Created by Lauren Kilefner (lkilefner)
Hi there! I’m Lauren! I'm a former K–12 math and robotics teacher who believes that educators are the magic, and that AI should support great teaching, not replace it. This repo shows how I evaluate AI-generated responses for clarity, accuracy, tone, developmental appropriateness, and usefulness for real students and teachers.
This is the same lens I use in the classroom:
- Does the explanation make sense to the learner?
- Is the tone supportive and confidence-building?
- Are the steps clear and free of unnecessary complexity?
- Will this help a student move forward?
| Folder | What’s Inside | Why it Matters |
|---|---|---|
| /prompts | Sample K–12 academic + feedback prompts with ground truth exemplar answers | Demonstrates how I define correctness and clarity |
| /rubrics | A simple scoring rubric for evaluating AI responses | Shows I can evaluate consistently and transparently |
| /datasets | Small JSON + CSV datasets used for scoring practice | Reflects readiness for LLM testing workflows |
| /dashboards | Notes on how I would visualize and track evaluation results | Shows how I think about patterns & quality over time |
-
Fractions with Pizza (Grade 3)
https://github.com/lkilefner/llm-quality-evaluation-examples/blob/main/prompts/01_fractions_grade3.md -
Photosynthesis (Grade 5)
https://github.com/lkilefner/llm-quality-evaluation-examples/blob/main/prompts/02_photosynthesis_grade5.md -
Encouraging Math Feedback (Supportive Tone)
https://github.com/lkilefner/llm-quality-evaluation-examples/blob/main/prompts/03_tone_student_feedback.md -
Apple Subtraction Word Problem (Edge Case)
https://github.com/lkilefner/llm-quality-evaluation-examples/blob/main/prompts/04_word_problem_edge_case.md -
Responding to Student Frustration (Emotional Support Case)
https://github.com/lkilefner/llm-quality-evaluation-examples/blob/main/prompts/05_emotional_support_message.md
When reviewing LLM outputs, I look for:
- Accuracy — Is the content correct?
- Alignment — Does the response follow the prompt and grade level?
- Clarity & Cognitive Load — Is the language simple and structured?
- Tone — Does the response encourage and support the learner?
- Actionability — Can a student or teacher use this to learn or improve?
In other words:
The explanation should feel like something I’d proudly hand to my own students.
LLMs are incredibly powerful, but great teaching is personal, kind, and thoughtful. The work here reflects how I help ensure that AI explanations are not just correct, but also human, supportive, and age-appropriate.
- Former accelerated math teacher & robotics instructor
- Curriculum designer & instructional leader
- Passionate about making learning feel possible for every student
- Excited about how AI can reduce teacher overload and expand student support
If you'd like to connect:
LinkedIn: www.linkedin.com/in/laurenkilefner
Thanks for being here for caring about thoughtful, responsible AI use in education!