In this session, Jana Christopher, Image Data Integrity Analyst at FEBS Press, offers an authoritative overview of current issues and tools in detecting image manipulation.
This discussion was part of 'Publication Misconduct and Fraud Day' hosted by COPE during Publication Integrity Week 2023.
Image fraud, beautification, and genuine errors
Rapid changes in technology in conjunction with businesses like paper mills have created an ‘arms race’ with tools for detection. This is an important issue because images are a reflection of underlying data and methods in fields like biomedicine. It is therefore vital that the scholarly community can trust what it sees. Jana discusses issues from fraud, to image ‘beautification’ and genuine errors, all of which can distort results, pointing out that journals have variable practices for detection.
Raw data availability
The market for screening tools is growing, but these should be used alongside expert verification. Detection is very difficult when raw data are not available, and this in itself should be a flag that further investigation is necessary. Paper mills routinely use falsified or AI-generated fake images which may need deep familiarity with manipulation issues to identify. It is good practice always to ask for the raw data behind images, but journals need to be specific in what they require and how this should be presented. Ideally we would have universal guidelines for raw data requirements, but these may require expensive infrastructure that is beyond many research institutions.
Related COPE resources
- Image manipulation in a published article COPE flowchart
- Login to your account or register
to post comments