When working with Generative AI tools like ChatGPT, we type questions or requests in a chat box to communicate with the tools enabled and powered by the Large Language Models (LLMs). These instructions, known as prompts, can range from simple questions to long texts with carefully designed structure. Clear and effective prompts are critical to ensure accurate and high-quality responses from those LLMs. Crafting prompts is an empirical science that requires an iterative process of designing and refining them for generative AI tools to harness their full potential. Various strategies, techniques, and practices can be used to optimize the performance of LLMs, and this field has quickly evolved into a relatively new field: Prompt Engineering.
A carefully crafted prompt can help you unlock a wealth of information to support your research projects, however without a well-formulated problem most prompts will fall short in producing the right outputs.
What is problem-formulation?
Problem formulation is the process of identifying, analyzing and defining a problem (Acar, 2023). Before you begin searching for sources or using generative AI tools be sure to identify the scope and underlying elements of the research problem you are trying to solve. Effective problem formulation involves breaking down complex issues into manageable components, understanding the context and constraints, and reframing the problem to explore it from different perspectives.
Writing effective prompts can feel overwhelming at first with tons of guidelines and techniques that seem specific to different contexts and tools. Though there is no magical one-size-fits-all formula, certain fundamental components can be applied and adapted to most use cases.
This is the heart of a prompt and the only required block. Be clear and specific about what you need the AI to achieve and the scope of information it needs to search for. Choose strong action verbs for clarity: provide, summarize, suggest....
Provide essential background information so the AI tool can understand the goal of the task better. It can also include relevant data or a specific role you want the AI to play.
Add detailed requirements or constraints to shape the task completion. It includes specifying output type and format, what to include and exclude, target audience, writing styles, etc.
Carefully review the initial output and refine your prompt as needed. This may require:
Ready to write powerful and effective prompts? Start by incorporating these components and applying them strategically in your prompt engineering.
Inspired and adapted from:
Tamsin, S. (2023, January 14). The Art of Writing ChatGPT Prompts for Any Use Case. https://sarahtamsin.com/the-art-of-writing-chatgpt-prompts
Ramlochan, S. (2023, April 5). Master Prompt Engineering: Demystifying Prompting Through a Structured Approach. Prompt Engineering Institute. https://promptengineering.org/master-prompt-engineering-demystifying-prompting-through-a-structured-approach