Your AI Needs a Logic Model: Prompt Engineering for Smarter Outputs

This presentation explores how principles of logic modeling can inform effective prompt engineering (i.e., crafting specific instructions for generative AI tools to maximize output relevance) for AI-assisted evaluation tasks. We posit that (1) treating generative AI tools as a student needing feedback and instruction, instead of an expert that knows the best answer, helps us identify the utility and limitations of these tools, and (2) we can leverage evaluation methodologies to enhance the use of generative AI. By systematically aligning prompt structures with elements of logic models (e.g., context, inputs, activities, and outputs), evaluators can leverage their expertise to promote methodological clarity and ethical intention into AI-assisted tasks. Our exploration of this approach, connecting evaluative thinking with prompt development, will introduce attendees to practical strategies for designing prompts that support evaluative tasks, helping them responsibly integrate generative AI into participatory evaluation practice.

Margaret Jaeger – The Rucks Group, LLC; Christopher Cox – The Rucks Group, LLC