A theory-first approach toward using generative AI for humanities research

Andrew Piper proposes theory-first guidance for generative-AI-assisted humanities research: explicitly define research problems and analytical constructs, distinguish AI as an object from AI as an instrument, and use theoretical commitments to guide evaluation and interpretation. The article argues for methodological rigor rather than demonstrating an independently verified performance gain.

Workflow level n/a Source Project-documented Outcome Not empirically evaluated here

Evidence & verification

Source level

Project-documented
Grounded in primary project documentation, not necessarily an empirical evaluation.

Outcome verification

Not empirically evaluated here
A documented resource or intended workflow, not a demonstrated academic outcome. No empirical benefit is established by its inclusion.

Sources checked

2026-10-17 (UTC)

Publisher title, author, DOI and online date match identity; journal Research Article in Computational Humanities Research, volume 2, e18. Abstract supports theory-first proposal; introduction and Step 1 support object/instrument distinction and construct/unit specification. Benefits are methodological arguments, not tested gains; no quantitative outcomes carried forward. Publisher-hosted methodological article; peer-review process was not independently established from the inspected excerpts. The evidence label does not imply full-text review.

Source-checked does not mean independently reproduced. See the Method page for how these labels are assigned.

Sources (1)

Research applicability

Direct methodological guidance for scholars using generative models to study literature, history and culture, including construct validity and theory-guided research design.

Getting started

Editorial suggestion, not an executed recipe: read the article, state whether AI is the study object or instrument, specify the construct and unit of analysis, and document evaluation criteria before using model outputs as evidence.

Maintenance

Mutable documentation captured at access time; future compatibility not established.

License

Check original license and dependency/data rights before reuse; no independent legal audit.

Workflow & human oversight

Resource-level automation unspecified; researchers define aims and verify outputs. No autonomous end-to-end outcome claimed.

Workflow level is not an evidence score; higher is not better evidence.

Limitations

Publisher abstract, plain-language summary, introduction and problem-formation excerpts reviewed; not a full-text adjudication despite the archived HTML containing further text. Examples and cited studies were not independently verified. No software execution or outcome reproduction.

Reproducibility

Source review by AI agent only; no installation, experiment or claimed outcome reproduced.

Adoption requirements

A humanities research question, explicit theoretical commitments, suitable texts or cultural materials and domain expertise. Any model use also requires appropriate data rights, privacy review and documented model/prompt choices.

Comments

No comments yet.

Add a comment

Comments are held for moderation before appearing. Critical or corrective comments are welcome — especially evidence that an entry overstates what has been verified.