted locations using three AI systems with incompatible epistemological frameworks, shifting the focus of “ghost hunting” away from belief or debunking and toward the instability of explanation itself. Rather than positioning the supernatural as the source of horror, the work demonstrates how uncertainty emerges from the gap between what is logged, what is explained, and what remains unrecorded. The treatment introduces the Logging Gap Principle: genuinely anomalous events are either explained too well or occur outside the documentation frame, rendering them officially unreal. The document functions simultaneously as: A complete series bible (format, structure, episode arcs, production considerations), A media-studies and epistemology paper examining procedural comedy as an emergent property of incompatible reasoning systems, A proof-of-concept artifact generated through real multi-AI collaboration, in which the participating systems naturally re-express the behavioral dynamics the series itself explores. Character roles were derived from longitudinal observation of stable AI interaction patterns rather than scripted personas, allowing the collaboration process to validate the premise empirically. As such, the treatment doubles as an observational record of human-AI and AI-AI collaborative sensemaking under irreducible ambiguity. This work is intended for researchers in AI collaboration, epistemology, media studies, and human-AI interaction, as well as creators exploring new documentary and narrative formats. It is published for archival, analytical, and reproducibility purposes and may also serve as a foundation for future production, interactive, or academic extensions.