Core 1 · Foundations & Capabilities
Core 1 · Foundations & Capabilities
What generative AI is, which Drexel tools to use, and how to prompt for useful results
One hour on Zoom | Have your browser open next to Zoom | Runs weekly in the fall term and repeats in spring — dates announced by school email
What You’ll Leave With
No experience needed. If you have never opened ChatGPT or Copilot, this session is for you. By the end of this hour, you will be able to:
- Say in one sentence what tools like ChatGPT do, and what they get wrong
- Know which Drexel tools to use and what information never goes into them
- Sign in to ChatGPT Edu, type a request, and ask a follow-up
- Use four simple questions (PTCF) to get a better answer
Run of Show
| Time | Segment |
|---|---|
| 0:00–0:08 | What AI is, which tools to use, and the one data rule |
| 0:08–0:17 | What it can do: six real tasks (instructor demo) |
| 0:17–0:23 | How AI gets things wrong, and who is responsible |
| 0:23–0:33 | Guided tour: ChatGPT Edu, step by step (follow along) |
| 0:33–0:44 | Activity 1: Your first three requests |
| 0:44–0:55 | Activity 2: Better requests with PTCF |
| 0:55–1:00 | Wrap-up and one thing to try this week |
- Put Zoom and your browser side by side, or use two screens if you have them. You’ll watch the instructor’s shared screen and try each step in your own browser.
- React with 👍 in Zoom when you finish a step, so the instructor knows when to move on.
- Stuck? Say so in the Zoom chat. We’ll help you get signed in and get an answer back.
- Prompt: what you type into the box. A question, a request, or instructions.
- Chat: one conversation. The tool remembers what was said earlier in the same chat, but not in other chats.
- Model: the AI system underneath the tool. You can use the default.
- Hallucination: when the AI states something false as if it were true.
What AI Is, and Which Tools to Use (8 min)
What it is. ChatGPT and Copilot write text by predicting what words should come next, based on patterns in a huge amount of writing. That makes them good at drafting, summarizing, rewriting, and explaining. It also means they can sound confident and still be wrong. They are not search engines or databases.
Which tools to use. For Drexel work, use ChatGPT Edu or Copilot Chat, both signed in with your Drexel account. Other tools, including personal ChatGPT, Claude, and Gemini, are for public or low-risk information only. Full details are on the Tools & Access page.
Never paste in health information about a person, student records, Social Security numbers, or identifiable research data. When in doubt, leave it out or make the details up.
Think of AI as a fast, eager assistant who has read a lot but is sometimes wrong. You tell it what you need, it writes a draft, and you check the draft before you use it.
What It Can Do: Six Real Tasks (9 min)
The instructor shows what these tools can do on work like ours. Watch first; you will try your own requests in the activities.
| Task | What you ask | What to notice |
|---|---|---|
| Analyze data and make a figure (live) | Upload a spreadsheet and ask which speed limits have the highest share of severe pedestrian crashes, with a bar chart, three findings, and any data problems. | It writes and runs the analysis itself, and flags missing and impossible values. You still check the numbers. |
| Review your own draft | “Act as an NIH study section reviewer. List the 5 biggest weaknesses, most serious first. Do not rewrite it.” | It finds design and framing problems before a real reviewer does. It critiques your writing; it does not write for you. |
| Make or update a presentation | “Turn this abstract into a 5-slide outline for a community meeting, with speaker notes.” | A fast first structure. Your expertise decides what stays. |
| Sort an email thread | “Pull every task out of this thread: who, what, and when. List anything nobody agreed to do.” | It catches the item that was raised twice and never assigned. |
| Find sources | “Give me 5 peer-reviewed studies on leading pedestrian intervals, with DOIs.” | In our test, one of two DOIs pointed to an unrelated paper. Open every source before you cite it. |
| Form student groups | Replace names with IDs in Excel first. Upload only IDs plus program, concentration, and coding experience: “Make 6 groups of 4. Each group needs a strong coder, at least two programs, and no more than 2 students from one concentration. Show a table and check each rule.” | Even in ChatGPT Edu, send the least information needed. Match IDs back to names yourself, and count each rule yourself. |
Try the data demo yourself later: download the fabricated crash dataset (1,045 crashes in a fictional city), upload it to ChatGPT Edu, and paste this prompt:
This file has pedestrian crashes in our city. Which speed limits have the highest share of crashes where someone was killed or seriously injured? Make a bar chart, give me 3 plain-language findings, and tell me about any data problems.
Try the groups demo too: download the fabricated class roster (24 students, IDs only, no names), upload it, and paste the groups prompt from the table. Then count each rule yourself: did every group really get a strong coder?
Agentic tools such as Claude, OpenAI Codex, and Copilot agents can open files, run code, and check their own work across many steps. They are not Drexel-supported: use them with public or low-risk information only. Start with ChatGPT Edu.
How AI Gets Things Wrong, and Who Is Responsible (7 min)
It makes things up. Because it predicts likely-sounding text, it can invent facts, numbers, and citations. Check anything that matters before you use it.
It can be biased. It learned from writing that reflects society’s inequities, so its answers can repeat stereotypes. In public health, that matters: our work affects real communities. (For a well-known health-care example, see the Bias & Equity Audit track.)
You are responsible. If your name is on it, you own it: accuracy, fairness, and tone. “The AI wrote it” is never the answer. We return to these rules in Core 4: Verify, Disclose, Decide.
Guided Tour: ChatGPT Edu, Step by Step
Guided Tour: ChatGPT Edu, Step by Step
Goal: Everyone signed in and comfortable with the screen before we try real tasks.
Can’t sign in to ChatGPT Edu? Open Copilot Chat with your Drexel account instead, and say so in the Zoom chat if you need help. Everything today works in either tool.
React with 👍 in Zoom after each step.
Sign in. Go to ChatGPT Edu and sign in with your Drexel account. Check that you are in the Drexel workspace (the name appears near the top or bottom left).
Find the message box. It is at the bottom of the screen. This is where you type your prompt.
Type a test prompt and press Enter:
In two sentences, what can you help a public health staff member with?
Ask a follow-up in the same chat:
Give me three examples.
Notice that it remembered your first question.
Copy the answer. Hover over the response and click the copy icon, then paste it anywhere.
Start a new chat. Click “New chat.” The new chat does not remember the old one.
Find your history. Your past chats are listed on the left. You can rename or delete them there.
Activity 1: Your First Three Requests
Activity 1: Your First Three Requests
Goal: See what the tool does well by giving it everyday tasks.
Copy each prompt below into ChatGPT Edu, one at a time. Read what comes back. There are no wrong answers here.
Request 1 — Summarize
Summarize this in 3 bullet points for a busy colleague:
Between 2018 and 2022, pedestrian deaths in the United States rose 18 percent, the highest level since 1981. Most of these deaths happened in cities, often on wide, fast roads without sidewalks or crosswalks. Black and Hispanic pedestrians and adults over 65 were killed at higher rates. Many cities have Vision Zero plans, but most focus on education campaigns instead of redesigning streets, which the evidence shows works better.
Request 2 — Draft an email
Write a short, friendly email reminding my committee that their report sections are due Friday at noon. Keep it under 100 words.
Request 3 — Explain
Explain what a “data use agreement” is in plain language, in about 100 words, for someone new to public health research.
Then try a follow-up. Pick the answer you liked most and type one of these: “Make it shorter.” “Make it more formal.” “Add a sentence about ___.”
Quick Reflection
- Which answer was most useful as-is?
- What did you need to change or add that the AI couldn’t have known?
Activity 2: Better Requests with PTCF
Activity 2: Better Requests with PTCF
Goal: See how a little more detail gets you a much better answer.
Four Questions to Ask Before You Type
| Letter | Question | Example |
|---|---|---|
| Persona | Who should it act as? | “You are an experienced department administrator.” |
| Task | What do you want it to do? | “Write an email announcing a staff meeting.” |
| Context | What does it need to know? | “The meeting is Tuesday at 10 in Room 204. We will discuss the new travel policy.” |
| Format | What should the answer look like? | “Under 100 words, warm tone, date and time in the first line.” |
You don’t need all four every time. Adding even one or two makes a difference.
Try It
Type this short prompt and look at the answer:
Write an email about a meeting.
Now start a new chat and type the full version:
You are an experienced department administrator. Write an email announcing a staff meeting. The meeting is Tuesday at 10 in Room 204, and we will discuss the new travel policy. Keep it under 100 words, use a warm tone, and put the date and time in the first line.
Compare the two answers. Which one could you send with the fewest changes?
Your turn. Think of a real email or short document you need to write this week. Write a prompt for it using the four questions. Use made-up names and details if needed.
If you run the same prompt twice, you may get slightly different answers. That is normal. When a prompt works well, save it in a document so you can reuse it.
Wrap-Up (5 min)
- AI writes drafts. You check them and decide.
- For Drexel work, use ChatGPT Edu or Copilot and follow the data rule.
- Before you type, ask the four questions: Persona, Task, Context, Format.
Try this before Core 2: pick one task you do every week—an email, a summary, a set of notes—and do it once with ChatGPT Edu using PTCF. Bring it to the next session.
Looking Ahead
Core 2: Workflow Enhancement — we take the one task you tried this week and make it part of your routine: summarizing, rewriting, and making tables in ChatGPT Edu and Copilot Chat, and saving the prompts that work.
Give Feedback on This Workshop
Your feedback directly shapes how these workshops are revised and what gets scheduled next.
In the first question, select the session you attended.