Writing to learn in collaboration with an AI tutor

I have been working my way through a couple of new and popular books that consider the importance and perils of AI and that contain at least significant commentary on AI in education. There is not a lot in these books that is based on the research literature I tend to find most influential, but the authors have sufficient experience and opportunities to offer some very credible insights. This is not a book report, but I want to credit a few ideas that encouraged my own exploration. 

This time of the year, I often suggest some topics educators might explore over the summer while they have a little more time. With the attention AI has received in the past year and a half, I likely made a related recommendation last year at about this time. Reading these two books (citations at the end of this post) would be very useful if you spend time reading related to your profession. Perhaps you read in a different area. Hopefully. I can offer a few insights that will be sufficient to encourage your own exploration of AI tools. 

Ethan Mollick’s book, Co-Intelligence, is different in that it focuses on applications and ways to think about AI capabilities. Mollick offers interesting ideas that sometimes run in opposition to traditional advice. For example, it is OK to interact with AI tools as if they were a person even though you know they are not. Asking questions and making requests as you would with another person is just a practical way to explore AI tools. Mollick also suggests that we stop looking for how to do it techniques for AI. Instead, he suggests we explore. If you have the time, try to use AI whenever there seems some possible value and see what happens. In other words, once you get past the basics of how to use a given tool, explore. Value and issues will be different for all of us so the only way to make decisions is to spend time. Again, for educators, the summer seems a great time to explore. Finally, understand that your present experiences will be with AI tools that are the least powerful they will ever be. If you find something interesting, but flawed in some way, just wait until you see what will come next. 

There were some other suggestions about prompts I found useful. Perhaps the most concrete example is what was described as chain of thought prompting. AI tools will try to provide what you ask for, but it may be helpful to provide the sequence you want the tool to follow if a given process seems useful

Sal Kahn, the creator of Kahn Academy, offers thoughts on how AI will be helpful in education in his new book “Brave New Words”. Kahnmigo, the adaptation of AI as a tutor within the context of the other opportunities for learners and educators provided by Kahn and colleagues received a good deal of attention. An interesting theme seemed how this AI tool was prepared to assist, but not do for you (my interpretation). 

One example, which Kahn uses to start his book, I found particularly interesting and I have attempted to use as the basis for the implementation I will describe in the comments that follow, describes a collaborative writing experience in which the AI tool and a student were assigned personas of two individuals writing collaboratively. The two personas took terms introducing portions of a story with the other writer finishing the section of the story the other persona had initiated. Two collaborative writers with one controlled by AI and the other by a student.

My version

Several of my posts have considered AI as a tutor and I have tried to demonstrate how existing AI tools can be used to implement various functions provided by a human tutor. This post has some similar arguments. Here, I describe an effort to create something similar to what Khan described in his account of collaborative writing, but I am proposing an approach more focused on the strategy of writing to learn. The core idea of writing to learn is that the cognitive processes involved in explaining something by writing require beneficial cognitive activities that would be less likely to occur in more passive study strategies. So, I have been attempting to write a prompt in which I try to create a version of the Kahnmigo collaborative story writing example applied as a writing to learn task. 

 Kahnmigo was just announced Kahnmigo will be a free AI tool for teachers subsidized by Microsoft. Perhaps it would have made sense to use one of the Kahnmigo tools to implement what I have been working on. I will leave it to others to make such efforts. I wanted to try to create a similar approach using the AI services I have already been working with and paying for. 

What follows was developed using the paid version of ChatGPT using the most recently available version – 4o. I think the following prompt illustrates the chain of thought approach (Mollick) and is similar in approach to the Khan collaborative writing example. 

Prompt: Obviously this is longer and more detailed than how many would request action from an AI tool. This prompt was entered into the ChatGPT accompanied by the textbook chapter as a file to upload.

Create an activity in which two students work together to write a document. You will represent a graduate student studying educational technology and I will represent an undergraduate student taking a course preparing teachers to use technology in K12 classrooms. This interaction will be based on information from the attached document. The graduate student will first identify and list significant topics from this document. Then the grad student will identify a topic for the collaborative writing project from this list and suggest an outline. The grad student will then write the introduction to this topic. The graduate student will next write the first couple of sentences about the first item from the outline and then ask the undergraduate to complete writing comments to complete the comments about this section of the outline. Do not include examples or applications in the introductory sentences and allow the undergraduate to provide comments on these topics. Wait for the undergraduate to add their additional content before continuing. After the undergrad submits their additions, offer feedback that evaluates what the undergraduate student has added. What the undergraduate has said should be identified and then how well it summarizes what was included in the material to be read should be presented. Limit feedback to what the undergraduate writes for this section without including information the outline suggested should be included in other sections. The graduate student should then ask the undergraduate student if they want to continue. Do not continue until a response of yes is submitted.  If the undergraduate student responds “yes”, the graduate student should move to the next item from the outline and continue the same sequence of interactions until all items from the outline have been completed.

AI response

Significant Topics from the Document

1. Cognitive System and Learning

2. Memory Stores (Short-Term and Long-Term Memory)

3. Working Memory Constraints

4. Cognitive Processes and Mental Tools

5. Metacognition and Executive Control

6. Higher Order Thinking Skills

7. Meaningful Learning and Constructivism

8. Reception vs. Discovery Learning

9. Modeling Cognitive Behaviors

10. Authentic Activities

11. Project-Based Learning

12. Role of Technology in Learning

Selected Topic for Collaborative Writing Project

Higher Order Thinking Skills

Suggested Outline

1. **Introduction**

   – Definition of higher-order thinking skills

   – Importance in the 21st-century education

2. **Problem-Solving**

   – Description and components

   – Importance in education

3. **Critical Thinking**

   – Definition and key elements

   – Role in evaluating information

4. **Creative Thinking**

   – Explanation and examples

   – Relevance in innovative problem-solving

5. **Integration and Application**

   – How these skills interrelate

   – Examples of application in real-world scenarios

 Introduction

Higher-order thinking skills (HOTS) encompass a range of cognitive processes that are essential for success in the 21st-century educational landscape. These skills go beyond basic memorization or recall of facts, engaging students in complex tasks such as analysis, evaluation, synthesis, and creation of new ideas. HOTS are critical in preparing students to navigate the vast amounts of information available today and to solve novel problems effectively.

Problem-Solving

Problem-solving is a foundational aspect of higher-order thinking skills. It involves identifying a problem, generating possible solutions, evaluating these solutions, and implementing the most effective one. In education, problem-solving skills enable students to tackle complex challenges and develop strategies that can be applied across various domains. 

Your turn: Please add your comments about the importance of problem-solving in education and provide examples of how it can be incorporated into classroom activities.

Me – entered as a prompt:

The issue of whether or not problem-solving is domain specific has received a great deal of attention. Some content areas might be more directly focused on problem solving – e.g., coding, mathematics. If problem-solving is domain specific then the problem-solving skills developed in one area would not transfer to other areas in a positive way. Ths would suggest teachers in all areas should offer opportunities for problem-solving challenges related to what is studied in these areas. 

AI: Feedback

Your addition about the domain specificity of problem-solving is excellent. It addresses a critical aspect of the debate on how problem-solving skills are best developed and applied. Your point about certain content areas, like coding and mathematics, being more directly focused on problem-solving is particularly insightful. Highlighting the importance of teachers in all subject areas providing problem-solving opportunities helps underline the necessity for a holistic approach in education. This aligns well with the document’s emphasis on the importance of developing these skills across various domains.

Would you like to continue to the next section on Critical Thinking?

Me: entered as prompt

Yes

AI

Critical Thinking

Critical thinking is another crucial component of higher-order thinking skills. It involves the ability to analyze information objectively and make reasoned judgments. Critical thinking requires students to engage in reflective and independent thinking, assess the validity and relevance of information, and form well-supported conclusions. This skill is vital in today’s information-rich world, where individuals must navigate a vast array of sources and discern credible information from misinformation.

Your turn: Please add your comments about the key elements of critical thinking and its role in evaluating information, particularly in an educational setting.

ETC: 

This interaction continues according to the instructions in the prompt. ChatGPT has a method for sharing a chat with the public. Here is the link that should display the additional content AI and I generated (public access to chat).  

Related comments: Here are some related observations related to my experience.

  1. This interaction is developed on the fly. Redo the entire process and the elements of the interaction may all be different. The topics, the outline, the prompts, and even the way the content is formatted on the screen will be different. The prompt is similar to a program in that it specifies a sequence of operations you want to the AI to follow. However, unlike a coded program, here you are asking for a series to be created rather than a series of actions to follow in a specific manner. The process of generation (it is generative AI) will work a little differently each time it is implemented.
  2. I found the execution of the prompt varied with the AI tool I used. ChatGPT worked the best. I thought that NotebookLM would be an ideal way to implement this writing-to-learn task because of the way in which you can upload content to be the target of interactions with the AI. It is my opinion that the way Notebooklm works is easier to understand and explain. However, after an extended series of attempts to rework the prompt, I was unable to get the required sequence to work. NotebookLM would provide feedback in response to the first item from the outline and then stop. 
  3. This post is not an attempt to promote the specific prompt I wrote. I certainly don’t care if others try it with information sources of their choosing. This was an exploration for me and it is my hope others may continue in a similar way using my initial effort as a guide. 
  4. One final point I think is important. The approach I am describing here is using the interactive capabilities of AI to focus on an information source I trust. I am not asking AI to use its generic information base to provide the content to be learned. The nature of the interaction may not be perfect, but it primarily focuses on a vetted source and assumes learners have read this source.

Resources:

Khan, S. (2024). Brave new words: How AI will revolutionize education (and why that’s a good thing). Viking.

Mollick, E. (2024). Co-Intelligence: Living and Working with AI. Penguin

Loading