- University of Wisconsin-Madison
- Research Guides
- Generative AI
- Using Generative AI Effectively
Generative AI : Using Generative AI Effectively
Guidance and resources for AI chatbots and other types of Generative AI
Effective Prompting
Ethan Mollick, Wharton School Professor and generative AI advocate offers the following general prompting advice:
- Prompting has a major impact on quality of output.
- Don’t rely on “magic” words or phrases (e.g., offering rewards to the AI for doing something well, telling it to take a deep breath before responding, being polite).
- Becoming really good at prompting is a waste of time.
Mollick offers three strategies that he has found consistently lead to good results:
- Add context. For instance,
- assign a persona (e.g., “you are a brilliant astrophysicist who is experienced at teaching complex theories to the public,”)
- assign an audience (e.g., “you are speaking to a client”)
- assign an output format (e.g., “set of bullet points,” “friendly email”)
- Include “few shot” prompts – provide examples of the kind of output that you want
- Include “chain of thought” prompts – ask the AI to complete instructions step-by-step (“First, outline the results; then produce a draft; then revise the draft; finally, produced a polished output”)
Incorporating AI Chatbots into Your Research Strategies
Although it's not advised to use generative AI directly to find sources on your topic, AI chatbots may be helpful for some parts of your research process. For instance:
- Prompt a chatbot to help brainstorm ideas or to narrow or broaden your topic. (Just be sure to confirm accuracy as you research further.)
- Prompt a chatbot to generate a list of keywords associated with your topic. Use these terms as you search online, in the Library catalog, or in our Library databases.
- Use a chatbot to translate or summarize sources for improved readability.