Part 2: How should we bring AI into the classroom?

In the first piece of this series, we established how disruptive AI has been to education, as well as touched briefly on potential benefits should we begin to bring AI into the classroom. Most importantly, we referenced the findings that there need to be drastic moves made in our education system so that we can improve literacy, and this is likely a driver for exploring AI as a solution.

It entered, uninvited

I opened my last piece with the idea that students brought AI into the classroom and students were forced to respond. In the weeks between writing that piece and sitting down to this one, the AI-cheating scandal at Brown erupted. After unusually high grades and a surprising number of perfect scores on a midterm exam that was administered online, an economics professor suspected AI cheating. When he shifted to in-person exams for the finals, many students who did well or got perfect scores in the midterm dropped the course or didn’t take the final. And where the average score for the midterms was 96, it fell to 48 in the in-person finals.

Article continues after this advertisement

The simplistic, and often the usual approach when we hear about cases like this is, “How do we catch them?” Teachers will ask for AI detection software, in the same way that before we had plagiarism detection software. But such detection software is an arms race, where the models and the detection will continue improving. In the meantime, there is the very real possibility of false positives, which could lead to awful outcomes for students accused of cheating when they did not.

What I’m saying here is perhaps it’s not a technological solution that we need to be looking at. Just because the disruption is technological, it doesn’t mean that we need to answer it with another technology. One thing worth pointing out is, as I said in the first piece where I talked about DepEd’s DO3 putting a lot on teachers, again we are looking at teachers as the frontline who need to figure this problem out.

Here’s my first invitation to zoom out. I’ve been pinning a lot of this on students walking into the classroom with Generative AI. Because that’s exactly what it felt like. But underlying all of this is the way that ChatGPT was deployed. It was dropped onto the public as a consumer technology without sufficient testing or thinking about its impacts. In fact, it could be argued that even though OpenAI might claim Iterative Deployment, this was an act of carelessness.

Let me go into the possible differences further, because how you view this will color how you might think about deploying AI in an educational setting. When OpenAI launched ChatGPT in November of 2022, they claimed that they were releasing it to the public so that everyone could participate in its development. Iterative Deployment was there to signal that this was just one iteration, and that there would be further iterations. In product development, you’ll often launch and test with audiences, and then use audience feedback to then improve the product, then launch improved iterations.

Article continues after this advertisement

This might sound sensible. But then you have to ask, why was OpenAI ready to launch this product in this way, when other, bigger companies like Google and Meta had created similar chatbots, but were not launching them out of concerns for safety? In the case of Microsoft, they launched a chatbot, Tay, in 2016, but quickly pulled it down after users got it to say all kinds of awful things. There was a safety problem that OpenAI was either overlooking or ignoring altogether.

They claim it’s in the spirit of Iterative Deployment, but at the same time, they claim that this is such a powerful, potentially world-breaking technology. Remember that OpenAI was founded due to fears of Superintelligence and they wanted to build an AI that would be beneficial to all. If they truly believe in this transformative power and existential threat, then more caution might have been warranted.

Of course, an easy response to all this is, who could have imagined that if we released a powerful AI that students would use it to cheat? And then other people would use it for intimate purposes? Due to the decision to launch the way it did, we have not just rampant cheating of students, but the many cases of what’s being called AI Psychosis, and other damage that people interacting with chatbots has caused.

Article continues after this advertisement

How does this bring us back to education? Because OpenAI launched the way it did, then everyone else launched their products, sometimes to laughable results (remember early Bard), or even more overtly harmful outcomes (like all the nonconsensual sexual material being pumped out these days by Grok). Bottom line is this has led us to this race mentality. And it’s this same energy that fuels the need for AI adoption in education.

AI or get left behind?

The race mentality, coupled with the circumstances around challenges in education outlined in the Part 1 of this series, lead to the sense that Philippine education, from the earliest levels all the way to post-graduate, vocational, and upskilling, need to rush into AI education.

How many times have we heard, “AI or get left behind?” Or “You won’t be replaced by AI, but you’ll be replaced by a human who knows how to use AI?” And it’s these overriding, dominant narratives that seem to be pushing a lot of high-level decision-making.

And as we established in the first piece in this series, there’s an overwhelming sense that we need to overhaul our education system. With that pressure and the feeling that we have already fallen behind, those promoting AI can point to it as one of the solutions to our challenges.

I will admit my own biases here. I believe that with the right approaches and deployment, I think that AI could be incredibly transformative. At the same time, I emphasize that thinking through the approaches and doing the deployments right are the challenges that we will all likely disagree on, and there’s no way of knowing what exactly the right thing is. If we move too fast, we might trigger awful consequences or compromise our position. We move too slow, and we are left even further behind than we already are.

I think this entry in the series is more than anything a call for pause and a call for context. Let’s look at the situation we confront and all of the different pieces that are going to influence our decision.

On one side here are AI companies who rushed into deploying this technology, with what at least I perceive to be a lack of care for safety. In their rush, now they have raised massive valuations and need ways for that technology to be implemented to justify those valuations, and one of those implementations is in education. Governments feel pressure from companies, from constituents who want education systems to improve, and from themselves, as they look across at other countries who are adopting and deploying AI. There is going to be a lot of FOMO here, and that will be a driver of decision-making.

On the other side are those playing defense, like teachers who need to confront student bodies with pretty much unlimited access to Generative AI, whether it is good for them or not, or whether they are prepared to use it. There are people who are worrying about the larger social effects that chatbots might have. Those looking at the environmental costs of aligning ourselves with hyperscalers and giving them fuel to advance their buildouts. And in general there are a lot of people who are cautious, wary, or even just outright oppose Generative AI for other ethical issues such as copyright violations.

I suppose what I can call for to close this piece is to establish a very intentional approach to AI when we bring it into our classrooms and education systems. Let’s not get swept up in the narratives, the sense of inevitability or FOMO on one side, or just the blanket assertion that all AI use is evil. We need to take a breath, allow ourselves to look at the opportunities and threats the technology presents, and then start choosing how we implement.

We can develop this intentionality through increasing our understanding of how different AI systems work. I’m not asking that suddenly everyone involved in education policy and teaching become AI experts and AI systems builders as well. But we all need to develop AI literacy (this is its own contentious and hard to define term, but let’s use it for the moment) because we need to know how AI works. We need to demystify it, we need to understand it, and then we can make better and more intentional decisions as we bring it into classrooms.

In the next piece in this series, I’ll discuss studies that show the impact of AI use in classrooms so far and explore how AI invites us to think about learning and education differently.

Your subscription could not be saved. Please try again.

Your subscription has been successful.