TLDR: An exploratory study with 20 novice programmers found that AI coding assistants boost efficiency and confidence but can lead to overreliance and hinder independent conceptual understanding. Students valued the two-part exam format (with and without AI) for highlighting AI’s benefits and limitations, emphasizing the need for balanced integration that supports learning without replacing foundational programming skills.
The rapid rise of artificial intelligence (AI) tools, particularly AI coding assistants like ChatGPT and Perplexity, has sparked considerable discussion about their role in various fields, including education. A recent exploratory study delves into how these AI assistants influence the experiences of novice programmers, offering valuable insights for educators and students alike.
Conducted by Sergio Rojas-Galeano from Universidad Distrital Francisco José de Caldas, the research titled “Tool or Trouble? Exploring Student Attitudes Toward AI Coding Assistants” investigated how 20 undergraduate students in an introductory Object-Oriented Programming (OOP) course perceived and interacted with AI during a two-part exam.
The Experiment Setup
The study employed a unique experimental procedure. In the first part of the exam, students were tasked with developing a Java application from scratch and were explicitly allowed to use any AI-powered code assistant. This session lasted 90 minutes. A week later, in the second part, students had to extend their original application but this time without any AI assistance. This session was shorter, lasting 30 minutes, and was conducted offline to ensure no AI use. Following these practical sessions, students completed an anonymous survey to share their experiences and perceptions.
Student Perceptions: Benefits and Challenges of AI
The findings revealed a mixed but insightful picture of student attitudes. Most students identified as basic-level programmers and frequently used AI for debugging, code explanation, and code generation. They largely perceived AI as a tool that significantly increased their problem-solving speed and efficiency during the first part of the exam. Many also felt more confident with AI support, attributing it to the ease of problem-solving and clarity it provided.
However, the study also highlighted significant challenges. When AI assistance was removed in the second part of the exam, many students struggled to apply programming concepts independently. This suggested a potential overreliance on AI and gaps in their fundamental understanding. Confidence levels notably dropped in the unaided session, with students expressing uncertainty, particularly with debugging and formulating logic on their own. The research found a weak correlation between frequent AI use and improved conceptual comprehension, indicating that simply using AI doesn’t automatically deepen understanding.
Students also reported difficulties with specific syntax and language-specific knowledge (like Java) when working without AI. Some realized that while AI-generated code was functional, they didn’t always fully grasp its underlying logic or unfamiliar constructs, sometimes prioritizing a working solution over complete comprehension.
Reflections on the Exam Format and Future of AI in Education
Despite the challenges, students generally valued the two-part exam format. Many found it well-structured and educational, helping them to recognize both the advantages and limitations of AI assistance firsthand. This experience underscored the importance of using AI tools in a way that supports, rather than replaces, their own reasoning and hands-on practice.
Looking ahead, students expressed a strong belief that AI is a useful tool for learning programming. They envision AI primarily as a support mechanism for resolving doubts, debugging code, and accelerating tasks, rather than a substitute for core coding skills. Many emphasized the continued importance of understanding logic and syntax, even when adapting AI-generated output. They suggested integrating AI into courses as a support tool for explanations and debugging, but also stressed the value of traditional teaching and building strong conceptual foundations independently.
Also Read:
- Guiding Self-Taught Programmers: A Storytelling Approach to Learning
- AI Literacy Shapes Student Experience with AI-Powered Assessments, Study Finds
Conclusion: A Balanced Approach to AI Integration
This study, though small-scale, provides crucial preliminary insights into the complex relationship between AI coding assistants and programming education. It highlights that while AI can boost efficiency and confidence, educators must design activities that encourage students to internalize concepts and apply them independently. The goal should be to leverage AI for scaffolding and support, fostering active engagement and critical thinking, rather than allowing it to supplant the essential learning process. As AI tools become more prevalent, a balanced pedagogical approach that blends AI assistance with opportunities for independent problem-solving will be key to nurturing skilled and adaptable programmers.
