INSPECT-AI: LLM Tool for Transparent Research Integrity Assessments of RCT Publications
The arXiv preprint identified as 2608.07202 unveils INSPECT-AI, an interactive tool powered by LLM, aimed at aiding human reviewers in evaluating the integrity of published Randomised Controlled Trials (RCTs). This tool is grounded in the widely accepted INSPECT-SR framework and employs the Research Integrity Provenance and Evidence ontology (RIPE-O) to trace the assessment process's origins. Additionally, the authors introduce the Research Integrity Provenance and Evidence knowledge graph (RIPE-KG), featuring an initial collection of 140 expert integrity evaluations for 95 RCT publications. By offering a structured and transparent methodology, the paper tackles the issue of subjective integrity assessments, ultimately striving to safeguard clinical care guidelines from being swayed by subpar or fraudulent research outputs.
Key facts
- INSPECT-AI is an LLM-based interactive tool for research integrity assessments of RCT publications.
- The tool is based on the INSPECT-SR framework and the RIPE-O ontology.
- RIPE-KG is a knowledge graph containing 140 expert assessments of 95 RCT publications.
- The paper is available on arXiv with identifier 2608.07202.
- The tool aims to assist human reviewers in assessing research integrity.
- The assessments are documented with provenance using RIPE-O.
- The goal is to prevent low-quality or false research from influencing clinical guidelines.
- The paper was announced as a new preprint on arXiv.
Entities
Institutions
- arXiv