Learner Profiles
This lesson is for researchers who are exploring AI-assisted coding workflows, whether they have been using AI tools informally for a while or are just starting out. The profiles below describe the kinds of people we had in mind when writing it.
Amara Osei
Ecology postdoc, data cleaning, field survey analysis
Amara works at a research university studying wetland plant communities. She uses ChatGPT regularly to help draft Python cleaning scripts, but her current workflow is: generate code in a browser tab, paste it into a file, run it, and debug manually when something goes wrong. It has mostly worked, but last semester a script silently dropped rows with missing coordinates and she did not catch it until her advisor asked why the sample sizes looked off. She wants a tighter loop between generating code and verifying it, and some way to document what the AI actually did so she can describe it in her methods section.
Diego Vargas
Public health PhD student, Python basics, cautious first-timer
Diego picked up basic Python in a Data Carpentry workshop two years ago and uses it for simple data wrangling. He keeps hearing colleagues mention AI coding tools but has not tried any. He is not sure whether to treat it as a shortcut that will make him sloppy or a legitimate part of a research workflow. He wants to see what AI-assisted coding actually looks like in practice before deciding whether to adopt it, and he wants to know what questions he should be asking before trusting the output.
Priya Nair
Research data analyst, library data services, validation-focused
Priya supports researchers across her institution with data management and analysis. She tried GitHub Copilot for a few months and found it useful for repetitive tasks but unreliable on anything that required understanding her specific data. She stopped using it after it generated a filtering script that ran without errors but used the wrong column name, producing results that took two days to track down. She is skeptical of any workflow that does not include a clear verification step, and she wants to know how to build one around AI tools rather than just hoping they get it right.