Behavior-based persona and course-grounded memory architecture for simulated students¶

Design · All designs
Evidence · 1 claim (1 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

EducaSim's student-agent architecture synthesizes generative agents with personas and memories, domain-contextualized memories aligned with actual course material, and a classroom dynamic tree using LLM as a judge. Personas are defined by pedagogically relevant behaviors rather than demographic identity traits, including "(1) name, (2) engagement style, and (3) speech style". Course texts are chunked chronologically, tagged with low/medium/high engagement levels, and prepended with strings describing what the student does and does not know; retrieval scores memories by recency, importance, and cosine-similarity, returning the top-k.

Design Implications¶

Context¶

Requirements¶

Course materials must be ordered chronologically based on when a student would encounter them so teachers can refer to past material

Constraints¶

Agents are an approximation, not a perfect representation, of student behavior and cannot fully represent the diversity of human understandings or confusions

Target Learners¶

Teachers-in-training practicing instruction

Learning Goals¶

Realistic simulated student interactions grounded in actual course content

Claims¶

In a preliminary 2×3 controlled lesson study across five backbone LLMs, structured student agents produce more differentiated mastery and misconception traces than a baseline simulator [+W]

Educasim Simulated Section Teacher Practice Tool
CogEvolution framework: ICAP perception, IRT-based memory retrieval, and evolutionary cognitive state update
Reusable simulation architecture for dynamic relational learning environments

Examples¶

-

Key Sources¶

Cameron Mohne, Nicholas Vo, Dora Demszky, Chris Piech. (2026). EducaSim: Interactive Simulacra for CS1 Instructional Practice. https://arxiv.org/abs/2603.11444

Page metadata

Field	Value
`type`	`design`
`id`	`educasim-agent-persona-memory-architecture`
`title`	`Behavior-based persona and course-grounded memory architecture for simulated students`
`description`	`EducaSim's student-agent architecture synthesizes generative agents with personas and memories, domain-contextualized memories aligned with actual course material, and a classroom dynamic tree using LLM as a judge.`
`status`	`review`
`generated`	by `process:settle-candidates` on `2026-10-10`
`trust`	unverified — no one has confirmed this page's content
`sources`	1 — `cameron-mohne-2026`

Frontmatter is the page's machine-readable record — see the Schema & Guide.