A collaborative project led by UC Davis, working with four California Community Colleges and three California State University campuses, has won a $1.5 million grant to help address equity gaps in writing support and AI literacy.

The awarded project, Peer and AI Review + Reflection (PAIRR), aims to build student competence and confidence by pairing human feedback with that of artificial intelligence (AI). 

“PAIRR emphasizes reflection and self-assessment and encourages students to be active and critically engaged users of AI feedback, rather than passive recipients who merely follow directions,” said Lisa Sperber, a continuing lecturer in the University Writing Program (UWP) at UC Davis. 

The grant is funded by the California Education Learning Lab’s AI Grand Challenge: Leveraging AI for Teaching and Learning, which helps foster innovation in the use of AI to improve curriculum and teaching in the state’s public higher education system. A total of five projects were selected to receive grants of up to $1.5 million each. 

The California Learning Lab award will fund the development of PAIRR for three years beginning in January 2025.  

Evidence put into action

The PAIRR project builds on a 2023-2024 UC Davis study that looked into how generative AI feedback could be used in the peer review process. 

“Our project was about how students can use GenAI writing technologies in smart, ethical and critical ways rather than using it to replace their own learning or take away their agency as writers,” said Carl Whithaus, professor of writing and rhetoric in the College of Letters and Science and UWP director. 

Generative AI is trained on pre-existing media — images, videos, text, etc.  — and creates a new piece of content based on various user prompts. 

“Evidence shows that cycles of drafting and revising are crucial for student writers’ growth,” added Sperber. “Feedback on student drafts has been shown to improve writing outcomes when it is clear, actionable, timely, and specific.”

But the reality is that teachers and teaching assistants are often strapped for time and can’t always provide this thoughtful feedback to everyone. 

“When used alongside human feedback, technology has the potential to help make this work more efficient and increase equity by getting more and better feedback to more students,” Sperber said.

Moreover, the study found that most students prefer receiving both types of feedback. 

When the human and AI suggestions overlapped, students felt reassured, increasing their confidence in the validity of the feedback. When the two forms of feedback differed, they complemented one another, offering different perspectives to students, according to the study’s investigators. Students learned how to critically assess the AI feedback too, noting when it was inaccurate or didn’t serve the student’s intended purpose, building their AI literacy.

“When students use AI in their writing process matters too,” Marit MacArthur, a continuing lecturer at UWP who is leading the project alongside Sperber, Whithaus and co-investigator Hao-Chuan Wang, associate professor of computer science. “We decided that students should receive AI feedback on their writing after they have completed a draft and an initial peer review, rather than during the drafting process, because this puts them in a strong position to assess AI output, which is crucial to building AI literacy.”

More than 650 undergraduates from 10 composition classes and three writing-intensive STEM courses across campus participated in that study, also led by graduate students Nicholas Stillman (Ph.D. candidate in English) and Sophia Minnillo (Ph.D. candidate in linguistics). The research was awarded a $3,000 Research Initiative Grant from the Association for Writing Across the Curriculum.

Moving forward with AI technology

The larger PAIRR project will take lessons learned from that study and put them into practice, drawing on the strengths of both approaches. The PAIRR process is a five-part plan in which students discuss short readings on AI, complete a peer review of their draft, prompt GenAI to review the same draft, critically reflect on and assess both types of feedback, considering their goals and audience, and, finally, revise. 

This way, MacArthur said, using AI supports them in their writing process without undermining their learning.

“We cannot ignore GenAI because it is being incorporated into the workflows of many careers. Failing to prepare students to collaborate with it in safe and appropriate ways could deepen the digital divide,” MacArthur said. “But what’s appropriate for GenAI in the workplace is not necessarily appropriate for education. LLM-powered chatbots like ChatGPT were not designed for education, so students need guidance to learn to use them more like AI tutors and less like ghostwriters. And whenever possible, we need to try out affordable tools developed by educators, rather than commercial tools.”

In the initial study, the team protected student privacy as much as possible, however, they were using ChatGPT, which collects a lot of data and is not completely private. Going forward, PAIRR will now use MyEssayFeedback, a not-for-profit GenAI platform, which has more privacy protection built in. MyEssayFeedback was designed by Eric Keane, a former math instructor, in collaboration with Anna Mills, who is also a member of the PAIRR leadership team and teaches writing at the College of Marin.

The project will involve 114 instructors from UC Davis and partnering institutions including CSU Bakersfield, Sacramento State University, CSU Maritime, American River College, Glendale College, College of Marin and LA Mission College. In addition, PAIRR will host public webinars on its ongoing findings during the summer of 2026 and 2027, and host a one-day training for 20 new college writing instructors during its Summer Symposium in 2027. 

Whithaus, Sperber and MacArthur are also co-investigators with the Center on AI and Emerging Futures (CAIEF) at UC Davis, which was recently funded by a $500,000 grant from the National Endowment for the Humanities. 

The CAIEF is also providing some support for the PAIRR approach at UC Davis beyond UWP courses. Faculty interested in trying PAIRR in their writing-intensive courses should reach out to MacArthur via email at mjmacarthur@ucdavis.edu.

YOU MAY ALSO LIKE THESE STORIES

Unraveling the Social Impacts of Artificial Intelligence

UC Davis experts Martin Hilbert and Marit MacArthur from the UC Davis College of Letters and Science answer important questions about advancements in artificial intelligence at a recent Team Research Forum.

Reimagined Davis Humanities Institute Slated for 2025 Launch

The Davis Humanities Institute is on track to start anew. By fall 2025, it will become the campus hub for making connections and sharing ideas across the arts, humanities and beyond. Its goals include making a deeper impact on campus and in the community, providing more research opportunities for undergraduate students and supporting cross-collaborations between disciplines within the College of Letters and Science as well as other colleges within UC Davis.

Stories Archive