TurboCurator Shines at Dataverse 2025: A Conversation with ICPSR’s Kelly Doonan-Reed
TurboCurator, a Research Data Ecosystem (RDE) tool that uses AI to generate metadata recommendations, was featured in six sessions over three days at the Dataverse Community Meeting 2025 at the University of North Carolina in June, drawing significant interest from Dataverse Administrators. Notably, TurboCurator is now listed in the newly developed Dataverse Marketplace, making it easier for organizations to discover and adopt new tools. Today, we caught up with ICPSR Product Owner/Manager Kelly Doonan-Reed, who played a prominent role representing TurboCurator at two sessions: “AI in Action - Innovation That Respects the Past and Looks to the Future,” as well as a collaborative presentation sharing use cases from other data archives.
Kelly, could you briefly describe how TurboCurator leveraged AI to support data curation, as discussed in your sessions at the Dataverse meeting?
TurboCurator’s initial prototype harnessed the power of ChatGPT to help researchers create high-quality, FAIR (Findable, Accessible, Interoperable, Reusable) data, keeping the human in the loop, providing critical guidance, and ensuring the results meet the highest standards.
Our team’s strong collaboration and deep curation expertise were foundational to getting this right. With TurboCurator, we dramatically shortened development cycles, rapidly deploying solutions that respond to the evolving needs of researchers and repositories. We are thrilled to share best-in-class metadata recommendations for key fields like “Title,” “Summary,” and “Keywords” — making the deposit process faster, smarter, and more sustainable.
A key differentiator is transparency: TurboCurator doesn’t just provide recommendations — it explains the rules and logic behind them. This empowers users, turning metadata creation into a learning experience and ensuring trust at every step.
Throughout development, we put TurboCurator through iterative, real-world testing — adjusting everything from model parameters to domain-specific guidelines — so that the AI’s suggestions are both actionable and tailored to our community’s needs. This innovative blend of machine intelligence and human insight not only raised the bar for research data curation at ICPSR but has also provided valuable lessons and techniques adopted in other data-driven initiatives.
During the conference, TurboCurator was featured multiple times. What kind of feedback or reactions did you receive from Dataverse Administrators and others?
The feedback we received was both enthusiastic and insightful. Attendees expressed excitement about TurboCurator’s potential to streamline workflows and improve metadata quality, while also raising thoughtful points about the complexity of supporting a range of disciplines and meeting diverse global standards. These conversations reinforced how important it is for AI tools to be flexible, customizable, and designed in close collaboration with the broader data community.
We saw a strong global interest in incorporating AI into data repositories and operational processes. It was especially rewarding to connect in person with our Dataverse partners, many of whom supported the initial TurboCurator prototype and experimented with other AI offerings like our integrated Chatbot.
Overall, we sensed real momentum and a shared commitment to responsible AI adoption, with many administrators eager to continue exploring and implementing these innovations in their own repositories in alignment with their repository rules, standards (e.g., controlled vocabularies), and compliance needs (country-specific rules and regulations).
In your presentation, you touched on “Innovation That Respects the Past and Looks to the Future.” How did you address balancing new AI capabilities with traditional archival values?
In developing TurboCurator, we were very intentional about ensuring that new AI capabilities were rooted in traditional archival values — especially those that support the creation of FAIR data. The foundation of TurboCurator is its rules and logic, which directly reflect established best practices in metadata curation and the core principles of responsible archiving.
By embedding these archival standards into the heart of the tool, we made sure that AI recommendations would uphold the same quality and integrity expected by the archival community. It was especially validating to see other archives test TurboCurator and achieve similar improvements in areas like title, summary, and keyword generation. This demonstrated that we could innovate and embrace the future of AI while still honoring and preserving the values that have guided our work for decades.
In short, TurboCurator blends AI innovation with deep respect for archival tradition — demonstrating that it’s possible to advance and modernize our practices without losing sight of what matters most.
Now that TurboCurator is available in the Dataverse Marketplace, what initial steps would you recommend to institutions interested in trying out these tools?
For institutions interested in responsibly integrating AI into their curation workflows, we shared a few practical recommendations during the event:
First, ensure you’re using Dataverse version 6.1 or higher, as TurboCurator requires this or a newer release. You’ll also need your Dataverse Administrator to enable the tool within your installation.
In terms of workflow, we found that the more information you provide to the AI — such as utilizing our more information box or including a summary — the better the results. Just as with your own experiments using tools like ChatGPT, refreshing or adjusting the input in TurboCurator can improve the metadata recommendations you receive.
We also recommended if you are creating a deposit on behalf of a researcher, to check-in on their specific needs.
We asked potential TurboCurator users to try it out and share feedback with the team on their experience and ideas.
What were some of the most significant outcomes or takeaways for you and your team from participating in the Dataverse Community meeting?
One of the most significant outcomes for our team was the opportunity to connect face-to-face with our testers. Those in-person conversations were incredibly valuable for gathering candid feedback, sharing challenges, and building stronger partnerships.
Another major takeaway was learning about the spirit of “coopetition” — where former competitors collaborate on shared projects to achieve more than any single group could alone. This collaborative approach opened our eyes to new ways that we, as a community, can pool our expertise and resources to advance innovative solutions together.
We were also inspired by hearing how colleagues around the globe are already utilizing AI in their daily workflows. Learning about the diverse approaches and creative uses of AI in data curation reinforced the potential for responsible, community-driven AI adoption — and gave us fresh ideas to bring back to our own work.
Contact: Dory Knight-Ingram
Jul 23, 2025