Documentation

This page provides the metadata schema (codebook), coding procedures, and best practices for contributing new studies to AI SHARE. For beta, content is embedded here; it can later be sourced from docs/schema.md and docs/contributing.md.

Metadata schema

AI SHARE organizes data at three levels: Paper, Survey, and Question.

The portal normalizes Airtable field names into stable API fields (e.g. question_text, country, year) so that consumers get a consistent schema regardless of base column names.

Coding procedures

Coders read papers and survey documents and enter structured information into Airtable. Decisions about how to map free-text (e.g. response scales, topics) into categories follow project-specific coding guidelines maintained by the research team. The web portal is read-only and does not alter coding procedures.

Best practices for contributing new studies

To add a new paper to AI SHARE:

  1. Obtain access to the project Airtable base (contact the team).
  2. Create a record in the Paper-level table with DOI, title, authors, year, source, and country/region as applicable.
  3. Add Survey-level records linked to the paper (country, fieldwork date, sample size).
  4. Add Question-level records linked to the survey or paper (question text, scale, category/topic, and any percentage breakdowns).
  5. Do not expose internal fields (coder names, status, deadlines) in any public-facing export.

Data limitations & quality notes

The dataset is curated by humans and may have gaps, delays, or inconsistencies. Country and year are derived from paper or survey metadata where available. Categories and topics may vary across studies. The portal is provided as a beta for exploration and export; always verify critical values against original sources when citing.