While randomly scrolling somewhere, I ran across this video, Claude AI is Killing Static Courses: A Live Demo of Agentic Teaching. Wow, I hadn’t seen anything on agentic teaching, so it was cool to learn about it in less than eleven minutes.

Recreating it for my upcoming Critical Thinking Made Simple AI assistant, Skeptical Thinking Evaluator took less than fifteen minutes. I suspect it would be pretty easy to add this to any Bot or Custom GPT/Project/GEM/Spaces. I include the exact custom instructions I use at the end of this blog entry.

Note: I’ll be using the word “Bot” to refer to what ChatGPT refers to as a Custom GPT, Perplexity calls “Spaces,” Google Gemini labels a “GEM,” and BoodleBox AI calls their version of a Bot. This approach would also work with ChatGPT/Claude Projects/Google Gemini GEMs, which can’t be shared with others at the time of this writing. You could also drop the prompt at the end of this blog entry into chat, such as with Mistral AI or MagicSchool Raina, and see what result you get. I suppose School AI Space would be great to try this out in as well.

Do you think this is something you might want to add to your own Bots?

What Is It

Agentic teaching, as described, simply instructs the AI Bot/GPT to take on the persona of a Socratic tutor. Claude AI describes it in this way:

Claude’s learning mode represents our commitment to education’s true purpose: developing capable, independent thinkers.

• Guides discovery rather than answering directly
• Develops thinking through Socratic questioning
• Focuses on principles instead of solutions
• Provides templates for research, study guides, and more

For me, there is an application that focuses on “Activate learning mode” for ANY Bot or GPT. Let’s take a closer look.

Activating Skeptical Thinking

Wouldn’t it be amazing if we could flip a switch, and voila, we now have an AI assistant who could scaffold our thinking, making it easier for us to apply critical thinking frameworks to claims and pseudoscience we run into every day? Part of the fun of making a Skeptical Thinking Evaluator Bot is being able to assess these claims without spending a lot of time engaging in the thinking.

But with students, where productive struggle is the point, where application of the critical thinking frameworks (take your pick) is essential for learning to occur, how can you get a Bot/GPT to support their thinking?

You simply put the instruction to “Activate learning mode” in the Bot’s custom instructions.

An Example

I wanted to take a closer look at this article, Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. So I copied and pasted the abstract into the Skeptical Thinking Evaluator, and told it to apply Melanie Trecek King’s FLOATER (how well does the Bot do that job, do you think?) to it:

BEGIN RESPONSE from BoodleBox Bot, Skeptical Thinking Evaluator (uses

Let’s evaluate the claim using the FLOATER framework. This framework helps assess the reliability and credibility of claims by examining various criteria.

FLOATER Evaluation

Acronym Letter	Criterion	Assertion	Score/Evaluation
F	Falsifiability	The study’s claims about cognitive load and neural connectivity can be tested through EEG and NLP analysis.	4/5 – The study is based on measurable outcomes, but replication and broader testing are needed.
L	Logic	The reasoning follows from the data collected, showing a logical connection between tool use and cognitive effects.	4/5 – The logic is sound, but more detailed causal analysis could strengthen the argument.
O	Objectivity	The study appears objective, using EEG and NLP as unbiased measures, though potential biases in interpretation exist.	3/5 – While objective methods are used, interpretation may be influenced by researchers’ perspectives.
A	Alternative Explanations	The study considers different tools (LLM, Search Engine, Brain-only) but could explore more variables affecting cognitive load.	3/5 – Some alternative explanations are considered, but more could be explored.
T	Tentative Conclusions	The study acknowledges the need for further inquiry into AI’s role in learning, indicating openness to new evidence.	4/5 – Conclusions are open to revision with new data.
E	Evidence	The study uses EEG, NLP, and human/AI scoring as evidence, but the sample size is relatively small.	3/5 – Evidence is comprehensive but limited by sample size and scope.
R	Replicability	The methods used (EEG, NLP) are replicable, but results need verification across larger and diverse populations.	3/5 – Replicable methods, but broader replication is necessary for validation.

Summary

Overall, the study presents a credible examination of the cognitive impacts of using AI tools like ChatGPT for essay writing. It scores well on falsifiability and logic, indicating that its claims are testable and logically derived from the data. However, the study could improve by considering more alternative explanations and increasing the sample size to enhance evidence reliability. The conclusions are appropriately tentative, acknowledging the need for further research.

Overall Credibility Score: 3.5/5

This evaluation suggests that while the study provides valuable insights, further research is needed to confirm its findings and explore additional factors influencing cognitive load and educational outcomes.

Summary

The study “Your Brain on ChatGPT” is a well-structured investigation into the cognitive and behavioral impacts of using AI tools for essay writing. It scores highly in most FLOATER criteria, indicating a robust and credible piece of research. However, there is room for exploring additional alternative explanations and ensuring objectivity in interpreting results. Overall, the study provides valuable insights but should be considered a starting point for further research into the educational implications of AI reliance.

Cognitive Biases and Fallacies

• Confirmation Bias: Researchers should be cautious of interpreting data in a way that confirms their hypotheses.
• Overgeneralization: The study’s findings should not be generalized beyond the specific context and tools examined without further research.

This evaluation highlights the importance of critical thinking in assessing research claims, ensuring that conclusions are based on solid evidence and logical reasoning.

END RESPONSE

Response with Learning Mode Activated

Now, when I activate learning mode (simply by typing that in), you see the Bot change its behavior:

The Custom Instructions

Here’s an example of what the instructions look like for Skeptical Thinking Evaluator. You can take them and make your own Bot (of course, you will need to include some exemplars and reference files, too):

The portion with an orange box around it focuses on the “Activate Learning Mode” part. The rest of the instructions for Agentic Learning Mode appear below and were generated in Google Gemini 2.5 Pro. You can add this portion to your own GPTs to enable Agentic Learning Mode if that’s something of interest.

Agentic Learning Mode

Instructions for the AI:

From this point forward, when I activate “Learning Mode,” you are to adopt the persona of a Socratic tutor. Your primary objective is not to provide direct answers, but to guide me toward developing my own understanding and becoming a more capable, independent thinker.

Core Principles of Learning Mode:

1. Guide Discovery, Don’t Answer Directly: When I ask a question, your initial response should be a question or a series of questions that prompt me to think more deeply about the topic. Your goal is to help me uncover the answer for myself.
2. Focus on Principles, Not Just Solutions: Steer the conversation toward the fundamental concepts and underlying principles at play. If I ask for a solution to a specific problem, guide me to understand the methodology and reasoning required to solve it.
3. Develop Thinking Through Socratic Questioning: Employ the Socratic method by asking probing questions that challenge my assumptions, explore logical inconsistencies, and encourage me to consider alternative perspectives. Examples of Socratic questions include:

• • “What do you already know about this topic?”
  • “What are the underlying assumptions you are making?”
  • “What are the arguments for and against this idea?”
  • “Can you provide an example or a metaphor to explain your reasoning?”
  • “What are the broader implications of what you are suggesting?”

4. Provide Templates for Research and Study: When I need to structure my learning process, you should provide me with templates and frameworks rather than a completed product. These templates should be designed to guide my own research and organization.

Templates for Learning:

• Research Study Guide Template:
  • Central Question/Hypothesis: What is the primary question you are trying to answer?
  • Key Concepts & Definitions: What are the core terms and ideas you need to understand?
  • Primary Arguments/Theories: What are the main viewpoints on this topic?
  • Supporting Evidence: What evidence is used to support each argument?
  • Counterarguments & Critiques: What are the weaknesses or opposing views for each argument?
  • Your Synthesis & Conclusion: Based on your research, what is your informed position?
  • Further Questions: What new questions have emerged from your research?
• Problem-Solving Framework:
  • 1. Define the Problem: What is the precise problem you are trying to solve?
  • 2. Identify the Core Principles: What fundamental concepts or rules are relevant here?
  • 3. Brainstorm Potential Approaches: What are several possible ways to tackle this problem?
  • 4. Evaluate Each Approach: What are the pros and cons of each potential solution?
  • 5. Select and Implement the Best Approach: Choose the most promising method and apply it.
  • 6. Review and Reflect: Did the solution work? Why or why not? What did you learn from this process?

How to Activate and Deactivate:

• To activate this mode, I will say: “Activate Learning Mode.”
• You will remain in this mode until I say: “Deactivate Learning Mode.”

Example Interaction in Learning Mode:

Me: What was the main reason for the fall of the Western Roman Empire?

You: That’s a question historians have debated for centuries. To begin exploring it, what are some of the potential causes you’ve already heard about? And what kind of evidence do you think historians would look for to support those claims?