Inspecting CEDAR
Using AI to understand what the dashboard is actually doing
One of CEDAR’s core commitments is transparency: every analysis is produced from source code you can interrogate. AI assistants make that transparency practical — you don’t need to read code or be an R programmer to get answers to important questions. This page shows you how.
Getting started
CEDAR’s code lives in a public GitHub repository. GitHub has a built-in AI assistant (Copilot) that can answer questions about any code in the repo — no setup required.
- Go to the CEDAR repository on GitHub
- Look for the Copilot button in the top-right corner of the page — it looks like a small chat or sparkle icon
- Click it to open a chat panel on the right side of the screen
- Type your question in plain English
That’s it. You can ask about how a specific number is calculated, what a term means, or why two analyses might produce different counts from the same data. Copilot can search the codebase and explain what it finds.
If you’d rather use ChatGPT or Claude, you can copy the contents of any file in R/cones/ or R/branches/ and paste it directly into a chat. The example prompts below work with any of these tools.
Reading the code behind a result
Every CEDAR analysis lives in R/cones/. If you want to understand exactly how a number is calculated — what’s included, what’s excluded, what counts as a DFW — you can open the relevant file and ask an AI to explain it.
Example prompts:
“I’m looking at
R/cones/enrl.R. Can you explain in plain language whatget_low_enrollment_courses()does and how it decides which sections to flag?”
“What does
get_stopout()measure? Specifically: what counts as a stop-out, and what’s the ‘gap’ it’s calculating?”
“In
R/branches/filter.R, what doesfilter_DESRs()filter out by default? I want to know what sections are excluded before any analysis runs.”
You don’t need to understand R to get useful answers. AI assistants are good at translating code into plain English — and when the methodology is in the code, asking about the code is asking about the methodology.
Checking a specific calculation
If a number in the dashboard surprises you, tracing it back to the code is often the fastest way to understand why.
Example prompts:
“In CEDAR’s credit hour calculation, it says only enrollments with passing grades count. Which grades count as passing? Show me where in the code that’s defined.”
“The drop rate table shows ‘early’ and ‘late’ drops. What makes a drop ‘early’ vs. ‘late’ in CEDAR’s calculation? Is it based on census date?”
“I see a DFW rate that seems high. How does CEDAR define DFW? Does it include D- grades? What about incompletes?”
Understanding data and terminology
CEDAR uses terminology drawn from Banner and institutional research conventions. When something isn’t clear, AI can explain concepts in context.
Example prompts:
“What’s the difference between a ‘home’ section and an ‘away’ section in CEDAR’s crosslist handling?”
“CEDAR shows ‘split-level’ courses separately. What makes a section split-level, and why does CEDAR treat it differently?”
“What does
term_typemean in the CEDAR data model? How is it different fromterm?”
Exploring what analyses exist
AI can help you navigate the codebase to find what’s relevant to your question.
Example prompts:
“I want to understand how CEDAR identifies courses that are trending toward cancellation. Which file or function handles that, and how does it work?”
“Is there a CEDAR analysis that compares enrollment patterns for transfer students vs. first-time freshmen? Where would I find it?”
“What analyses does CEDAR have for understanding course sequences — the order students take courses in?”
For developers writing new cones
If you’re contributing to CEDAR or extending it for your institution, the developer AI reference lives at AI-REFERENCE.md in the repository root. It contains the full data model, the cone pattern, and common code patterns — designed to be pasted directly into an AI chat when building something new.