Prompt Programming: A Platform for Dialogue-based Computational Problem Solving with Generative AI Models
Project Overview
The document explores the use of a platform named Prompt Programming that aims to enhance students' interactions with generative AI models, particularly for programming tasks. It highlights the significance of equipping students with the skills to formulate effective natural language prompts, which are essential for generating and assessing code. The platform promotes dialogue-based interactions and accommodates complex, multi-function problems, thereby encouraging iterative problem-solving practices among learners. Analysis of student engagement indicates high levels of participation and success, revealing that students are not only actively involved but also demonstrate critical thinking skills when evaluating the AI-generated code. Overall, the findings suggest that generative AI can significantly enhance educational experiences by fostering essential programming skills and promoting deeper cognitive engagement in students.
Key Applications
Prompt Programming platform
Context: Introductory programming course for university students
Implementation: The platform was deployed in a large introductory C programming course, enabling students to interact with AI through multi-turn dialogues and on-request code execution.
Outcomes: Students showed high engagement and success rates in solving programming tasks, with the platform fostering critical thinking and problem-solving skills.
Challenges: Challenges included students' reliance on AI-generated code and the need for careful evaluation of the AI's output.
Implementation Barriers
Educational Barrier
Students may overly rely on generative AI tools, potentially undermining their coding skills.
Proposed Solutions: Teaching strategies should evolve to prioritize a hybrid approach combining natural language prompting and traditional coding skills.
Technical Barrier
The testing framework for multi-function problems currently limits modular testing and independent function verification.
Proposed Solutions: Improve the testing framework to check each component independently rather than collectively.
Project Team
Victor-Alexandru Pădurean
Researcher
Paul Denny
Researcher
Alkis Gotovos
Researcher
Adish Singla
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Victor-Alexandru Pădurean, Paul Denny, Alkis Gotovos, Adish Singla
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