Skip to main content Skip to navigation

Generative AI for Data Science 101: Coding Without Learning To Code

Project Overview

The document explores the use of GitHub Copilot, a generative AI tool, in a Data Science for Business course at the University of Southern California, emphasizing a unique pedagogical approach that allows students to engage with data science concepts without the necessity of mastering coding syntax. Instead of learning R code directly, students articulate their ideas through English prompts, which Copilot then translates into functional code. This innovative method aims to enhance student engagement and foster curiosity about data while addressing the difficulties associated with coding education. The findings indicate significant benefits, including increased student interaction and lowered barriers to experimentation, though challenges persist, such as the variability in AI-generated outputs and the importance of crafting precise prompts. Overall, the implementation of generative AI in this educational context illustrates a promising shift towards making data science more accessible and interactive for learners.

Key Applications

GitHub Copilot as an English-to-code translator for R

Context: Data Science for Business course in an MBA program for non-major students

Implementation: Students were taught to write prompts in English that Copilot would convert into R code, starting with basic commands and progressively tackling more complex tasks.

Outcomes: Students engaged more deeply with data and developed a broader perspective on data analysis tasks, leading to innovative solutions without prior coding experience.

Challenges: Inconsistent outputs from Copilot based on prompts, students' frustration with variable responses, and the lack of transparency in how Copilot functions.

Implementation Barriers

Technical Challenge

Variability in AI outputs based on the specificity of prompts, leading to inconsistent experiences for students. Students also struggle with understanding how to effectively interact with AI tools.

Proposed Solutions: Encouraging students to write specific prompts and emphasizing the importance of the context window in generating desired outputs. Developing guiding principles for writing effective prompts.

Learning Curve

Students struggled with understanding Copilot's capabilities and the distinction between writing prompts and coding.

Proposed Solutions: Possibly introducing coding concepts before using Copilot to enhance understanding.

Project Team

Jacob Bien

Researcher

Gourab Mukherjee

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Jacob Bien, Gourab Mukherjee

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies