Skip to main content
AI at Oxford

Streamlining coding and editing workflows with ChatGPT Edu

User Case Study

DPhil student, Department of Physiology, Anatomy and Genetics

I’ve found ChatGPT Edu is a useful tool for assisting with editing both R code and text. For R code writing, I used to edit my code manually and amend it based on advice I could find in online resources, such as Stack Overflow forums. However, this advice was usually not directly related to my application and it would require some rewriting to work with my existing variables.

For editing text, I used to do this manually by reading my paragraphs sentence by sentence and replacing jargon words with layman terms (if I was writing text for a more generic audience) or summarising text into shorter more concise sentences (if I was writing text for a given word count). I would usually just search in a thesaurus for different suggestions for phrasing.

Now, using ChatGPT, I can provide code that I’ve written and ask in the prompt for specific modifications for a given output. For example, I might ask it to change code to modify ggplot   box width, or to rewrite code so that plots are generated in a loop rather than individually. It can then suggest changes that are directly relevant to the code I have already written and the variables I already use.

And now, using ChatGPT, I can provide the draft text that I have written and simply prompt AI to "change this text into layman language and in 150 words" and it can give some suggestions of how to rewrite some phrases more concisely or clearly while retaining the original message. I then work back and forth, tweaking some sections based on the suggestions I like best.

Both applications have accelerated my processes, helped suggest words I might not have thought of using a thesaurus or dictionary alone, and offered clearer ways of arranging text or scripting code. I find that ChatGPT is most useful for modifying or enhancing material I’ve already written, as there’s  already sufficient detail and original contribution in the starting prompt. This allows me to feel confident taking credit for the output, which only includes small additional contributions from the tool.