What is Prompt Engineering?
Generative AI tools (like ChatGPT and others) can act as powerful assistants in education and research – but how you prompt them matters. Prompt engineering is the art and science of crafting the instructions (prompts) that guide AI systems to produce the output you want. In plain terms, it's about learning to "talk" to AI in a way that gets useful, relevant responses.
The RICCE Framework
One helpful method for building effective prompts is the RICCE framework, introduced in USC's AI Fluency course. RICCE stands for Role, Input, Context, Constraints, Examples. Think of it as a checklist to ensure your prompt covers all the key elements for success.
Effective Prompts
This framework (taught in the Garnet AI Fluency 101 course) acts as a checklist to ensure you give the AI all the guidance it needs for a great response. Each component of RICCE is defined below:
Role:
- Specify who or what the AI should emulate. Giving the AI a clear role focuses its tone, expertise, and perspective.
- For example, “You are a curriculum specialist…” or “Act as a research librarian…”.
- This helps the AI adopt the appropriate voice and knowledge level for the task (e.g. a friendly tutor vs. a strict proofreader).
Input:
- Clearly state what you want the AI to do, and provide any data or query it will work with.
- This is essentially the task or question at hand – for instance, “Summarize the following article” or “Draft an email about X.”
- Include any specific input text if needed (like an excerpt to rewrite).
- A well-defined request prevents the AI from guessing your intent.
Context:
- Give background and details to situate the task.
- Context might include the purpose, audience, subject matter, or any situational details the AI should know.
- For example, mention the course name and level for an essay prompt, or the workplace setting for a policy draft.
- The AI uses context to tailor its answer – without it, responses can be too generic.
Constraints:
- Set any limits or guidelines for the output.
- Constraints keep the AI on track.
- This can be desired length (e.g. “in 200 words”), format or style requirements (“use bullet points and an informal tone”), or content to avoid (“no references to personal data”).
- Constraints act as guardrails so the response meets your needs (e.g. focusing on specific topics or maintaining formality).
Examples:
- Provide examples or reference outputs if possible.
- An example of a “good” result (or a specific format) shows the AI what you expect.
- For instance, include a sample answer, template, or style snippet: “Example: [provide a brief excerpt or format].”
- Examples are optional but powerful – they significantly improve relevance and quality of the AI’s output by giving it a target to mimic.
- Even if you don’t have a ready example, you can describe an ideal answer’s characteristics.
When you put it all together, a RICCE-based prompt might read like a detailed set
of instructions rather than a single question. The effort upfront is worth it: structured
prompts yield more tailored, high-quality answers on the first try. As you use RICCE,
keep these tips in mind: be specific, imagine explaining the task to a human assistant,
and don’t hesitate to include context or examples that clarify your expectations
Academic Integrity Reminder
This prompt library is meant to enhance learning and productivity – not to do your work for you. Always use AI ethically and in line with your instructor's guidelines. Do not use AI for unauthorized assistance on exams or assignments.
Use these prompts to brainstorm, study, and save time on routine tasks, but ensure the final work is your own. If you do incorporate AI-generated content, follow your course's rules on disclosure and citation. Using AI dishonestly is a violation of the Honor Code and undermines your learning.
Prompt Library
Gain access to our prompt libary by joining our Garnet AI Community in the Garnet AI Fluency 101 Blackboard course. This library and course are free to facutly, staff and students. Join here.