Who's the Real Author when AI is Used? The Moore AI Progression Scale Explained

Just because someone is a bad colleague does not mean that they do not still have things that you can learn from. One of the most important lessons in my life is that even if someone is hateful, they oftentimes still do have something they can teach you, whether directly or indirectly. Now, while we would all like to believe that teachers can work together in harmony and be kind together, supporting each other in the way that we are always trying to tell our students, anybody who has been in that environment unfortunately knows that sometimes some teachers are just a crabapple. For some reason they fail to apply the very lessons they are teaching their children to their own lives or to their own colleagues. But regardless of that fact, and regardless of the reason for their current state, that does not mean that we cannot benefit from their critiques. While it is important to have a thick skin if you happen to work with one, that does not mean that you should be fully immune to all things they say, as was true in my case.

Years ago, after I had recently written my first edition of Beyond Traditional Teaching: A Practical Guide to ChatGPT in the Classroom, I was putting my books on my bookshelf behind my desk when this teacher, whom I was not familiar with but who would be a new colleague, came up, introduced herself, and we began talking. She asked about the book. I showed it to her and explained that I had written this book and some of the things that I found interesting and why I had written it. After casually glancing through it for a few moments, she asked if I had used AI at any stage in writing it. I truthfully explained, of course, that I had. In fact, one of the points of having written this book using AI was as a demonstration model for how AI and people can work together to create something better than either of them could alone. She then casually tossed the book back onto my desk and very matter-of-factly stated, “Well, you didn’t write it then.” I retorted very clearly, “Yes, I did write this book. Getting input on writing and having corrections made has been part of the book-writing process forever.” She was very dismissive and resolute though in her statement as she repeated again, “If you used AI at all then you didn’t write it and it’s not a real book.”

Well, that would be an omen of things to come. At the moment, though, I was more flustered and annoyed. I had spent weeks and weeks working on the different chapters of this book, nearly my whole summer, developing ideas, working through hypothetical examples. You have to remember this was back the very same year that the very first ChatGPT 3.5 chatbot opened up to the public, and it was essentially the Wild West. Everything was a guess about how this could be used. Fortunately, nearly all of my predictions about its use, and its abuse, turned out to be true. The book went on to occasionally reach moderate popularity, especially in circles about theoretical teaching models and educational pedagogy on Kindle. However, for years her comment still needled me, and it suddenly dawned on me that she, with her dismissive critique, had actually put her finger on something very important: how can we determine the level of authorship in a given work when AI was used at some point?

It then dawned on me: we needed to have a scale. There had to be some way to be able to determine the amount of AI that was used in a given work and the amount of human that was used in a given work. With some kind of scale, we would actually be able to know whether a person was using AI in a way that might one day be considered correct—these things are still in flux literally as this article is being written—or whether a person had used it incorrectly. And of course, that will also depend on the context of the use. It was with this idea driving me forward that I began working on the Moore’s AI Progression Scale, lovingly referred to as MAPS. MAPS would provide a clear way of identifying, both for students and for their educators, how much AI both parties were essentially agreeing or allowed to use.

The problem was in designing such a scale. While I was able to finish writing a book over one summer, in this particular summer trying to create the scale proved far more daunting. Every single attempt that I made to figure out a proper metric ended up failing the moment I attempted to apply it. Too wordy, too vague. We needed clear-cut and obvious. Something people could actually use. Finally an idea began to coalesce around a scale with five separate levels, ranging from essentially spell check at level 1 all the way to level 5 where the human instigates a prompt and takes the work directly as it is written, and depending, the human may or may not actually even have originated the prompt either. The more I refined the model, the more I realized that no one else had created a tool quite like this—one that was designed to provide openness and clarity both for teachers when evaluating student work and for students themselves.

Forever ago we had already figured out that we should let students know when they can and cannot use calculators and the appropriate role of that tool. Now it is time for us to apply the mathematical lessons we learned when that technology was introduced to the field of artificial intelligence. Students don’t need to be banned from using it; they need to be taught when they can use the tool, when they cannot use the tool, what the purpose of the tool is, and why they might be allowed to use a level 3 for writing in one project but then be relegated to a level 1, or level zero—zero of course being no AI of any kind—for another project. Using the scale, teachers can clearly articulate what the purpose of the lesson is and why the tool can only be used for certain applications of that purpose.

But it also works the other way as well. It allows a student to have greater clarity from their educator about the degree to which the educator themselves will use AI. With a myriad of scandals occurring both at the high school and collegiate level of teachers simply using AI to grade their students’ work without ever looking at it themselves, that creates a trust barrier—especially when those students have been told that they cannot use AI for writing the project but then their professor never even lays eyes on it and simply has a machine do his work for him. That breach of trust hurts the educational process, and trust needs to go both ways. In their syllabus or course work, a professor should clearly articulate how they intend to use AI to assist them, what levels it will be used at for what applications, and there should always be a promise that at some level they themselves will actually operate at an AI level of zero at least for some assignments and look over the work themselves, just as students at some point in every project should also be working at an AI level zero to ensure that the work is authentically their own.

While the full definitions of the MAPS scale are laid out in the second edition of my book—and that was a major reason why I decided to write the second edition—in summary, the scale works like this. Level zero obviously is no AI of any kind. You’re doing it old school. And by no AI, that also includes traditional spell check tools and things like Grammarly, which are now becoming AI-powered. This means either you have turned those features off or you are working using manual academic tools like pencil and paper.

A level one, as we mentioned before, is something like spell check. For instance, a professor might allow his students to use level 1 of the MAPS scale. Level one just means that you had the AI perform functions like spelling, grammar, and correcting misused words. At this level, the AI functions as a glorified spell check, which has already been academically and ethically sound for nearly 20 years, especially in higher education. No one is going to claim, or at least should claim, that just because you had the AI go over and fix your spelling and your misused words and your grammar that somehow you are no longer the author of that work.

Level two is now brainstorming and idea generation, trying to draw some inspiration from the thoughts of the machine but not taking its words directly or its prompts, as no or extraordinarily little of the AI’s actual verbiage should occur in the text. This is mostly for inspiration, outlining, where the AI gives big broad outlines, concepts, or ideas, and the human actually does all the work to finish and fill them in.

Level three, something special happens that I refer to as synthescribing. While this level is not appropriate for every academic function, it is the one that I have found is the best for work, especially in my writing process, as it is where the human and the AI enter into a beneficial reiterative feedback loop. Synthescribing is not the AI replacing the human author; it is the AI working alongside a human-driven process. The human provides the ideas, the creativity, the direction, the questions, the critiques, and the decision-making. The AI provides speed, structure, counterarguments, expansion, and refinement. The human is still the driver of the entire process. Without the human’s ideas, questions, critiques, and creative direction, there is no feedback loop at all and there is no collaboration. The human is not being replaced; the human is being amplified. The process becomes a back-and-forth collaboration where the human proposes ideas, the AI responds, the human critiques and reshapes those ideas, and the AI helps refine and articulate them more quickly. Together they create a synthesis of ideas that neither could have created alone, but the direction, creativity, and judgment remain human-led throughout the entire process.

While I believe that for many work applications level 3 synthescribing might be the sweet spot, level 4 is where it begins to go sour. Level four is where a student or a person gives the AI a prompt and then either simply rewrites it into their own words or makes minimal adjustments or minimal reiterative process. They do a little bit of back and forth, they might give some ideas, they might even rewrite it into their own words, but essentially at level four the machine has taken control of the wheel and the human is a more passive passenger in the realm of creation.

Level five takes that to its natural conclusion in something I refer to as autoscribing. The machine simply does all the writing and the human contributes essentially nothing apart from either a prompt, and depending on the situation they may not even contribute that. Sometimes people have machines tell the machines to make things, in which case no human was involved in the creation at all.

We could even simplify these five levels one step further by simply saying it like this:

Level 1 – Did the AI help you mechanically?

Level 2 – Did the AI inspire you?

Level 3 – Did the AI collaborate with you?

Level 4 – Did the AI work for you?

Level 5 – I can see the AI… where are you?

While the full scale is available in my book and it will also shortly be available for my paid subscribers, I hope this outline of the scale itself has proved beneficial and will start to change your thinking about how to create guidelines around the use of AI that are beneficial both to students and to their teachers. While there will always be some naysayers who will suggest that anything more advanced than the pencil is not truly your work reasonable people using the same scale, agreeing and knowing how it was used, and to what degree, can help reinsert trust back into the academic process, which has rapidly been losing trust as students lose faith in the institutions and the institutions lose faith in their students. Re-establishing trust then is foundational to the success of the future of education.