Evidence codebook
Revision 3 — 20 October 2026

Unit structure
report_id identifies a publication, counted once across its versions.
study_id identifies an eligible main experiment or study in that publication.
outcome_id identifies an outcome-family record within that study.

Report fields
citation: short author/year citation.
doi: DOI for the cited report or specific discussion-paper version.
url: public source address.
scope_note (JSON only): report-level scope note, when needed.
studies (JSON only): nested eligible study units.

Study fields
n_reported: sample size as stated in the source. This is text so conflicts and multiple condition sizes can remain explicit. Do not sum it without resolving sample overlap and denominators.
object: what was viewed or offered for acquisition.
actual_source: reported production source of the stimuli.
attribution_design: information provided to participants, manipulated labels or participant classification.
context: recruitment and viewing setting.
source_locator: section, page, table, supplement or codebook location used for extraction.
appraisal: design-specific interpretation notes, discrepancies and uncertainties.
outcomes (JSON only): nested outcome-family records.

Outcome fields
construct: the practical measurement family used in this map. Despite this legacy field name, the categories are not asserted to be independent latent psychological constructs.
response_format: response scale, unit, perspective, range and aggregation where verified.
financial_consequence: whether the evaluative response carries a documented payment/allocation consequence; preserve the source-specific description.
note: qualification or classification rationale, including content embedded in a composite.

Allowed families
aesthetic_or_artistic_evaluation: liking, beauty, creativity, quality and related appraisal content.
noncurrency_economic_rating: economic wording answered without a monetary unit, including items within a mixed composite.
estimated_sale_price: legacy key for the family displayed as sale-price judgment. Includes a currency prediction or recommendation of a seller's price; response_format and note retain which perspective is elicited.
personal_stated_wtp: the respondent's own hypothetical monetary payment.
purchase_intention_or_relative_preference: acquisition likelihood or comparative buying preference.
auction_bid: bid under a documented consequential auction mechanism.

Missing information
Unverified detail is stated explicitly. A blank note or scope_note means no additional note is supplied, not that every possible property has been verified. Negative special codes in the Neef codebook are not rating-scale endpoints.

Counting
A study contributes at most once to a family but can contribute to multiple families. A mixed composite may therefore contribute two content-family records without supplying two independent statistical outcomes. Report-level counts collapse all eligible studies within each publication. Outcomes are not effect sizes, and the map is not an exhaustive population of AI-art studies.

Reproduction
Run reproduce_counts.py with Python 3 in this directory. It uses only the standard library. It checks identifiers, required fields, CSV/JSON agreement, the saved counts and all coverage exclusions. These checks validate the internal records; source accuracy requires reading the linked passages.
