Read the recap below or watch the recording here.
Setting the stage
Dustin opened with a gap that framed the rest of the discussion. By most estimates, only 3-5% of manuscripts raise genuine integrity concerns, yet desk rejection rates across ScholarOne are considerably higher, which means editors are making many more decisions about quality than about fraud. Miriam was quick to note that integrity work is far from finished, but even with fraud set aside, a harder question remains. “Does this piece of content make a contribution to scientific discourse? And I think there the debate is quite lively, because quality can be measured in so many different ways.”For Miriam, part of that debate is how best to use an already stretched reviewer pool. “Not everybody publishes articles with groundbreaking research, but does everybody deserve a quality share of attention from a reviewer?” Dawn added that quality judgments will always rest with people, blind spots, and all. “We’re going to get it wrong some part of the time. But I hope a good amount of the time, we get it right.” The opportunity, in her view, is to take mechanical tasks off reviewers’ plates, and IEEE has been exploring whether an AI agent could validate the algorithms within submitted articles.
Jonathan introduced a distinction that the panel returned to throughout the session. One kind of quality is about process, which he compared to parts per million in manufacturing: were the right checks, reviews, and approvals in place? The other is about caliber, meaning whether the research is good, applicable, and relevant to the person or AI tool looking for it. The challenge, he said, is that “in an AI workflow, it doesn’t really, today, have a good way to distinguish between the two.”
From speed to friction
Submission growth is putting pressure on every part of the editorial process. Dawn estimated that most publishers are seeing increases of 20-30%, depending on the field, while Dustin pointed to individual journals that have doubled and conferences that tripled year over year. Dawn was confident the industry can adapt, even with fewer reviewers available, and she was clear about her priorities. “I would rather be a little slow right now and get the quality judgment right.”Miriam described how much earlier those judgment calls now happen. “If there are any concerns about your manuscript, you will not get a review anymore,” she said, adding that this frees up resources for stronger content. Dawn sees the results at IEEE, where desk rejections are rising quickly while final acceptance rates are creeping up, which tells her the initial screen is working.
Both publishers described a clear shift in priorities. When Miriam joined IOP Publishing five years ago, she was asked to grow submissions and speed up peer review. Now, she said, the conversation centers on whether speed is a good sign or a warning sign, and on hurdles for authors whose manuscripts aren’t ready. Dawn described an IEEE volunteer group dedicated to adding friction to the submission process; a change she admitted takes some getting used to. “This is a change from what I’ve been doing for the past 30 years. Give me a minute. How do I figure out how to make it harder for authors?”
Sound science journals proved to be the hardest case. As Dustin put it, “it’s effectively free to create a research-shaped object,” and these journals are seeing the fastest submission growth. Miriam said the threshold between acceptable and good is hard to institutionalize, because editors recognize it when they see it, but applying that judgment at scale is another matter.
What manufacturing can and can’t teach us
Jonathan, who worked in lean manufacturing before moving into scholarly publishing, challenged the idea that slowing down automatically improves quality. He pointed to poka-yoke, a concept from the Toyota Production System that focuses on preventing mistakes from being possible in the first place. “It’s not about time, but it’s about trying to figure out how to manufacture errors out of the process,” he said, which suggests added friction doesn’t automatically improve outcomes.Miriam noted that peer review has been moving in this direction, using tools to help people get it right the first time. When manuscripts reach reviewers, the routine checks are already done, and reviewers can focus on what the paper contributes to the field. Dawn agreed but pointed to the constraint that technology can’t fully solve. “Now if only we were making widgets,” she said. Reviewers can’t simply double their workload, so the real challenge is finding more experts.
Quality in an agentic world
The conversation then turned to how quality signals will travel as AI agents increasingly find and read research on people’s behalf. Jonathan explained that in the search era, a publisher’s domain carried gravitas because search engines learned which sources to trust. “In an agentic world right now, literally, every agent out there is working off their own search algorithm,” he said. Over time, he expects agents to learn from human behavior which sources are reliable.Miriam emphasized that publishers can help shape that outcome, and she pointed to industry work on trust markers. In her view, trust markers work on several levels, helping agents understand what kind of content they’re looking at and what process it went through, while also giving authors, especially those from emerging research communities, a clearer signal of what to expect from a journal.
Jonathan outlined what it takes for those signals to work in practice. Trust markers need to be defined and tied to the content, delivered to agents with provenance intact, and interpreted correctly by those agents, with usage reported back to platforms, publishers, and authors.
Making room for novelty
The final part of the discussion focused on something that’s easy to lose in a push for efficiency. An audience member asked whether some permissiveness in publishing might be a feature rather than a bug, since the value of research isn’t always clear at the time it’s published. Miriam agreed. “As domain specific publishers, we have to create the space for that.” Dawn noted that the same concern is surfacing at IEEE, where one editor-in-chief is researching an AI reviewer while others worry about what that would mean for novelty.Looking back at the rapid publication cycles of the COVID era, Miriam drew a lesson for AI: getting it right brings real rewards and getting it wrong carries serious risks. Dawn offered a steady perspective on the moment. “To me, the current AI rage is not scary. It’s just another technology that we have to learn to work with.”
The panel left us with a clear message: integrity remains essential, but it is the beginning of the quality conversation rather than the end of it. The path forward involves investing in the people who make quality judgments, using tools to give them more time for that work, and working together across the community to make sure trust travels with research wherever it is read.
View all of the Platform Strategies 2026 session recordings here.