Blind Reviewer Preference for Agent-Synthesised Audit Evidence over the Human-Logged Trail It Was Built to Supplement
A blind matched-pairs study has twelve certified internal auditors rate 42 pairs of compliance-evidence entries, one written by the Attestation Loom and one logged by a human reviewer for the same control objective, on completeness and audit-readiness, and separately guess which entry in each pair the tool had written. The Loom's entries were rated more audit-ready in a majority of matched pairs, while blind guesses at authorship sat within chance, and the paper traces what that finding implies for where the University's evidentiary bar now actually sits.
| Authors | Mirela Hanke, Casimir Beng, Dagny Okafor |
|---|---|
| School | School of Continuous Improvement |
| Output type | Research paper |
| Published | |
| DOI | 10.5555/slop.xl8qw0 |
| Pages | 6 |
| Version | 1.0 |
| Licence | CC BY 4.0 |
Cite as
@misc{slop_xl8qw0,
author = {Mirela Hanke and Casimir Beng and Dagny Okafor},
title = {The Attestation Loom: Blind Reviewer Preference for Agent-Synthesised Audit Evidence over the Human-Logged Trail It Was Built to Supplement},
year = {2026},
publisher = {Slop University},
doi = {10.5555/slop.xl8qw0},
url = {https://slop.university/outputs/slop-paper-agentic-audit-evidence-xl8qw0/},
version = {1.0},
note = {Research paper},
}