This section offers some practical tips on generating effective prompts. Readers may refer to the references for more suggestions.
Provide GenAI with the main topic for your interactions. Or tell GenAI the core concepts you want to practice through interactions.
The goals of your interactions. Or it can be said, what do you want to obtain after the interactions?
For example, "The goal is to practice my negotiation ability through the interactions in the created scenarios."
Tell GenAI to act as a scenario creator and simulate detailed scenarios according to your requirements and instructions.
The role you want GenAI to play in the interaction.
The role you'd like to play in the interaction.
Break down complex tasks into subtasks and structured steps. Give GenAI more specific step-by-step instructions instead of general ones.
e.g. "Translate the text into Chinese." -->
1. "First translate literally to preserve meaning."
2. "Refine the translation to align with Chinese linguistic conventions."
Provide GenAI with more details which are relevant to your interaction topic.
e.g 1. You want to ask GenAI to write a cover letter for you. Then you'd better also provide relevant detailed job description and personal CV;
e.g 2. You want GenAI to help revise the text for you, then you need to add more details on the particular focus, such as writing style, word accuracy, content fluency, etc.
"Revise the text." -->
"Act as a top editor for top journals to improve the clarity and flow of the context."
GenAI is good at learning from examples (few-shot learning), so don't just tell.
e.g. "Create a chart for this data." -->
(upload a paper) "Create a bar chart for this data, similar to the one in figure 3 of the attached paper."
You can specify your desired output style, tone, format, such as bullet points, reading level, etc.
e.g. "Summarize the key findings." -->
"Summarize the key findings in bullet points and use language a high-school student would understand."
It aims to prevent GenAI from acting in unexpected ways.
e.g. You want to simulate interactions with LLMs by allocating different roles for yourself and LLMs. To prevent LLMs from acting your role, you could put "Do not play my role. Only play the xxx roles." in your prompt.
Take a look at the response from GenAI and assess whether they are in line with the expectations. If yes, the interaction can come to an end. If not, adjust your inputs.
After evaluation, according to your needs, you can make adjustment to your inputs.
Tips: 1. chain-of-thought prompting: Your could add instructions like "let's think step by step" in the prompt to get more improved response.
2. emotional prompting: Adding emotional expressions like "Take a deep breath--", "I will tip $200 for great responses" to get responses of higher quality.
Try again and again!
"Effective prompting is not formulaic...treat prompts as testable hypotheses, and use results to inform iterations. Not all attempts will succeed, but evidence accrues with each. With tenacity, optimal results will emerge." (Lin, 2024)
An Example for Writing Effective Prompt
References:
Lin, Z. (2024). How to write effective prompts for large language models. Nature Human Behaviour, 8, 611-615. https://doi.org/10.1038/s41562-024-01847-2
Lin, Z. (2024). Prompt engineering for applied linguistics: Elements, examples, techniques, and strategies. English Language Teaching, 17(9), 14-25. https://doi.org/10.5539/elt.v17n9p14
Lu, W. (2024). Training the stochastic parrot: Using generative AI to create textual materials for communication courses. Communication Teacher, 38(4), 315-322. https://doi.org/10.1080/17404622.2024.2385343
Mollick, E., & Mollick, L. (2023). Assigning AI: Seven Approaches for Students, with Prompts. https://doi.org/10.48550/arxiv.2306.10052