Hacker News new | ask | show | jobs
by Al-Khwarizmi 61 days ago
Rather than banning the use of AI and trying to catch students who are cheating, why aren't they creating schoolwork that requires AI?

It's an interesting debate, but I see several reasons not to.

1. As a human you need to learn gradually, e.g. in CS you need to learn the basics of programming before going into more complex stuff. If you embrace AI from the beginning, it can let you skip the basics (why would you code a simple 200-line program if the LLM can do it?) and then you don't have the fundamentals when you reach the more complex level where human thought is needed. It's a similar problem as firing the juniors (because AI can do their work) but then who will become senior if you have no juniors.

2. If you evaluate coursework with the expectation of students using AI, those who pay the $200 subscription will have an advantage over those who pay $20, and in turn over those who use free LLMs. The only way to make it fair would be to provide all students with the best available LLM.

3. While I have heard the analogy with calculators many times, I feel that LLMs are at a different qualitative level. The calculator doesn't really replace human thought, or if it does, it's only some very specific form of it. LLMs replace human thought in a very broad way, so I think they are much more dangerous for learning.

2 comments

I agree with all these points. The algorithm to multiplying being done by the calculator, is not the same because you still have to learn why you are multiplying. If you rely on ai you may not know why the ai is giving the answer. All you’ve learned is copy paste. Figuring out how someone can use ai and give the wrong answer will solve the a huge problem. But it feels quite difficult at the time being. Ai is so general it’s hard to think of how to pose the question in a way that ai can’t answer. Maybe submitting prompts and judging prompts that bypass understanding. What if the test was more how to teach the ai? Meh ai in education is filled with gotchas.
The issue I have is that people are still trying to shove AI into a pre-AI educational paradigm. And yes, if you introduce AI into a world where people are still trying to teach kids the way that they were taught in the 20th century, AI looks like a threat because it allows kids to cheat, both testing and themselves. But we have the option to stop trying to teach kids the way that they were taught in the 20th century, which I think people here are fundamentally not understanding.

It's an old book now, but Neal Stephenson's Diamond Age includes the vision that we should have for education. We literally have the tools today to build his fictional "Young Lady's Illustrated Primer." What he envisioned was not that far off from an iPad with a Claude subscription where Claude has specific goals for the conversation. It's not teachers lecturing a class, it's individualized education where an AI teaches students at their own pace using their own interests. And built into AI is the ability for precocious kids to go beyond the curriculum, either on tangents or to more advanced subjects. This is impossible in a world where a teacher is trying to shepherd dozens of students through a curriculum as a group.

In the 2010s, we got some of the way there with Khan Academy. It was genuinely new that a student could rewatch something until it clicked rather than having to digest a lecture and have any question that didn't immediately spring to mind go unanswered. AI offers the possibility to go a step beyond this. Instead of rewatching the exact same content, AI can present it to the student in multiple ways based on a student's confusion and keep explaining it until a topic clicks. It can find examples of things that a student finds interesting to show how what they're learning isn't just theoretical. If a student likes space, the AI can discuss how the trig concepts they're learning apply to the Artemis II mission. If they like sports, it could apply the same concepts to tennis. Students in literary classes could read different books according to their interests while AI ensures that they understand the same sorts of concepts while discussing them. By customizing based on the specific curiosity of the student, it can make learning far more engaging and actually fun.

To address your #2, schools should be working with Anthropic, OpenAI and Google to shape a new personalized paradigm of educating students. They should be working out deals that give access to AI to their entire district. If I were heading the Department of Education, I would go a step further and get companies to bid on a contract to put their AI in the hands of every public school child in America. A version of the AI where teachers input their curriculum into the AI and students work through it with the AI, either alone or in small groups and the AI reports back to teachers so they can intervene where they are most needed would allow school districts with staffing shortages to serve more students more efficiently and with better results.

Sometimes it feels like our current system of education is only secondarily concerned with students actually learning and the primary concern is testing students to sort them into different tiers to be absorbed into different strata of our workforce. AI does compromise this sorting process to some extent. But if we can get back to the true mission of education and think creatively to deploy AI to best educate students, we have the potential to transform education like never before. What if we don't need to test students? An AI can give an individualized assessment of how well a student has grasped what they're supposed to be learning based on weeks of individualized work. It's as if we can give every student their own private tutor who will report back to the teacher on the student's actual progress. When you have that, stress-inducing exams are a ridiculous substitute.

I've been pretty shocked at how closed-minded the responses to my comment have been. We're supposed to be a community that envisions radically better futures that can be built with technology. And here we have a revolutionary new technology that upends a staid and increasingly problematic part of our society and the majority of the responses are geared towards explaining why that staid and problematic institution should be maintained unchanged. AI is fundamentally a danger to our current education model, but that model can change radically for the better. And I would've hoped that more people here would have recognized that.

Have you tried to learn a new topic with ai? Not like, learn the answer to a fleeting question. An actual large topic one might take even a short course on.

I say because most of the experiments I have seen in this space have failed. Chatgpt education was quietly removed not that long ago. Khan academy recently said that their Khanmigo AI tutor was facing challenges because students don't want to use it. Its a long-standing observation in the field of education that the miracle of the internet, computers being ubiqutous, phones etc hasn't clearly resulted in improvements to education (theres some minor evidence for or against it, but no blindingly obvious effect)

I worked at one point in ed-tech and the longer I was there the more I realised that nobody wanted this. Students only used it if they were made to, teachers only did it if admins wanted, admins only did it if they were sold on it, and the sales people seemed to be the only people who actually thought it was helping anyone

People often seem to think that teaching is the process of neutrally presenting facts to a learner. That the better and more clear the facts are presented, the better. But any book can do that. The entire business of education, as it were, is the management of motivation in students. Exams function just as much as a tool in this as the institution of coming to a classroom at all, rather than sitting at home remote learning. You need a clear mind, distraction free - just bored enough to find the content acceptably appealing (content cannot be made more interesting).

The best things for education seem to be:

- getting enough sleep

- getting a good diet

- good exercise

- being around other people you respect who are also learning the content

- being around other people you respect who have already learned the content, and who you want to emulate

- having someone who visibly cares if you learn the content or not, who expresses and reinforces that they expect you to learn it. (both "you need to focus, stop messing around" but also "you are capable of this, its hard but I know you personally can do it")

- having a good reason to want to learn the content

- having time pressure to learn the content

note that literally none of this regards the actual presentation of the content. Books have existed for centuries. A well motivated learner with all of the above will find the content. That is in no way the problem

An argument might be "but why, given all of this, are teachers actually teaching, why arent they standing up and lecturing soley about the importance of learning the content" and to some extent thats fair (although lecturing does satisfy some of the above if you think about it), but they ALSO tend to assign reading, link to further resources etc. Probably half or more of teachers speaking time is given over to procedural stuff working towards the above goals. Explaining the process of upcoming exams, worksheets, homeworks that need to be done (time pressure), demonstrating their love and knowledge of the subject (people you respect who know the content), building rapport with students (people you respect), holding people to account/motivating, explaining why the content is important, and trying to build good habits in their students (organising study groups, project work, dealing with problems and creating distraction free conditions)

AI are a poor facsimile of the above. Though they may try to replicate it, their inherent lack of physicality and humanity makes much of the power lose its effect. I dont care if a chatbot is disappointed in me. I'm not inspired by a chatbot that claims enthusiasm about a subject. I don't care if a chatbot tells me I absolutely must do this by next week else ill be left behind (behind who?).