Studying Comics at Scale with Artificial Intelligence: The PAGE Framework and Three Studies of American Comics and Japanese Manga

TEXT START: Comics are one of the great visual records of the twentieth century: hundreds of thousands of pages in which a mass readership saw its leaders, its wars and its heroes.

The Dissection

The paper is not merely proposing a research tool. It is presenting AI-mediated interpretation as a benevolent extension of scholarship: faster archival search, scalable annotation, accessibility for blind and low-vision readers, preservation, and recognition for uncredited artists.

PAGE industrializes the full interpretive pipeline—parsing, labeling, sampling, comparison, and explanation. The human coder becomes the source of the codebook and validation sample; the machine performs the archive-wide labor. That is the important fact buried beneath the humanitarian language. The framework converts a slow expert practice into a reproducible, scalable production system.

The Core Fallacy

The paper confuses increased analytical capability with preserved human economic necessity.

PAGE may make comics research better, cheaper, and broader. Under the Discontinuity Thesis, that is precisely the threat. Once a model can perform visual reading, textual retrieval, classification, attribution, and descriptive accessibility work across thousands of pages, the demand for the human labor previously required to perform those functions contracts. Human annotation is presented as a safeguard, but sampling and calibration are not a moat against automation; they are the procedure by which automation becomes trustworthy enough to remove the bulk of the labor.

The framework also mistakes cultural preservation for social survival. An archive can become more searchable while the people who once catalogued, interpreted, translated, described, and taught it become less economically necessary. The medium is preserved. The labor circuit is not.

Hidden Assumptions

• Human experts will remain economically central because they design the codebook, even after models execute the codebook at archive scale.
• Human validation will remain extensive rather than shrinking to periodic quality control.
• Model outputs will retain scholarly legitimacy without creating a new concentration of ownership over the models, datasets, and infrastructure.
• More accessible archives will generate enough demand to offset the labor displaced by automated interpretation.
• “Explaining change” across an archive is treated as a neutral technical act, although the codebook determines what counts as visible, comparable, or historically meaningful.
• Credit for uncredited artists can be recovered automatically without transferring new gatekeeping power to whoever controls the recognition system.
• The benefits of scale will accrue to scholars and the public rather than primarily to institutions that own the computational stack and the digitized collections.
• Generative systems can extend human scholarship without reorganizing scholarship around machine-readable categories and machine-priced labor.

Social Function

Primary classification: transition management, prestige signaling, and ideological anesthetic, with a substantial partial truth.

The partial truth is real: PAGE can expose patterns that book-by-book scholarship misses, preserve fragile material, and improve access. The anesthetic lies in describing this labor substitution as an uncomplicated benefit “to the medium itself.” The paper packages displacement as stewardship. Its moral vocabulary—preservation, accessibility, credit, and scale—makes the power transfer appear charitable.

Its prestige function is equally clear. “Human annotation,” “written codebooks,” and statistical controls signal methodological seriousness, while the underlying direction remains unchanged: encode expert judgment once, then apply it mechanically across an enormous archive. The result is not human scholarship multiplied indefinitely. It is scholarship reorganized around a smaller class of owners, supervisors, and indispensable validators.

The Verdict

PAGE is a credible instrument for extracting more knowledge from comics and a warning about what happens to the labor that once extracted it. It does not refute the Discontinuity Thesis; it illustrates P1 and P3. Cognitive work is formalized, validated, and scaled until most of the human participation becomes optional. The archive survives in better condition. The mass scholarly labor model does not.