Learning Aloud

When a technology offers advantages and disadvantages, the decision-making process can be quite complicated, especially when oversight cannot be guaranteed. For example, many states now ban cell phones, making a use such as telling parents when the schedule for after-school activities has changed difficult. The advantages and disadvantages vary with the field of application and my interests have mainly been focused on education. Just to be clear, by this I mean learning in general, not just the type of learning that occurs under supervision or is associated with educational institutions.

The generic educational situation that raises concern involves tasks undertaken to encourage both skill development and knowledge, and includes a requirement that demonstrates that the task has been attempted by the existence of some product. In educational settings, such products might result from homework or class activities, or simply by visible demonstrations of activity. The issue with AI is that in many cases, such as problem sets or documents of various types, these same products could be generated by AI, avoiding the cognitive activity of the learners. The phrase “cognitive offloading” has been used to describe this alternative form of product creation. Teachers might simply call it cheating. Cognitive offloading itself can be a desirable or undesirable option, requiring decisions regarding when it is appropriate and efficient, and when it is a detriment. 

While cognitive offloading to avoid learning tasks seems an obvious problem, little actual research exists to demonstrate the damage done. Some would argue that if technology can replace an activity and that technology is readily available, why bother to “learn” the skill in the first place? Why learn information if your cellphone can allow you to search for information when it is needed? Why learn basic calculation skills when you cellphone can also serve to do mathematical operations? There are responses to these challenges, sometimes offered by students or parents, but this analysis would take this post in a direction I did not intend. 

Here, I want to focus on learning to write and writing to learn by discussing a different learning task. This may sound unnecessary, but at present, there is a reason to take this approach. The justification for being indirect is that writing is a complex skill consisting of multiple subskills, and we learn to become competent at even a basic level over years and not weeks or hours. We are investigating an alternative to the traditional methods of instruction that can be subverted now, and we cannot rely on experience to help us evaluate and tease apart how the development of subskills are impacted. The insights and evidence of the potential damage done would take to long to emerge. As one perspective, consider the lingering impact of COVID on learning. What about the move to online learning did we not anticipate and what consequences are we still trying to mitigate?

AI in Learning to Code

Shen and Tamkin had an opportunity to investigate the impact of AI with adult programmers learning to make use of a new library. Think of a library as a collection of functions (tools to perform specific and commonly used tasks). Instead of having to write code to accomplish common tasks each time a programmer encounters a need, libraries allow programmers to call prewritten code snippets. It takes some work to make use of a library – what functions are available, how do you call the function you want, what inputs and outputs are involved and how are these integrated with the code you write yourself? The researchers recognized that the learning coders had to do to make use of a new library provided an opportunity to study how AI could help and hinder learning a complex process. 

Shen and Tamkin studied actual programmers as they worked to learn a new library. They suggested that the process be viewed as a tutorial including both background information and simple programming tasks. Programmers were assigned to a control and a treatment group, with the treatment group having access to AI. The learning phase concluded with an assessment evaluating multiple concepts and skills. Video of treatment group participants was collected to document how each individual used AI and worked on the programming exercises.  

The researchers found that the treatment groups did not differ significantly in the time spent learning, which they found surprising. On the post-test, the largest group differences were in debugging skills. Smaller skill differences were found for code reading and conceptual understanding. Those without access to AI made more coding errors on the practice tasks, spent more time practicing debugging, and ended up with better skills on the outcome evaluation. How AI was used differed greatly with some simply asking AI to solve the coding challenges and others who only asked higher-level questions of the AI tool. Some users had the AI tool solve the coding challenges and then retyped the solutions themselves (rather than copying and pasting). This was not an effective strategy. 

Generalizing from the coding study

I have spent considerable time both coding and writing and I have always found the processes to have similarities. While others may find this a strange observation, I have always said that coding and writing were the two professional tasks I learned I could not perform later in the evening if I wanted to get a good night’s sleep. Reading was fine. Grading was fine. Something about both coding and writing was cognitively stimulating, making it difficult to sleep. 

The application of AI to complex skills is interesting, but difficult to study. Clearly, a single skill would seem very unlikely to be developed if a learner could completely substitute AI for practicing the skill. However, it seems possible that learning a multiple-component skill such as reading or coding might benefit from replacing specific components with AI under certain circumstances. We have limited cognitive capacity and substitution for some components of a complex task could allow the remaining components to receive more attention until well learned.

Learning to write might represent an example. I have often referred to Flower and Hayes’ writing process model when describing the components of writing and writing to learn. The use of AI to offer content to provide the basis for a writing task and perhaps even to offer a structure to guide the organization of a writing product could free up capacity to focus on lower-level skills such as spelling, grammar, and coherent paragraphs. In contrast, I typically use Grammarly while I write to allow to move more quickly while relying on this AI tool to alert me to possible spelling and grammatical improvements. 

Part of what Shen and Tamkin observed in their qualitative observations of the different learner-imposed focus of AI and the relationship of differences to what was learned or not learned offers a related perspective. Debugging is an important lower level coding skill and having AI debug code appeared to limit a coder’s ability to debug when working without AI. 

Suggestions for Learning to Write and Writing to Learn

AI can support both “learning to write” (developing writing skill) and “writing to learn” (using writing to deepen understanding), but depending on which writing skills are the goal best practices should differ.

Learning to write: skill development

Here AI should be thought of as a coach, not a ghostwriter.

Emphasize feedback: Tools like Grammarly give immediate feedback on grammar, syntax, cohesion, and organization, helping students revise iteratively while concepts are still fresh.

Structure and separate subprocesses: Generative tools can help students brainstorm ideas, outline structures, or identify expectations for different types of writing (e.g., sample introductions, transitions).

Process?first policies: “Writing first, AI second” approaches ask students to draft independently, then use AI for critique and revision. When coders used AI in the Shen and Tamkin, this is the general theme that seemed most successful. 

Writing to learn: thinking with text

When the goal is conceptual understanding of content knowledge, AI is best used to amplify reflection, not replace it.

Clarifying concepts for the writer: Students can ask AI to reexplain readings, generate examples, or pose practice questions, then respond in their own words, using writing as a space to consolidate understanding.

Challenge personal understanding: AI can generate counterarguments, alternative explanations, or “what if” scenarios that students must address in writing, pushing them beyond summary toward analysis. Why do others disagree with the summary I am creating? What can I offer to support my position and what are the limitations of the alternative?

Shared design principles

There are some guidelines these goals for writing. Across both purposes, similar design choices matter.

Make process visible: Require artifacts – notes, outlines, draft histories, and brief process memos about when and how AI was used. Document the transition from any use of AI to products student has generated. 

Align AI roles with goals: For skills (learning to write), let AI focus on feedback, exemplars, and mechanics; for content learning (writing to learn), keep generative help outside the main composing space and treat it as a prompt.

Previous analysis of technology and the writing process

Sources:

Flower, L., & Hayes, J. R. (1981). A cognitive process theory of writing. College Composition & Communication, 32(4), 365-387

Shen, J. & Tamkin, A. (2026). How AI impacts skill formation. arXiv preprint arXiv:2601.20245 (this study has yet to officially be published)

One thing I miss as a retired academic is going to the office daily and having the chance to share ideas on common interests. My background was in educational psychology and topics such as AI and learning would not only be relevant to me, but also to the people I had lunch with and passed in the halls. I would have been interested in my colleagues’ take on the pros and cons of AI in classroom settings. Were they concerned about cheating? Had they encountered students who cheated and how did they know for sure that their suspicions were justified? Had they modified the assignments they had always used or perhaps abandoned these tactics as untrustworthy? 

I still find myself thinking about such topics and despite no longer having firsthand experience, I wonder what I would do should I still be working. When something is that important in your life and self-view, it doesn’t leave you, and I cannot help but continue to explore such topics and share my findings and opinions through outlets like this. 

Beyond the internet, AI poses a tremendous challenge, with both its opportunities and its risks. Cheating obviously falls in the risks category. It challenges how accomplishments are evaluated and the results shared with learners and other interested parties (e.g., employers, those involved in competitive selection processes for limited-enrollment programs, the instructor in subsequent courses). It also poses a challenge to our efforts to craft assignments we are confident will contribute to student learning. If tasks are not completed as we assume, we cannot trust the markers we use to evaluate what students know, nor can we rely on them to guide our decisions about when to move on and what we can assume will make sense in new instruction. 

Without my colleagues, I now must rely more on what I can read or find online to form my own opinions. This is a difficult and relatively recent problem and little I would regard as proven seems to be available. There is plenty of advice and personal perspectives and folks willing to offer books on the topic. I might as well offer my own perspective on a specific instructional situation, since, at present, ideas focused on specific tasks in a specific type of classroom are the only ones I don’t immediately find myself arguing with. 

The Opposite of Cheating

I have been reading The opposite of cheating: Teaching for integrity in the age of AI (Gallant & Rettinger). It is well written and well referenced, but the type of source I find myself both rejecting and applauding when it comes to specific recommendations. Typically, a negative reaction stems from the impracticality of a suggestion given my own circumstances. I doubt it is reasonable I should expect authors to create a master model differentiating when a specific idea can be applied as that would add too much complexity and readers need to be active participants in finding what they should take from a resource. Anyway, this book made a point that sparked what follows. 

Writing to Learn

Written products played a significant role in some of the courses I taught. I assumed the products were a) an incentive to read the sources I expected students to read and listen to presentations I and students made, b) a way to demonstrate understanding and depending on the assignment consider applications, c) a task that involved the student in thinking in ways that led to understanding and retention, and d) a way to evaluate students. Having students write in isolation is one of those common tasks that has come under suspicion because of AI. 

Back to the “Opposite of cheating”. One of the authors’ general suggestions is to evaluate the process, not the product. I wrote a post some time ago making a very similar point. I think it helpful to explain why emphasizing what I would describe as subprocesses allows not only what might be described as surveillance, but also a superior instructional approach. 

Similarity to the strategy of showing your work.

    Yes, a requirement in what is probably math classes that you show your work was partially a check on whether a student had done the work, but just as important it was a record of the processing that was involved. A student and the teacher had access to the student’s externalized thinking. This visible record might be used by the student when the process breaks down and must be adjusted. It also provided someone else the opportunity to follow the student’s logic. The concept of externalized thinking has many applications for those who propose that cognitive research is useful to educational issues.

My long term interests in showing your work have focused more on writing and a specific application of writing often called writing to learn. Given a writing to learn task, assigned by a teacher or taken on as a personal strategy, a student could, of course, simply start writing or feed a prompt to their AI tool of choice. Here again, the “show your work” strategy can serve as both a check that you have done the work and a benefit to deeper thinking. 

I have been influenced by the logic and justification of advocates of personal knowledge management and the second brain. These concepts, when considered carefully, are clearly process-oriented: engaging purposefully and thoughtfully in specific processes benefits the products they produce, and externalizing processes that could be performed internally enables them to be performed more skillfully. I have long been interested in the Writing Process Model (Flower & Hayes) and variations. These researchers sought to develop a model that identifies the processes of writing and how the processes interact to create a written product. One benefit they proposed for such a model was the identification of component skills, allowing more efficient development of proficiency in individual skills. Identification of processes could guide both the topics researchers pursue and the instructional practices relevant to the classroom.

Connection with the topic of mitigating cheating

Let me start with this claim: an externalization requirement can serve both the purpose of ensuring that a process has been executed by a person and the educational goal associated with the assigned task. I think this works well when the goal is writing to learn.

I already indicated that writing to learn (or learning to write) can be broken down into subprocesses. Rather than relying on the Writing Process Model, allow me to offer a simpler approach for this situation.

In order to complete a writing to learn assignment, a student must:

Read the content

Identified what she felt are important ideas in the content

Processed this collection of ideas to understand and apply

I assume this is acceptable as a gross level description. If an educator relies on only the product turned in, with AI the educator must guess whether any of these tasks had actually been performed by a given student.

Those of us who make use of Personal Knowledge Management tools engage in these processes even though we are not accountable to an educator responsible for our skill and knowledge development. We do these things because we believe they deepen our understanding and strengthen our ability to craft better products.

We integrate a variety of tools while we read that would allow someone else to agree that we have in fact read.

Most of these tools involve highlighting and annotation as part of the reading process. The highlights and notes serve as an external representation of what we regard as important ideas in the content.

We then extract highlights and annotations from the original context so we can store and manipulate these elements more effectively. Having these elements separated and independent allows their long-term access and allows further processing such as linking, tagging, and secondary elaborations to occur. We value this growing and ever-modifiable collection as what has become popular to describe as a second brain that can be searched and explored for new insights and the generation of products.

The tools we use are ever improving and the skills in using these tools are being constantly scrutinized in search of greater efficiency and effectiveness.

The tools are there and it is easy to find free options. There is long-term benefit in learning to use such tools as skills relevant to lifelong learning. Why not teach these techniques to students and use the potential side benefit of accountability?

Hypothes.is as a starting point – try it you might like it

I first used Hypothes.is because I was interested in social note-taking with students. Simply put, this perspective argues that there may be benefits to a system that allows students to share notes. What did others find interesting or valuable in an assigned reading, and what might comments they made in response to what they highlighted as important reveal that others may not have considered?

This same tool could be applied such that an individual’s highlights and notes be available just to the instructor rather than the entire class. This covers “was it read” and “were ideas I thought important identified? 

The process for exporting from Hypothes.is works like this:

How to Export Annotations

Activate Hypothesis: Go to the webpage or document you’ve annotated and open the Hypothesis sidebar.

Open Sharing Menu: Click the “Share” button.

Select Export: Choose the “Export” tab.

Select Annotations: Use the dropdown to choose which user’s annotations to export (your own, a specific group, etc.).

Choose Format: Select your desired file type (e.g. HTML, plain text).

Export: Click the “Export” button to download the file, or “Copy to clipboard”. 

A screenshot of Hypothes.is in use. Hypothes.is is a browser extension so the content must be something online or something you can open in a browser (e.g., pdf). The content window on the left is where the reader highlights, annotates, and reads. The highlights and notes appear in the column on the right. 

Organize and Elaborate

At this point, I would now bring individual elements into a tool such as Obsidian, which I would not hesitate to introduce to college students. This tool is designed to store a large collection of individual idea notes, tag them, create links among them, and extend individual notes by generating secondary notes (elaboration). I raise this tool as an opportunity, not because there are no other options. Perhaps this mention of this tool will raise the curiosity of those willing to go a little deeper. 

There are other basic ways to do this. In the next stage before writing, you might open a document in any word processing tool and copy and paste individual notes or ideas from the notes or highlights into this document. As you proceed, you might cut and paste from this working document to better organize topics and integrate them into your final product. Even with the many personal knowledge management tools I use, I often take this simple approach when approaching the final stage of a project. I might cut and paste chunks of text and citations from the content I have accumulated into a common document. Often, this is not just about collecting ideas from a single source but bringing together ideas from multiple sources. I open this “collection” document in a separate word processing window and work from this narrowing of material into a draft of the product I am creating. 

Some writing tools even offer visible ways to do this. For example, Scrivener provides notes as note cards that can be moved around in a space to explore organizational options. Even if you do not intend to use a tool like this, visualizing the approach may be helpful. The “corkboard” option in Scrivener is shown below. Here you can see how individual project-related notes have been moved to this corkboard. The notes can be dragged around to create an optimal structure.

Summary  

This post focuses more on a concept for discouraging AI cheating more than on a detailed tutorial for using the tools involved. The core idea is that tasks can be assigned that are both beneficial for applying subskills to the writing-to-learn process and useful for documenting students’ completion of these subskills. I have identified specific tools and tactics, but there are likely alternatives for everything I have used as an example.

Source

Gallant, T. & Rettinger, D. (2025). The Opposite of Cheating: Teaching for Integrity in the Age of AI (Vol. 4). University of Oklahoma Press.