The Human Ball-Strike System (HBS) is a lightweight interactive course designed to activate human judgment before an AI-generated summary enters a consequential institutional record personnel files, committee minutes, accreditation reports. The course covers six AI summarization failure modes (overgeneralization, statistical misrepresentation, context displacement, flattening, anchor frame shift, hallucination), three live synthetic examples across different corpus types, and five governance principles drawn from Purdue Universitys Data Ethics Committee model. A companion AI Summary Watermark Tool generates a provenance label modeled on ANSI Z535 industrial warning conventions that travels with the document into whatever workflow receives it. The tool requires no login and no external dependencies. The course includes a Wikipedia dissemination pathway allowing practitioners to contribute documented failure mode examples to the public record under open license. Developed at the University of Pittsburgh, May 2026. CC BY-NC 4.0.
Gordon Mitchell (Fri,) studied this question.