3 ms·

I'm a (finishing) grad student and do a lot of teaching. While I definitely experience a lot of what's said in here I think a lot of people are getting things w

I'm a (finishing) grad student and do a lot of teaching. While I definitely experience a lot of what's said in here I think a lot of people are getting things wrong. It is easy to point at technology and just say the youngins are dumb/lazy/have no attention. > Students are literally finding it too hard to think. So they can’t learn new things. First, I have mixed feelings about this. The first part is right but the second part isn't. And it illustrates part of the problem. They can learn new things but we are not incentivizing them to. The thing is we need to rethink how we educate people. In fact, I think we need to rethink a lot of things. The problem here is you need to ask a lot of "why" questions. What I do like about the article is that it does talk about the addiction in short form mediums. That instant gratification. This is definitely part of the problem, but I'd argue extends far beyond students and isn't just from social media. Further upstream, I think a problem we have is over-metricization. Or what one might call bureaucracy. I've been calling it "Goodhart's Hell". The reason Goodhart's Law is a thing is because you can never perfectly align a metric to the thing you actually want to measure. So no matter how good your algorithm is[0] it won't be perfectly aligned. You need to evaluate beyond the metric and watch for drift. Goodhart's Law is more a warning about how people are good optimizers, so if you strongly incentivize the metric and only the metric, you end up with a billion cobras while trying to eradicate them[1]. So students are kinda doing what students have done for awhile. We've been having this conversation with cheating and we've had jokes about how you remember material for the test and immediately forget it after. These things are indicative of misalignment! Part of the problem here is that so much pressure is put on getting good grades that you will punish students for being honest. You punish students for struggling and gaining deep learning while you reward those who look up solution manuals and get homeworks from peers or even ask GPT. This is the root of the problem, and we can't solve these things until we rethink this. This is in the same way that we've created these addiction machines like TikTok, YouTube Shorts, and others. We strongly fit the metric at the expense of fitting our goal. We all know people who did really bad in school and on tests like SATs and stuff but are brilliant. Similarly we also know people who did amazing on all those things yet are bumbling idiots. Those are just examples of misalignment. Weirdly the more meritocratic you try to make things the less meritocratic things become. Not because you impose metrics, but because you overfit metrics. What's happening to students? It's the same thing that's happening to everyone else. I think we really need to rethink a lot of this. Luckily, there is action that a lot of you on HN can do! The problem is created through lots of small effects, thus the solution needs to be too. If you are just a junior engineer, ask questions about how well aligned a metric is to the goal. At worst you learn something, at best you make the product better. If you are a senior, think carefully about these things and encourage your team to too. Expertise is about understanding depth and nuance, so encourage that! If you're a manager, IT IS YOUR JOB TO CAREFULLY ENSURE THINGS ARE WELL ALIGNED. That's one of the most important things you can do! I know a lot of the pushback, and I hear "but you can't assign value to <x>". But here's the thing, all those numbers are fucking made up anyways. Use your expertise and try to make products better. Not better defined by what makes share value go up or your manager says is "value" but just make the user experience better! Make things that make peoples' lives better! That was the promise of computing in the first place and why it created these billion dollar companies. I promise the money will follow. There is nothing stopping us from making money hand over fist WHILE making good products that make peoples lives better. If you believe these are at odds with one another, then do you really just want to build a dystopian Sci-Fi future? Are you really just okay "sacrificing your soul" for a paycheck? While I understand the calculus isn't so easy for everyone, we are talking about a field where $100k-$400k salaries are not uncommon, and you definitely have a choice in those ranges. But importantly, be a little grumpy. After all, that's your job as a developer/engineer/researcher: to find problems and fix them. Tldr: think carefully if you want to enable others to think carefully [0] I'll admit the irony of me saying this. I am extremely math oriented and work in machine learning (https://news.ycombinator.com/item?id=43442631 https://news.ycombinator.com/item?id=43442631) [1] https://en.wikipedia.org/wiki/Perverse_incentive https://en.wikipedia.org/wiki/Perverse_incentive

A4ET8a8uTh0_v2 2y ago
I hesitated a little bit before attempted to respond, because I think this subject returns often, maybe even too often. I am not saying this because I disagree with you. I am not saying this because I really disagree with the premise presented by author. I am simply not sure we can really add anything truly new. << Tldr: think carefully if you want to enable others to think carefully It is not a secret on what works in education. There is no real secret sauce. Some things are well known and apparent even to the casual observer, but, and it is not a small but, it does not pay the bills of the cottage industry that has grown around education. For example, it is easier to throw money at everything, but ensuring small class sizes, which dramatically affect long term outcomes. I am not cynical. At least in US, I do not understand how anyone can pretend anything else is true.
- godelski 2y ago
  I'm not sure I really disagree much. We both agree on the main part about the perverse incentives. But > I am simply not sure we can really add anything truly new. I do not like these types of attitudes. Defeatist (I apologize if I'm misinterpreting). Our world has progressed so we have strong evidence that we can add new things. Not only that, but we can make things better! It wasn't long ago people shit in a bucket and threw it out their window. While there's a lot of shit now I have no doubt things, in the long run, have improved. Even if things are worse now than 5 or 10 years, in the long run things improve. The other part is that just because you're a small cog in a large machine doesn't mean you can't change things. I very much understand this feeling and I think it's natural too! But the cog is needed, even if there's some redundancy. You don't have to be a giant cog to have impact. The reason so much goes wrong is also why this last part is true. Things exist in an unstable equilibrium, like an inverted pendulum. It takes little to throw things off balance (I can explain more if needed, with examples). Which is the very reason why small things matter. But there's observation bias and we can't forget. When things work smoothly we often don't think about it and all the little things that needed to go right to make that happen. But we do think when things go wrong. Of course we have this bias, because we want to relax and reduce cognitive load. But the world has gotten more complex and so we probably need to think about this more. It's becoming more important. Because as things get more complex low order approximations become less useful. It's the bitter curse of advancement. The bitter curse of adding new things. But I believe we can do it

A4ET8a8uTh0_v2 2y ago

I hesitated a little bit before attempted to respond, because I think this subject returns often, maybe even too often. I am not saying this because I disagree with you. I am not saying this because I really disagree with the premise presented by author. I am simply not sure we can really add anything truly new. << Tldr: think carefully if you want to enable others to think carefully It is not a secret on what works in education. There is no real secret sauce. Some things are well known and apparent even to the casual observer, but, and it is not a small but, it does not pay the bills of the cottage industry that has grown around education. For example, it is easier to throw money at everything, but ensuring small class sizes, which dramatically affect long term outcomes. I am not cynical. At least in US, I do not understand how anyone can pretend anything else is true.

godelski 2y ago
I'm not sure I really disagree much. We both agree on the main part about the perverse incentives. But > I am simply not sure we can really add anything truly new. I do not like these types of attitudes. Defeatist (I apologize if I'm misinterpreting). Our world has progressed so we have strong evidence that we can add new things. Not only that, but we can make things better! It wasn't long ago people shit in a bucket and threw it out their window. While there's a lot of shit now I have no doubt things, in the long run, have improved. Even if things are worse now than 5 or 10 years, in the long run things improve. The other part is that just because you're a small cog in a large machine doesn't mean you can't change things. I very much understand this feeling and I think it's natural too! But the cog is needed, even if there's some redundancy. You don't have to be a giant cog to have impact. The reason so much goes wrong is also why this last part is true. Things exist in an unstable equilibrium, like an inverted pendulum. It takes little to throw things off balance (I can explain more if needed, with examples). Which is the very reason why small things matter. But there's observation bias and we can't forget. When things work smoothly we often don't think about it and all the little things that needed to go right to make that happen. But we do think when things go wrong. Of course we have this bias, because we want to relax and reduce cognitive load. But the world has gotten more complex and so we probably need to think about this more. It's becoming more important. Because as things get more complex low order approximations become less useful. It's the bitter curse of advancement. The bitter curse of adding new things. But I believe we can do it