With concerns that AI was undermining her students’ critical thinking, Roselyn Baronia didn’t ban it — she redesigned her assessment around it. In the second installment of this three-part interview, Baronia discusses the oral defense she created for a coding lab to produce a “virtual pet” and the rationale behind it. In part one, Baronia described her background and the problems she encountered with students that led her to realize something had to change in her teaching of AP Computer Science A (CSA). Part three will share some of Barnonia’s reflections on what she’s learned as she’s made these changes and on the implications for her future teaching. Baronia has been teaching at the Tsinglan School, a private international school in Dongguan, China, since 2020. Baronia’s also the Head of the Computer Science department, Technology Coordinator, and chair of the school’s AI committee. This interview has been edited for length and clarity and formatted by Adelaida Kim.
As a resource for teachers interested in adapting an oral defense into their work, a public website provides access to some of Baronia’s materials from the virtual pet project, including a timeline, a lesson overview, student examples, and a rubric.part
For previous posts related to AI, see: Could concerns about the academic pressure on students in China lead to real changes in conventional schooling? Stability & change in the education system in China (Part 4); Scanning the global headlines for recent news on AI, schools, and education: Rapid growth in uses, users, and concerns.
++++++++++++++
Thomas Hatch (TH): You’ve already talked about how, as you walked around your classroom, you could see that students were relying too much on AI when they were coding. When did you realize you had a problem? Can you pick up the story from there?
Roselyn Baronia (RB): This was in November and December of 2025. I had already done the first two units, which are the foundation units of AP Computer Science A (CSA). In Unit 3, students are asked to apply what they learned in Units 1 and 2 in a project-based learning activity. In that activity, I’m not teaching in the traditional sense. The students will be building something, and I will guide them, but I won’t be delivering lessons. That was the time I realized my students were heavily depending on AI, because there were so many times I could see it. I would know because the students would try to hide it. They would try to close a window on their computer, or they’d look surprised when I walked by. I would just tell them, “Why does it feel like you’re cheating? Why do you think you’re cheating?” I’d ask them, “Why don’t you want to admit to me that you’re using AI?” They felt ashamed about using AI. That was the turning point. I was like, this has to change. Somebody has to make these kids understand that they are losing a very, very important aspect of learning if they totally depend on AI.
“Somebody has to make these kids understand that they are losing a very, very important aspect of learning if they totally depend on AI.”
I told them, AI is a tool that should help us, but not something we can solely depend on. There are some kids that really need AI. For instance, I have a student who is struggling with English, and some teachers are concerned he might not graduate. But in my class, this student is really doing his best. It’s difficult for him because it’s in English and his English level is low — but that’s why he needed AI. For that particular instance, I think that’s fine. But for other students, AI is just an easy way to get the answers, and this is not right — that’s why I needed to change the assessment.
TH: What steps did you take once you realized you needed to do something?
RB: I had a conversation with my students. I told them about recent MIT research in an article called “Your brain on ChatGPT” that suggested that AI can be helpful. However, if you just use it to do the work for you, it can undermine your learning. I also shared some statistics from McKinsey, the World Economic Forum, and related research from MIT about the problems with relying too much on AI. I told my students: you are taking computer science, and you are at the front line of the technology used in the classroom, so you have to do it right and influence your peers. That’s why I need you to really think about why there’s a need for using AI. Surprisingly, they acknowledged the problem. They said they rely too much on AI, and I was happy that they understood there is an impact from using AI — it’s a double-edged sword.
“They said they rely too much on AI, and I was happy that they understood there is an impact from using AI — it’s a double-edged sword.“
After that, I asked them: How do you think we should modify our assessments to make sure you are not relying too much on AI and are actually learning? I tried to involve them in the process, because I have teenagers myself, and I know this is a sensitive matter. That’s why I realized maybe I should involve them — ask them how they feel about changing the assessment, how they feel about giving lower weights to final products and instead focusing on the process. Then, when I suggested doing an oral defense, they said, ‘Yes, we’re fine, we’re excited, we want to do the oral defense.’
Another reason they agreed to the defense was because, as they were doing their first lab, some of them were already using AI, and when they took the unit test and progress check afterward, there was a gap between what they were able to show in their final product and their performance on the free-response questions. That means some of them really relied on AI — maybe the students couldn’t admit it to me, but they admitted it to themselves — because the test results didn’t match the quality of the product they submitted.
“The test results didn’t match the quality of the product they submitted. The students couldn’t admit it to me, but they admitted it to themselves.”
The Oral Defense
TH: So you realized you had to do something last November while they were doing one of the project-based learning in Unit 3, and then you had the discussion with them in December, in preparation for their second Lab. After you had the discussion and got them on board, what did you do?
RB: When they said it was fine to change our assessments, I presented them with a plan for an oral defense to come at the end of their 2nd lab, and they were okay with it. They told me they were excited and looking forward to the oral defense. We brainstormed how they would be assessed and who would be part of a panel of judges who would assess them. Then they continued working on their projects. For this lab, all the students chose to create, and they worked on it between December and February. There actually wasn’t that much class time, because in January we had our final exam for the first semester, so there was an interruption; and then one week after their exam, we stopped again because of the Chinese New Year holiday — that was a three-week holiday. So we held the oral defense about two or three weeks after the Chinese New Year.
TH: Can you describe for us how the oral defense went?
RB: The panel was composed of two adults — I was one of them, and the second was one of the other computer science teachers. I also invited the head of curriculum, teaching and assessment, and the division directors, because in my mind I also wanted to offer this learning opportunity that might help change assessments in other subjects. Unfortunately, the administrators were not able to come because of their busy schedules.
TH: Just so I’m clear: you and one other computer science teacher were on the panel. Was there anyone else?
RB: There were two students. One suggestion from the students was to allow everyone in the audience to ask questions, but that would take too long, but instead I was able to include two students on every panel.
TH: Did every student have a chance to be a panelist?
RB: Yes. Because I wanted them to also experience what it’s like to ask questions.
TH: Did you do any preparation for them in terms of being a panelist, or did you just assign them and let it go?
RB: That’s a very good question. I shared with them some of my experiences as a panelist, but I did not have a formal session to practice asking questions — and that’s something I have to reflect on. I experienced one or two students who, when they were supposed to ask questions, either didn’t want to ask questions or asked questions that were more connected to emotions and not very technical, like, “Did you enjoy it?” or “Did you have fun?”
TH: Can you describe for me your experience with the oral defense? How long did it take to do all of them? How did it work?
RB: The students worked in groups to create their virtual pets, and every group had 15 minutes for the defense — it was actually quite short. So basically, each group had 3 minutes for the presentation, and they could do it in any way they wanted.
A lot of students in the first class did a traditional presentation — they demonstrated their app and what it could do. But in the second class, it was more video-based. They created a video, and one of them did something like a podcast, as a way to share what their project was about. After the 3-minute presentation, the panelists started asking questions. Every panelist could ask up to 2 questions, for a total of eight questions. I designed for about 1 to 1 and 1/2 minutes per question, plus the presentation, so it took about 15 to 20 minutes per group. But that meant I could only have two groups do the defense in one 40-minute class block. In my first class, it took me 2 days — or 4 blocks — to finish the oral defense. The second class was smaller so it only took 1 day. Then they did a reflection the following day.
The teachers and students on the panel get a copy of the manuscript before the day of the defense so they can go over their answers and base their questions on what the group has written. It also comes with a link showing a demo video of how the code works. Due to limited time, review of students’ work happens before the defense, so what we focus on during the 20-minute activity is more on students proving their work, justifying their choices, sharing rationale behind, understanding how AI is used or influenced in their choices, and checking if the work is originally done by them with AI assistance.
The Assessment
TH: That’s really helpful. You created a Summative Coding sheet for the Virtual Pet Lab, which has a set of questions in the different sections. What was your rationale for this — why did you structure it that way?
RB: I created it because I want to focus on the process. If you look at the summative assessment guidelines, it focuses more on brainstorming — what is the algorithm, how did they come up with it? I was looking at the debugging ability of the students, and I wanted to see what bugs they encountered and how they fixed them. I was looking at the revisions they made — what are the changes, what are the new features they added? So if you look at the rubric, you’ll see the weighting is more on the process itself.
I created the sections so that students also know where the questions will come from during the oral defense — so they don’t feel blindsided. They also know from the rubric how much each section weighs, which tells them what’s more important.
TH: And these criteria are connected to your learning goals and also to the expectations for the AP exam?
RB: Yes, I make sure this is aligned to what is being asked in the external exam as well. It’s like hitting two birds with one stone. I want them to have fun, I want them to be responsible in using AI, but at the same time I want to make sure they are prepared for the external exam. And true enough, when they reflected on this, that’s actually what I saw in the reflection.
“I want them to have fun, I want them to be responsible in using AI, but at the same time I want to make sure they are prepared for the external exam. It’s like hitting two birds with one stone.”
Next week: “Be Honest, Grow With Your Students”: An Interview with Rose Baronia on her experiences with AI, Assessment, and a New Oral Defense (Part 3)
