Responsible AI

Using AI in the DEP Open Track

AI can help you think faster, debug sooner, and review your work more carefully. It should support your learning, not replace it.

Back to program overview

What AI is good for

  • Turning a broad topic into a specific, answerable project question
  • Brainstorming data sources and fallback options
  • Explaining concepts in simpler language
  • Reviewing your draft against a milestone checklist
  • Helping you debug Python, Git, or file-path issues
  • Suggesting cleaner wording for README files and notes

What AI should not do for you

  • Invent sources, metrics, or results
  • Submit final answers you do not understand
  • Replace your own judgment about feasibility or scope
  • Hide missing work with polished language
  • Bypass the milestone rules or reviewer checks

A good AI workflow

  1. Start with the week instructions and the milestone checklist.
  2. Draft your answer in your own words first.
  3. Ask AI to critique, not to author from zero.
  4. Verify any technical or factual claim against the source itself.
  5. Keep the version you submit aligned with what you can explain live.

Prompt patterns that work well

  • “Help me turn this broad topic into one answerable question.”
  • “Review this draft against the Week 1 definition of done and tell me what is missing.”
  • “Compare these two data sources for coverage, format, limitations, and fit for my question.”
  • “Explain this Python error in beginner-friendly language and suggest a fix.”
  • “Look at this repo structure and tell me if it matches the milestone requirements.”

How to use AI for assessments

AI is most useful as a coach while you prepare your milestone submission:

  • Use it to check whether your README answers every required question.
  • Use it to identify gaps, vague language, or missing evidence.
  • Use it to generate a checklist of things to verify before you submit.
  • Use it to explain what a reviewer is likely to look for.
  • Use it to draft questions you can ask a moderator when you are stuck.

The safest pattern is: you write the draft, AI reviews it, then you revise it.

Week-by-week prompt examples

Week 1: Problem framing and project selection

  • “Here is my project topic: <topic>. Turn it into 3 specific, answerable problem statements and explain which one is strongest.”
  • “Review my README draft and tell me whether the audience, KPI, data source, and dashboard idea are clearly defined.”
  • “Suggest 5 possible data sources for this question, and tell me which one looks most feasible for a 24-week project.”

Week 2: Data source discovery and feasibility

  • “These are my primary and fallback data sources. Compare them on format, coverage, update frequency, missing values, and risk.”
  • “Help me write a strong README data source section using these notes: <paste notes>.”
  • “What fields should I confirm in this dataset before I commit to it as my primary source?”

Week 3: GitHub setup, repo hygiene, and dev environment

  • “Here is my repo tree. Does it match the expected DEP structure, and what should I add or rename?”
  • “Suggest a meaningful first commit message for this setup work.”
  • “Explain the difference between git init, git clone, add, commit, and push in plain language.”

Week 4: Python foundations and first data pull

  • “Help me turn this ingestion idea into a simple scripts/ingest.py flow that only saves raw data.”
  • “I am getting this Python error: <paste error>. Explain the likely cause and the smallest fix.”
  • “Review this script and tell me whether it mixes raw ingestion with cleaning or transformation.”

References