AI as a Classroom Ally: Simulated Group Discussion

Background

One of my current classes, like so many classes across the country, is designed around a blended learning model. It's a high school diploma program with a mix of limited in-person synchronous time and the majority of the program available asynchronously via digital courseware.

Digital courseware specifically for high school diploma programs can be tricky. While there are many benefits to an all-in-one curriculum solution, aspects of the learning experience may lack the personalization, scaffolding, or other modifications necessary to connect with our students.

And then there's the elephant in the room—academic integrity.

How do we design for an authentic learning experience that leverages the strengths of digital courseware while limiting the amount of assignments that set off the plagiarism detectors?

So I dove into the courseware, looking for an assignment that seemed interesting and that I could reimagine and redesign, and came across a group discussion activity meant to be completed in person.

In the redesign, I decided to explore how AI can support collaborative learning in adult education—particularly in contexts like mine, which are typically individualized (HSE, ASE, distance learning)—to allow for a personalized experience that encourages authentic participation.

🧠 The Core Idea

Adult education often relies on independent learning models, while still expecting some level of participation.

That creates a tension:

How do we design for interaction when learners are not co-present?

In this activity, interaction is introduced through a simulated group discussion facilitated by AI.

🤖 What the AI Does

Learners choose a topic, add a small set of sources, and enter a discussion with AI-generated peers.

The AI:

• introduces multiple perspectives
• moves the discussion forward
• pauses to invite participation
• prompts reflection at key moments

The goal is not to generate responses, but to create a guided interaction.

🗣️ How the Discussion Works

The interaction begins before the discussion itself.

Learners start by identifying a debatable topic. The AI helps refine it—clarifying scope, suggesting angles, and prompting for questions worth exploring. Learners then add a small set of sources, which the AI uses to ground the discussion.

This setup step matters. It gives the conversation something to work with.

Once the discussion begins, the AI introduces a moderator and several fictional peers with distinct perspectives, backgrounds, and communication styles.

Two elements shape how the discussion unfolds:

Pacing
The AI pauses and waits for learner input before continuing. This creates space for participation—even in asynchronous settings.

Variation
Each persona contributes a different viewpoint and way of speaking, introducing disagreement and nuance into the exchange.

The discussion moves forward—but not automatically. The learner has to enter, respond, and re-engage along the way.

⚙️ How This Was Built

The Process

This work involved a mix of instructional design, prompt development, and iteration.

I used AI to:

• draft and refine the activity structure
• test how the simulation behaved
• identify where additional constraints were needed

The process was iterative:

generate → test → adjust → constrain

Over time, the focus shifted from producing strong responses to shaping how the interaction itself unfolds.

The Platform

I decided to use Google's NotebookLM app for several reasons.

• Students are already using the Google Workspace environment in other school-related areas, including other assignments in my class
• The app is free and accessible
• The walled garden format of the app is ideal for the scope of the activity
• AI literacy can be introduced in a tangible, valuable way

Delivery

SkillBlox really shines here because what I'm trying to do is build a clean 'playlist' of information and assets ready to be shared with both educators and students. For learners, it doesn't require an additional login and can be easily shared by teachers via link, QR code, Google Classroom, or a variety of other integrated channels.

🔶 What Shaped the Design

Several elements had a noticeable impact on how the activity functioned:

Persona design
The peers carry the interaction. Differences in perspective, language, and experience create a discussion that feels less uniform and more representative of real participation.

Exemplars
A sample discussion helps establish tone and expectations. Without it, responses tend to become more generic.

Tone
The AI is guided to be conversational, respectful, and accessible—reducing friction for learners who may be hesitant to participate.

⚠️ Implementation Considerations

This type of activity depends on how well it aligns with actual learning conditions.

Access remains a factor—mobile use, connectivity, and the ability to re-enter an activity all affect participation.

Learners also need orientation to working with AI. Clear expectations and simple entry points make a difference.

The instructor role shifts as well: less time delivering content, more time designing and supporting the learning experience.

🔁 A Reusable Pattern

The activity follows a structure that can be adapted:

Individual task → AI-mediated interaction → structured reflection

This pattern can extend across contexts:

• ESL conversation
• writing and revision
• research-based discussion
• workforce preparation

In practice,
-Jerry