Description, Explanation, and Prediction: The Step-Changes at the Heart of Leadership


The Learning Engine

9 June 2026

Description, Explanation, and Prediction: The Step-Changes at the Heart of Leadership

Your post-observation conversation with a teacher is happening in a few minutes, so you gather your notes.

You had reviewed the student performance dashboard before walking into the classroom; an AI agent had flagged the students for this teacher as underperforming on reading comprehension for three straight months.

Your observation showed the same pattern: Students were speeding through the text, answering questions quickly, then moving to the next section to finish the assignment.

The teacher enters; after exchanging pleasantries, you ask the teacher how the lesson felt.

“I think they got through the material – they answered most of the questions.”

You nod, then ask what the students seemed to struggle with during the lesson.

“Some of them just didn’t seem engaged; a few were rushing.”

You nod again, then the conversation continues for another 10 minutes. The conversation confirms the pattern from the dashboard and your observation: Students are underperforming on reading comprehension. The teacher recognizes this; everyone agrees that change needs to happen.

Later that evening, you are reflecting on the situation. There seems to be a deeper issue that you cannot reach – but what is the deeper issue?


This piece follows from Philosophy, Learning, and AI Agents – The Structure Under Every AI Interaction, which introduced the philosophical structure for both leadership and AI agents; we are going deeper into using AI agents well.


The deeper issue is the kind of pattern.

There are three levels a pattern can reach: Descriptive, explanatory, and predictive.

A descriptive pattern names what happened – “students are underperforming on reading comprehension.” This description was verified from the dashboard, your observation, and the post-observation conversation.

An explanatory pattern names the mechanism – “students are speeding through the text because they have no strategies for monitoring their understanding.” The explanation connects the observable behavior to a mechanism that is inside the student’s Personal World 2.

A predictive pattern names a way to test a specific mechanism – “if students practice a specific self-monitoring strategy during reading three times per week for six weeks, comprehension scores will improve.” To be meaningful, the prediction must be tested and updated against real classroom results.

Each of the three patterns has a sliding scale from “thin” to “robust,” but require a step-change to go from one pattern to the next pattern.


AI Agent as the Other

AI agents can only create descriptive patterns.

The AI agent works entirely within the Conceptual World 3 – trained on the written record of human knowledge and given data from various sources. As the image below shows, the AI agent has no direct connection to the Physical World 1.

The result is that AI agents can move from thin descriptive patterns to robust descriptive patterns, with the density of description increasing with more data sources. However, the step change into the explanatory pattern requires a mechanism in the Physical World 1, which cannot happen because the connection is broken.

The AI agent can create a statement that sounds explanatory – “students lack self-monitoring strategies” – but that statement cannot be tested by the AI agent. The AI agent has no way to see if the statement works with these students by this teacher; the testing must come from the teacher and leader.


Person as the Other

A person can create all types of patterns.

As the image below shows, the person (in this case, the teacher) has full access to the Physical World 1 – plus a connection to the leader through the World 2-to-World 2 arrow.

The result is that the teacher can move along the sliding scale in each type of pattern and make the step-change from descriptive to explanatory to predictive. The teacher can check to see if the students have any self-monitoring strategies, then use that information to take action.

In addition, both the teacher and leader can discuss which mechanisms to test. This discussion will deepen the teacher’s and leader’s understanding of classroom practices, helping the teacher and leader grow as educators.


Back to your reflection on the deeper issue.

Your observation, the AI agent, and the teacher confirmed a descriptive pattern – but the post-observation conversation stopped there.

Your role as the leader was to push the conversation towards explanations and predictions, listening to the teacher’s response.

If the teacher restates the description – “they’re not engaged” – then you must guide the teacher into explanations for the lack of engagement. As the teacher moves towards explanations and predictions, your role shifts to co-experimentalist: Design a classroom move that will be tested with the students.

Both the leader and teacher continue to learn and grow by moving from description through explanation and into prediction – compounding effort into expertise.


Before the next post-observation conversation, locate the breakdown yourself. The Learning Breakdown Diagnostic is a scenario-based tool that gives you two coordinates in twelve minutes – where the breakdown lives and how to make the effort compound.

Have comments or questions about any part of this newsletter? Please reply and let me know I respond to every email!

Did someone forward this newsletter to you? If you want to start receiving the newsletter sign up here.

600 1st Ave, Ste 330 PMB 92768, Seattle, WA 98104-2246
Unsubscribe · Preferences

The Learning Engine

We teach you the principles of learning, helping you understand and apply the principles of learning in your leading, coaching, and teaching. By using the principles of learning, your leading, coaching, and teaching will be more effective!

Read more from The Learning Engine

The Learning Engine 2 June 2026 Your AI Agent Found a Pattern; Now What? The Instructional Leader’s Role in Connecting Solutions to Your Context Your district’s new AI agent is now connected to a student performance dashboard. You ask the AI agent to find patterns in the dashboard; one of the patterns is that students seem to be underperforming on reading comprehension. Without checking on the students and teachers, you ask the AI agent for recommendations. The AI agent answers quickly and...

The Learning Engine 28 May 2026 One Surface Issue, Two Underlying Challenges: A Diagnostic for Every Instructional Leader Your conversation with a teacher looked productive. The teacher was engaged, asking good questions and leaving with a clear next step. You sat with the notes after the conversation, feeling a disconnect between the productivity in the conversation and what would happen next. There was a pattern from the previous three times together – productive-looking conversations with...

The Learning Engine 24 May 2026 Philosophy, Learning, and AI Agents – The Structure Under Every AI Interaction The morning started brightly. Each prompt to the AI agent gave you information, so you seemed to be making progress towards your goal. But, frustration was mounting. Some part of the information was off; sometimes the mistakes were obvious, with other errors arriving subtly. [Five hours later] Frustration has turned to anger, both at yourself and the AI agent. What am I doing wrong?...