Elevating the AI Conversation with the AID Framework

Restoring academic integrity through AI attribution statements

Posted on January 13, 2025 in Blog Posts

As we pass the two-year anniversary of ChatGPT’s launch in November 2022, much of the initial hype and panic surrounding generative AI has dulled. We have a better sense of the problems and uses of AI and have new frameworks for developing AI literacy. But even with more knowledge and lower emotional levels, a spirit of pessimism has also taken hold, especially in higher ed. How can we trust that students aren’t using AI to complete their assignments? How can we ensure that scholars are using AI in ethical, transparent, and responsible ways in their research? How can we continue to advocate for the importance of individual learning and education? It has been easier to pretend that these problems don’t exist or consider them impossible to resolve than develop practical solutions.

The University of Waterloo has taken some of the most exciting steps in higher ed to keep pace with generative AI and rebuild confidence in academic integrity. Its research guide is a top-notch resource for students and researchers incorporating AI into their scholarship. More recently, Kari D. Weaver, a librarian at University of Waterloo, published an article in College & Research Libraries (CR&L) introducing the Artificial Intelligence Disclosure (AID) Framework. The Framework offers a highly customizable, yet legible, means for students and scholars to capture the specific AI tools and usages they employ in their work. Kari’s article breaks down the components of the Framework and its uses, but we wanted to talk to her to learn more about its development. Below is our interview with Kari.

📄 Read “The Artificial Intelligence Disclosure (AID) Framework: An Introduction” in CR&L

---

LibTech Insights readers will remember you from your earlier work on generative AI citation (and webinar!). In a nutshell, what is the AID Framework, and how does it build on that foundation? 

In research, we have two ways of crediting others’ ideas and efforts: citation and attribution. Citation acknowledges a direct output or a particular fixed form of an idea. Attribution is more flexible and can address nuanced contributions to a work. We regularly see both of these elements in published works either as a citation in a reference list or as an attribution in the acknowledgements section.

Generative AI, also known as GenAI, complicates the information landscape. Students and scholars are using these tools in combination with their own judgment and for a much wider variety of tasks than traditional citation practices might cover. The need for transparent disclosure is also strongly expressed in journal editorial policies related to GenAI use, but there has been a lack of clear guidance on how to implement this disclosure.

The AID Framework addresses this gap in a few key ways. First, it helps define the range of possible acceptable ways in which a student or scholar might use GenAI. Second, it provides a clear and flexible structure for including this information in a student paper or publication. Finally, it allows everyone to provide this transparency in a way that is consistent and not onerous. When we leave students and scholars to disclose without guidance, we make it a difficult and time-consuming process with little clarity on the accepted practice.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• All Hail Our New Meeting Notetaker: Zoom AI Companion
• Five Choice Interviews with Digital Archivists and Curators
• Libraries, Education, and Video Games: Press X to Start

---

Can you walk us through an example of a statement a scholar can build using the AID Framework? 

Every AID statement has two main components: a statement about the AI tools used—including specific model and use date—and a brief review of the specific ways those tools were used in the research process. There are also some light formatting guidelines to make it easy to read by both people and computers.

Let’s take the example of a psychology researcher. They use their institutional instance of support for Microsoft Copilot for AI a few different ways through the process—for example, refining and critiquing study design, assisting with creation of the study instrument, and helping with framing out and editing parts of the final publication. They might also take advantage of Grammarly, which is now an AI tool itself, to assist with correctness elements like verb agreement, spelling, and punctuation.

The AID statement for this work could look like this:

Artificial Intelligence Tool(s): Microsoft Copilot (University of XX institutional instance) accessed November 8-14, 2024 and Grammarly (Pro personal account), November 22, 2024; Methodology: Microsoft Copilot used for critical feedback and analysis of study methods design; Data Collection Method: Microsoft Copilot used to design survey instrument; Privacy and Security: Microsoft Copilot institutional instance ensured study materials were protected from AI data training sets; Writing—Review & Editing: Microsoft Copilot used to establish paper outline and edit at the paragraph level, Grammarly used for verb agreement and detailed editing at the sentence level.

AID statements for researchers, like this example, are typically a bit longer and more detailed than those for students. If an element of the AID Framework does not apply to the project, you simply omit it. There is also a lot of flexibility for growth and supplementary resources, including a rubric, are available to help scholars and students in building their AID statements.

What I like about your previous work and the AID Framework is that it takes a pragmatic approach to the educational conundrums of generative AI. If we can establish a system of disclosure for generative AI use, we are better able to adjudicate ethical and unethical uses in alignment with our institutional values. It restores the concept of academic integrity, which some have intoned is now a lost cause. It also helps to dispel the culture of suspicion and fear that is plaguing the student-professor relationship.  

Thank you! The response to the AID Framework has been enthusiastic and overwhelming, in large part because it is designed to be practical. At many institutions, decisions about what to do have been left to the individual instructor, which results in inconsistencies in the student experience and a widespread lack of clarity. As an educator and librarian, I am always thinking about what I want my students to know and to do from my instruction. While there has been a lot of conversation around what we think students might need to know about GenAI, we have been woefully light on direct recommendations about what to actually do.

One of the major trends we see in response to this lack of clarity is instructors outright “banning” the use of GenAI tools in their classrooms. This simply does not work as some students will covertly use these tools. However, using the AID Framework allows instructors to set their course policies from a more nuanced perspective. In some cases, the instructors weren’t fully aware of the capabilities of GenAI tools, and the AID Framework helped them identify that gap. In other cases, it helped faculty better articulate what level of use could support learning their established outcomes for the course or assignment. In a couple of cases, I have spoken with instructors where GenAI tools really would have interfered with the intended student learning, and the AID Framework helped them to think through and better articulate why they were asking students to complete their work without GenAI. These directions build trust and confidence as well as enhance the culture of integrity at our institutions.

On that point, beyond institutional values, what is the pedagogical value of having students work through the AID Framework and write AID statements? 

One of the growing conversations we see in higher education is the need for AI literacy and critical thinking around the use and value of AI. By defining the different ways in which GenAI might be used in education, it opens essential conversations we need to have in our classrooms and with ourselves. What is the purpose of learning? How can we use the capabilities of GenAI tools to support or enhance learning? What are the limitations of these tools? How do we select a GenAI tool that meets our particular need? When do we already have optimized tools for a specific function that would be preferred? The AID Framework gives everyone a common vocabulary and starting point for these discussions. It also contributes to building trust between faculty and students, scholars and publishers, graduate students and their supervisors and committees.

The reality is GenAI is here to stay, and it is already having an impact well beyond the education sector. In higher education, one of our major contributions is teaching students to think in the manner of their area or areas of study. Considering critically how GenAI fits into these disciplinary and interdisciplinary systems of thought and practice truly is the major pedagogical task of the next 5 to 10 years. What we need now are tools, like the AID Framework, that encourage us to step beyond our fear or trepidation and actually engage with this new way of doing things in our world.

What is next for the AID Framework?

We are still in the early days of GenAI technology and its application to education. Just like with the technology itself, my focus for the AID Framework today is getting it into the hands of those who need it, seeing what works and what doesn’t work, and determining how it may need to be adapted from there. Like I mentioned earlier, the response so far has been enthusiastically positive, and it’s worked great for the vast majority of cases where it’s been used. I’m sure edge cases will arise, and I want to make sure those are also considered and the learnings shared with those who need it.

I’m also engaging with multiple standards bodies to discuss how the AID Framework can fit into their present and future guidelines. For those readers who are on the editorial boards of journals or support scholars who are, recommending the AID Framework as the approach to transparent disclosure can help accelerate this conversation.

It’s time to move the conversation forward about GenAI in the world of higher education. Adoption of the right tools for thinking about and discussing these topics will help everyone use this technology to its best potential in an ethical and transparent way.

Related Posts