LLM usage log

LLM tools were used in the following way for the tasks below

Brainstorming

  • Coming up with the title of this project
  • To write an outline / structure of project tp guide me
  • to help me derive a schedule for this project
  • brainstorming ways to manipulate temporal data to obey IID assumptions
  • effective data visualization tools
  • deciding and implementing the best evaluation metrics for model comparisons in supervised learning

Writing:

  • Summarizing my messy notes and incorrectly worded interpretations into concise sentences
    • such as my notes on the Tier placement scores– the policy descriptions were long but I needed to cut this down short and concise
  • Mostly grammar check
  • Text summarization for literature review
  • Outline for the literature review (asked what common breakdowns look like)

Code:

  • ChatGPT was consulted for:
    • parsing data using the World Bank API
    • using BeautifulSoup to partse Wikipedia data for the Henley Passport
      • lots of complications and trial and error
      • helped me debug the intricacies in the HTML
    • writing out the ISO codes to represent and map to European countries
    • DEBUGGING in many parts
    • Markdown syntax and shortcuts that I’d forgotten (footnotes and in text links)
    • linking datasets to qmd
    • how to retrieve accidentally deleted files in github
    • implementing the data manipulation to turn temporal data into static
  • ClaudeAI was consulted for:
    • creating the interactive maps in the EDA section using plotly.express
      • implementing the ISO codes for mapping
      • creating color mappings
      • just introducing me to the tool
    • deciding and implementing the best evaluation metrics for model comparisons in supervised learning
    • visualizing the results in supervised learning
    • helping to write the pipeline; I fed Claude my code and it caught mistakes that were causing errors
    • LOTS of debugging help

General:

  • I was confused on some hyperparameters specifics up to this point, but ChatGPT explained clearly so that I could clearly document the information in the descriptions.
  • Cleared up some information