Skip to content

R0029/2026-03-27/Q001/SRC02/E01
Research R0029 — Plural Voice Attribution
Run 2026-03-27
Query Q001
Source SRC02
Evidence SRC02-E01
Type Factual

IBM AI Attribution Toolkit: structure, functionality, and self-assessed maturity

URL: https://research.ibm.com/blog/AI-attribution-toolkit

Extract

The IBM AI Attribution Toolkit (released May 2025, available at aiattribution.github.io) is a questionnaire-based tool that generates formal attribution statements. It captures three dimensions:

  1. Relative contribution: How much work did the human do relative to the AI?
  2. AI's contributions: What specifically did the AI contribute?
  3. Review and approval: Who reviewed and approved the final work?

Users can generate short-form or long-form attribution statements with a single click. The toolkit was inspired by the CRediT (Contributor Role Taxonomy) used by major scientific journals.

Critically, IBM Research describes the toolkit as "a first pass at formulating what a voluntary reporting standard might look like" — explicitly positioning it as experimental and pre-standard.

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 Supports A working, publicly available tool demonstrates that formal frameworks exist
H2 Contradicts The toolkit's multi-dimensional structure goes well beyond binary disclosure
H3 Supports The "first pass" language and "voluntary reporting standard" framing confirms pre-standardization status

Context

The toolkit draws on findings from the CHI 2025 study (SRC01) with the same research team. This means SRC01 and SRC02 are not independent — they share a common origin.