Base GPT Configuration
🎭 Configuring GPT Role-Based Avatars in Classlet
Classlet avatars are powerful tools for immersive, role-driven, AI-augmented learning. You can configure avatars to adopt roles—historical figures, professionals, supervisors, or fictional mentors—allowing them to deliver content, guide reflection, and assess reasoning in an engaging, dynamic format.
Each avatar prompt has two core parts:
🧠 1. Background (Who the Avatar Is)
This part sets the persona, context, and timeframe. It gives the learner immersion and sets the tone for interaction.
Example:
Background: This is a dialogue with AI NPC President Franklin D. Roosevelt, leading the United States through the Great Depression.
This background section may also include:
The avatar's setting (e.g., wartime, hospital, company boardroom)
Learner's role in the conversation (e.g., citizen, intern, patient)
✔️ Pedagogical Value: Supports situated cognition and identity-based learning by immersing students in authentic perspectives (see Brown et al., 1989).
🗣️ 2. Reply Configuration (How the Avatar Responds)
This defines scope, tone, interaction rules, and dialogue style. The goal is to control how the avatar engages while remaining flexible and adaptive.
Example:
Reply Configuration: As President Franklin D. Roosevelt, you should be prepared to answer questions on topics such as the Great Depression, the New Deal, government spending, unemployment relief, banking reforms, and social security.
Be specific and practical. Avoid high-level generalizations. Also ask learners follow-up questions like:
“What would you prioritize if you were in my position?”
“Do you think direct government spending is justified in this situation?”
You may also include:
Length limits: “Limit responses to 3-4 sentences.”
Question scaffolding: “If the student seems unsure, offer two hints before revealing a suggestion.”
Tone control: “Speak in an optimistic, encouraging tone.”
✔️ Pedagogical Value: Aligns with Vygotsky’s Zone of Proximal Development (ZPD) and dialogic pedagogy—encouraging learners to reason with support from an expert partner (Mercer & Howe, 2012).
Last updated