In 2020, AIP Publishing formed a DEI Committee, setting out targets for increasing the diversity, equity, and inclusion of our organization and publishing programs. We wanted to ensure all communities are adequately represented in our organization and content; that everyone, regardless of their background, has an equal chance to contribute and be recognized; and that our content would be enriched with a wider range of perspectives, reflecting a more diverse pool of authors, reviewers, and editors.
To that end, our editorial sub-committee, along with our publishing partners, collaborated to embark on a major benchmarking exercise. Soon, we’ll be releasing a report that lays out high-level data encompassing the geography, gender, and ethnicity of our authors, reviewers, board members and editors. It includes all AIP Publishing journals, as well as journals from several partners: Acoustical Society of America (ASA); American Association of Physics Teachers (AAPT); The Structural Science Society (ACA); AVS: Science & Technology of Materials Interfaces, and Processing (AVS); and The Society of Rheology (SOR).
Tracking diversity, equity, and inclusion is critical to making sure all voices are being heard. Yet tackling an exercise like this is no small feat, so we wanted to share our feedback on the process — and what we learned through doing it.
As a part of the Joint Commitment for Action on Inclusion and Diversity in Publishing, we’ve been actively considering the right approach to gathering and sharing our data. While self-reported data would provide a more inclusive and detailed demographic snapshot of our community, putting a system in place to securely collect this information takes time. In the interest of gaining an initial view and baseline, we worked with the consultancy Delta Think to undertake data analysis using estimating software relying on existing publication metadata from the Peer X-Press (PXP) web-based editorial system for manuscript handling.
The analysis comprises three years of data, from 2019-2021. One of the biggest hurdles we faced was to share the huge amount of data — some 500,000 “person/activity” records — sourced from PXP. We wanted to analyze corresponding authors, reviewers, editors, and editorial board members, and required a unique record combining name and country information for each. Ensuring secure data transfer in line with GDPR and other privacy regulations was paramount. An aggregate and anonymized results set was produced to allow safe reporting — but at the same time, to be able to spot check the accuracy of the automated data analysis, we had to ensure there was secure access to individual records for a small number of nominated AIP Publishing stakeholders.
The task of cleaning the data was gigantic, and you can find a Delta Think blog on the subject here. They consolidated unique records, taking those 500,000 entries down to 200,000. The data was then processed using software that predicts likely gender and ethnicity based on names and countries, providing categories and match probabilities.
While insightful, we know this process has its limitations. For instance, because the software relies on older records like census data, it can only classify gender in binary terms and cannot identify non-binary identities solely based on names. Ethnicity classifications also align with the primary categories used by the US Census. These broad categories were chosen for compatibility with other reports, yet we acknowledge their inherent limitations. Fortunately, manual data verification upheld a 93% match for gender and a 94% match for ethnicity, ensuring a high level of confidence in our findings.
In essence, what we’ve compiled is a good proxy for diversity across AIP Publishing’s journals. There is a high degree of confidence in the accuracy of the analysis we have conducted, based on the manual spot checking and self-declared information we have compared the analysis to.
So, what comes next? The report itself, and our initial benchmarking, will be shared soon, along with our thoughts on what the data tell us. Beyond this first report, we are also making progress on developing a system for consistent self-reporting of data through PXP, which we will also share more information about in due course.
As we go forward, we pledge to continue to keep our community up to date on our progress. Our thanks go out to the many AIP Publishing team members and our partners who helped enable this process.