Pilot prototype for educator co-design. AI-generated questions and scores require educator review and should not be used as the sole basis for consequential grading.

Oral Assessment, Designed With Educators

Move beyond multiple-choice. CALLIOPE (Conversational Assessment for Learning, Listening & Intelligent Oral Performance Evaluation) supports spoken dialogue, source-grounded probing, six-dimensional draft evaluation, and educator review in one pilot workflow.

Calliope AI Dashboard Interface showing evaluation transcripts and radar charts

Advanced Platform Architecture

A responsive prototype for designing, conducting, and reviewing conversational assessments.

Multi-Source Context Engine

Add public web pages or plain-text files. CALLIOPE uses the extracted material as approved context for questions and draft evaluation; educators remain responsible for checking accuracy.

Real-Time Turn-Taking

Powered by local WebAudio VAD. Calliope listens, waits for the student to finish their thought, and responds conversationally without requiring UI clicks.

Six-Dimensional Evaluation

After a session, configured AI evaluators produce draft evidence across six dimensions for educator review and adjustment.

1. Content Accuracy

Estimates factual correctness against educator-provided source material, with the result presented as a reviewable draft.

2. Explanatory Depth

Measures comprehension beyond rote memorization by extracting causal reasoning chains directly from spoken transcripts.

3. Scientific Reasoning

Reviews Claims-Evidence-Reasoning (CER) structure, falsifiability awareness, and possible logical gaps after the session.

4. Systems Thinking

Analyzes semantic breadth across domains and the student's ability to articulate complex feedback loops and emergent behaviors.

5. Communication Clarity

Evaluates logical cohesion, pacing (Words Per Minute), and filler word frequency to ensure effective articulation.

6. Metacognitive Awareness

Reviews observable self-correction, stated uncertainty, and strategy reflection as evidence of metacognitive monitoring; it does not infer a learner's internal confidence.

Built for Educators

Prototype tools for creating learner-specific sessions, reviewing transcripts, adjusting AI draft scores, and publishing feedback.

Intuitive Builder

Assign a session to a learner, configure its duration and opening prompt, attach source material, and set rubric weights.

Actionable Analytics

View aggregate classroom performance charts or drill down into individual student transcripts to identify systemic knowledge gaps.

Role Management

Secure Role-Based Access Control (RBAC) ensures students only see their own assessments while educators maintain full oversight.