The use of AI in Software Engineering

16 Dec 2024

---

I. Introduction

The role of AI in education has very mixed reviews in the sense that many instructors advise not to use any sort of AI when doing their coursework. However, with software engineering, in ICS 314 specifically, we were encouraged to take advantage of many forms of AI to aid our learning process. With this encouragement, I leaned more into using ChatGPT and looked into the benefits of Co-Pilot. The use of AI was very fundamental in my learning experience in software engineering due to its easy access allowing me to ask for help whenever I needed it.

II. Personal Experience with AI:

My personal experience with AI ranged throughout the course with the following assignments:

• Experience WODs: For the experience WODs, I tended to use ChatGPT to get a rough outline of the code and fixed it up to get the desired results. At the beginning of the course when we had assignments that could be done on TypeScript Playground, AI was very convenient in easily pinpointing the tiniest error that may have been causing the results to not come out correctly. For WODs like the “E35: Island Snow with Bootstrap 5”, where it heavily relied on styling, I would send ChatGPT a prompt saying something like “My navbar spacing is a bit off, how do I fix it? Here’s my current .css file: . And nine times out of ten, they were able to figure out what was off and provide a solution.
• In-class Practice WODs: The in-class practice WODs were usually a time for me to try to do the assignment with very minimal use of AI. However, the closer it came to the end of the WOD, I resorted to ChatGPT to help fill in the blanks that were forming in my head on how to finish the assignment to meet the criteria. For the UI design WODs, there would be a problem with the button not properly redirecting to its proper route so I would send the prompt “The ‘Go Back’ button is not working. How do I fix it?” and ChatGPT would send me a list of things that I should double check to ensure that everything is set up correctly.
• In-class WODs: The in-class WODs were a definite experience for the use of AI because of the fact that we were being timed and would get a 0/100 if we didn’t finish in time. I copy and pasted the instructions of the WOD into ChatGPT and prompted it to “Create an outline”. After creating the outline and working on it on my own, If I was still getting errors I would type the prompt “ Error: ” and ChatGPT would assess the problem.
• Essays: Essays were one of the few courseworks that I didn’t use AI for. I thought that essays were very self-explanatory and most of the time reflecting on our learning experiences so I didn’t feel the need to ask ChatGPT or any other AI for anything.
• Final project: The final project was definitely difficult, even with the use of AI. I used AI to try to figure out some major issues within the code (such as syntax errors, routing errors, etc.) but even when prompted “ What needs to be changed or added in order to fix ?” ChatGPT tended to give a more generic answer because it doesn’t know the whole program so it could only answer based on the information given, which doesn’t amount to much. More basic syntax errors were easily fixable with the use of ChatGPT.
• Learning a concept / tutorial: When learning a concept, I found that watching videos (either given to us or searching up online) were better for learning than it would be to ask an AI. While an AI could’ve given the concept in simple terms to understand, I didn’t see the importance of asking questions when it was probably answered in a video or could be asked in class.
• Answering a question in class or in Discord: When given a question in class and Professor Port would say “you can ask your favorite AI”, most of the time I would and it showed us that sometimes AI gives really good answers and sometimes the answers are a bit off or really off. This was a good learning experience.
• Asking or answering a smart-question: I’ve never used AI to form a question or a response to a smart-question. I think this ties in together with the “learning a concept” question because when we learned about smart-questions in the beginning of the semester, we were given a layout and guideline for them and after learning about them, I was able to pick up how to form them on my own so I just never felt the need to ask an AI after that experience.
• Coding example: This was very useful in software engineering because its similar to asking for an outline but you could see how everything should look after you properly format everything. For example, I have asked ChatGPT “Show me an example of a fully functioning navbar” and it would give me an example then I would compare it to the navbar that I currently had, making it easy for me to understand what was wrong with my own code.
• Explaining code: When explaining code, it was very useful to use AI because sometimes it would give you a piece of code as a solution that you never would have thought of on your own. By using the quoting feature on ChatGPT, I often highlighted the section I wasn’t understanding and typed a simple “Why do I do this?” and it would give me an explanation which would lead me to understand it better.
• Writing code: If the code was a bit more complex and it was taking me an unreasonable amount of time, I would ask ChatGPT to “ Provide a completed .ts file.”, copy and paste it into VSCode, and see if it ran. If it did, I would read over the code generated and figure out what I was having trouble with.
• Documenting code: When it came to documenting code, AI created very apparent AI-generated comments while providing code, so it was easy to tell when something wasn’t written by a human. For example, the top of the file would just read “// .tsx file” and that would be it for the comments. I found myself having to adjust these comments most of the time.
• Quality assurance: Quality assurance was a big factor that I used AI for. Specifically when getting ESLint errors, I would prompt ChatGPT with “ ESLint errors: Regenerate this code to resolve the ESLint errors.” and it would regenerate the code without changing anything important, solving the errors and making the code cleaner.
• Other uses in ICS 314 not listed: Other than the uses that I just went over, I didn’t specifically use AI for anything else other than as an aid to assignments and questions given in class.

  III. Impact on Learning and Understanding:

  The incorporation of AI has influenced my learning experience in ICS 314 by allowing me to have a more visual understanding of what I’m learning. When getting stuck on an issue and not knowing how to solve it, AI has been able to answer my prompts, give me step-by-step solutions, and provide a solid explanation. An example of this would be with all the new programs like Vercel and Prisma that were added to the course during this semester. AI was able to tell me what commands I needed to input into my terminal in order to properly install, create, or run something. This increased my comprehension, skill development, and problem-solving abilities all while challenging me to try and figure out a solution even if most of it was laid out for me through AI. I can confidently say that AI technologies have enhanced my understanding of software engineering concepts.

  IV. Practical Applications:

  Outside of ICS 314, I have used AI to create a list of ideas based on a prompt I have given it. For example, on a final paper where we were allowed to pick any topic we wanted, I asked ChatGPT to “Give me a list of ideas relating to ” in order to narrow down what I could write about. The effectiveness of AI applications in addressing real-world software engineering challenges is proven to have a positive impact, although often given a negative connotation when being talked about.

  V. Challenges and Opportunities:

  A challenge that I encountered with the use of AI within ICS 314 is the limitations on what it could do in terms of fixing a code with a given prompt without the other components getting affected from the updated “fixed” code. For example, for one of the Aloha Beer WODs, I asked ChatGPT to “fix the icons on the navbar” and although their code was technically correct, when inputted into VSCode and ran, the formatting was all off and effected the whole thing making it look worse than it did before the issue was “solved”. Potential opportunities for further integration of AI in software engineering education could possibly be teaching proper ways to use AI in a way that would be the most effective. Considering that most of the time people just copy and paste instructions into an AI application and hope for the best, it may be of best interest to find ways to ask them questions that would give the best possible solution.

  VI. Comparative Analysis:

  Traditional teaching methods definitely have its pros and cons compared to AI-enhanced approaches in software engineering. One pro is that instructors would know that the class is retaining knowledge based on if they’re able to solve an issue themselves, rather than wondering if they just copied and pasted a solution from an AI application. However, in the context of software engineering, engagement would and practical skill development would be easier to come by due to the fact that it allows students to feel more secure in what they’re learning and provides a fallback or aid to them if they feel like they need one. In traditional teaching methods, if a student doesn’t understand the content, they’re essentially on their own and risk failing more of the assignments than if they were given the option to ask an AI application for help.

  VII. Future Considerations:

  I do believe that AI is the future and its development will be considered an essential part of software engineering itself. However, when it comes to software engineering education, I think that there should be a structure in how AI is taught or should there be some kind of guideline that would prevent students from shortcutting their way through their assignments rather than trying to understand the content for themselves. I like the idea of including if we used AI within our submissions when we turn in something like a WOD or an assignment. This helps the instructors to be able to identify whether AI is helping or hurting students and their progress towards getting a better understanding of software engineering.

  VIII. Conclusion:

  In conclusion, the use of AI in ICS 314 was a very fundamental aspect to both me and the rest of my classmates. Although some definitely relied more on it than others, we were able to use it effectively to get us to where we are now. I am very appreciative of the opportunity to be able to use AI and not be bashed for it but rather encouraged to do so. Apart from learning about software engineering, I felt like we were able to learn about how to use AI in an effective way and allowed us to open our eyes to things we may have never been able to understand without it. Optimizing the integration of AI in future courses should depend on what the development of AI has come to. If AI progresses further than it has now, the guidelines for using them should change along with it. Everyone should also be taught to be aware of the dangers that come with relying solely on AI and its pros and cons, allowing people to decide on their own if and how they want to use it.