AI Prompting
For Faculty and Staff at USF
Prof. & Chair Allison Thorson, athorson@usfca.edu
April 7, 2026
How is AI changing the playing field?
Since access to “an answer” is now easier than ever, the smartest people in the room are now those who ask the right question(s).
What is prompt engineering?
Prompt engineering is the practice of designing and refining inputs (prompts) to guide AI models toward generating the most useful and accurate responses. It involves understanding how language models interpret instructions and experimenting with phrasing, context, or structure to optimize outputs. In essence, it’s about learning to “speak the model’s language” so it performs better for a given task.
Common Large Language Models (LLMs) & data
** Founded by former OpenAI employees
Things to consider…
Today’s focus: prompt engineering!
Prompt engineering techniques we’ll focus on today:
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P.,
Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-
Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D.
M., Wu, J., Winter, C., … Amodei, D. (2020). Language models are few-
shot learners. Advances in Neural Information Processing Systems, 33,
1877–1901. https://arxiv.org/abs/2005.14165 (**not entirely attributed to
one set of authors)
This field is relatively new—mainstream adoption began around 2022.
Zero-shot prompt – give the model a prompt with no examples
Zero-shot
Zero-shot
Data centers can be harmful to the environment primarily because they consume enormous amounts of energy and resources. These facilities run thousands of servers continuously, requiring vast amounts of electricity not only to power the machines but also to cool them and prevent overheating. Much of this electricity still comes from fossil fuels, which leads to significant greenhouse gas emissions and contributes to climate change. Additionally, data centers use large quantities of water for cooling systems, which can strain local water supplies, especially in dry regions. The production and disposal of hardware also create environmental issues, as manufacturing servers requires rare minerals and generates pollution, while outdated equipment contributes to electronic waste. Together, these factors make data centers a growing environmental concern in an increasingly digital world.
Zero-shot – how an instructor might use this
Data centers can be harmful to the environment primarily because they consume enormous amounts of energy and resources. These facilities run thousands of servers continuously, requiring vast amounts of electricity not only to power the machines but also to cool them and prevent overheating. Much of this electricity still comes from fossil fuels, which leads to significant greenhouse gas emissions and contributes to climate change. Additionally, data centers use large quantities of water for cooling systems, which can strain local water supplies, especially in dry regions. The production and disposal of hardware also create environmental issues, as manufacturing servers requires rare minerals and generates pollution, while outdated equipment contributes to electronic waste. Together, these factors make data centers a growing environmental concern in an increasingly digital world.
One-shot prompt – give the model a prompt with one example
One-shot in a series of prompts
One-shot in a series of prompts
Rough draft
Few-shot prompt – give a few examples of what you want before you ask your prompt (set the tone with the style you want)�
Few-shot
How a student could use few-shot prompting…
How might we need to change our assignments to adapt to this new technology?
CLEAR framework, Lo (2023)
Act as a college professor who needs to create ADA-compliant image descriptions (alt text) for the visually impaired using the image attached. Keep descriptions short. Focus on the information the image is meant to convey within the context of the page. Describe what is physically present. Describe the image based on why it is there. Include any text in the description. Present output as unformatted text that can be easily copied and pasted.
Reflective: think critically about the AI tool's answer. Does it make sense? Are there perspectives or experiences that are missing? �
More advanced techniques – building your own tools:
“Act as an internship class instructor for a college course. I want to use CHATGPT to act as a mock interviewer. I want CHATGPT to ask an interview question and have students answer this question by clicking on the microphone in the chat. Based on students' responses (a weak answer or a strong answer) adjust what the next follow-up question will be as if you were having a conversation with them. “
“One thing I would like you to add to this assignment is - after asking 7 questions, provide feedback to the student on the strengths of their answers, weaknesses, and suggestions on how to improve. A strong answer will use the STAR method (a structured technique for answering behavioral interview questions that involves the answer to ‘Tell me about a time when...’ detailing a specific Situation, Task, Action, and Result)”
Output: (Instructions for students: Copy and paste this prompt into an LLM and respond verbally.)
Some things to notice:
It’s all about communicating effectively with the LLM.�
Prompt engineering techniques from today:
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P.,
Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-
Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D.
M., Wu, J., Winter, C., … Amodei, D. (2020). Language models are few-
shot learners. Advances in Neural Information Processing Systems, 33,
1877–1901. https://arxiv.org/abs/2005.14165 (**not entirely attributed to
one set of authors)
Questions?
AI Prompting
For Faculty and Staff at USF
Prof. & Chair Allison Thorson, athorson@usfca.edu
April 7, 2026