CV Google Scholar MSU

AI and Research Practice

A practical guide to using large language models in research, for researchers in management and adjacent fields. It starts from what these models are actually doing, and builds from there to how to get output you can trust.

Why this site

Researchers are already using these tools. Most of the conversation about that is about permission and consequences. Am I allowed to use AI? Journals are answering that in their AI policies. What will AI do to the field? Editors are debating that in editorials.

Far less of it is about the question most of us face at the keyboard. How do I get good output, output I’d stand behind in my research? That’s what this site is about.

Where it starts

Getting good output starts with knowing what these models are doing. Two principles carry most of the guide.

PRINCIPLE 1LLMs are predictive, which means all they compute is a probability for every word that could come next.

PRINCIPLE 2The output is stochastic, which means what you get is one draw from those probabilities.

THE IMPLICATIONLLMs are optimized for fluency, so they sound just as certain when they’re wrong.

Upcoming lab sessions

The labs are on Zoom, open to anyone, and recorded. Sign up for the labs

What’s here

Last updated September 2026. To cite: Harmon, D. (2026). AI and Research Practice. https://derekharmon.com/ai