Tag: higher education

In Praise of Random Questions

I’ve written about this before: I am in the habit of asking my students to ask me five random questions at the end of each class. The questions can be about absolutely anything, the only condition being that they cannot be about the topic we have discussed in class.

The reason I bring this up is because Tim Harford’s column in the Financial Times serves up it’s own share of delightfully random questions:

Last year, inspired by Randall Munroe’s delightful books What If? and What If? 2, I invited the good folk of Twitter to ask me absurd hypothetical questions about the economy, to which I would attempt some serious answers. This year, we’re going to do it all again.

https://timharford.com/2024/01/more-of-your-crazy-economics-questions-answered/

Tim Harford covers five questions in this column, and while I encourage you to go read his answers in his original blogpost, I’ll list out the questions here:

  1. How big would an asteroid made of precious metal have to be for it to be worth doing a space mission to bring it back?
  2. Could a universal currency ever be based on electricity, or currents?
  3. What if your tax bill was discounted by the distance you lived from the centre of London (eg if you lived in Kingsway, you paid the full amount; if you lived in Shetland, you would pay no tax)?
  4. What if inflation was made illegal? Could we legislate that no prices could ever rise?
  5. In the UK, we used to print banknotes on paper, now it’s a horrible slippery plastic. Could we use a more environmentally friendly material, like leaves? Or perhaps something edible — printed on some sort of simple flour and water biscuit? No waste!

Magnificent – just magnificent questions. In answering them, Tim Harford covers (without always naming them explicitly) scarcity, relative prices, optimization, the functions of money, volatility, Jane Jacobs, relative prices, price as a signal (not to mention an incentive!), Douglas Adams (!) and the supply of money.

One needn’t stop there, of course. You could introduce many other topics into the discussion, in addition to those already covered by Tim Harford. But the point here is that all of these topics could be introduced in response to the questions asked by the students. These topics aren’t being introduced “because they are a part of the syllabus”, but because these are helpful concepts that help in formulating sensible responses to delightfully random questions.

Sensible responses to delightfully random questions is exactly what “What If” and “What If 2” are all about, as Tim Harford points out in his column. And that, of course, is the guiding spirit behind such questions – students should tap into their natural curiosity about the world, and ask mind-bending and patently ridiculous questions.

This forces the person answering the question to build up their response from first principles, in order to show why and how the answer to the question cannot help but be ridiculous. What a remarkably powerful way to teach, no?

Why is it remarkably powerful? It is remarkably powerful because we ask the students to imagine a better world, if only we could do x. Look at the fifth question in the list, for example.

And then the question becomes, well, why can we not do x? Well, here’s why: reason 1, reason 2 and reason 3. In order to explain our reasoning, we introduce concepts. We don’t introduce concepts Because That Is What We Do So That The Syllabus Is Completed – we introduce concepts in order to answer questions that students came up with.

And I don’t know about you, but both my theory of how the world work and my experience for having taught introductory economics for many years tell me that students are much more likely to be attentive when it is weird questions being answered sensibly.

Or here is another way of putting it: memorizing a textbook, and occasionally asking if it is applicable to the world can get pretty boring. But looking at the world, asking questions about how we can make it better, and then learning about concepts that either validate or invalidate the premise of the question – that’s not boring at all.

Would you rather memorize the functions of money, for example, or would you rather wonder about why we cannot use electricity as a currency? Or time, for that matter! Now, if you were to ask me why we couldn’t use electricity or time as money, I may have to end up explaining the functions of money to you, and also explain why these functions can’t be carried out by time or electricity.

Ah, you might say, that makes sense. Now you know why time can’t be money (hehehe), plus you’ve ended up learning about the functions of money. Fun discussion, plus learning. What’s not to like, I ask you.

If you aren’t asking weird questions, you aren’t learning.

Ask weird questions!

The Death of the Classroom, *NOT* the University

This post is a continuation of my post from yesterday, available here.

I’ve been predicting the death of the classroom for three years, and wishing for it for far longer. We have classrooms, and continue to have classrooms, for the same reasons that factories organized themselves around the steam engine: it was the best response to available technology. That is, in a world without the internet and AI, in person synchronized learning with a figure of authority leading the class was the most efficient way to make learning happen.

And while we do have the internet and AI now, we have become so used to the status quo of the classrooms in universities that we find it difficult to reimagine what a classroom will look like given today’s technology.

Here’s one way to see classrooms: they’re solutions to coordination problems. A lot of students wish to learn a particular topic, a professor has knowledge about said topic, and a class in that classroom is a way to coordinate the dissemination of that knowledge to those students at the same time. For many years, if not centuries, it was the only way to disseminate that knowledge efficiently. There were, to use the language of the economist, no substitutes available that could achieve the same result at lesser cost.

But with the advent of the internet, and especially AI, that is no longer true. I can turn up tomorrow at a time and place of your choosing and give any number of students a class on, say, the principle that incentives matter. But all the students who attended that class by traveling to that place at that time could just have sat at home and learnt about this principle from ChatGPT. They could have had discussions with economists, sociologists, theologians and anthropologists and linguists about incentives. Pretty soon, they can build customized videos of their favorite sports personalities giving them a class on this topic.

Now, you might bristle at the thought and say that this couldn’t possibly be better than learning from a human. Not now, not forever, you might insist. But what if one of the students has to travel for this class from, say, Gadchiroli? What if this student cannot understand English all that much. What if this student learns best when concepts are sung to him, rather than spoken? What if he doesn’t relate to some of the examples being discussed in class? It needn’t be a student from Gadchiroli, and it needn’t be this particular list of problems. The point is that no human, no matter how good and multi-lingual she may be, can ever hope to achieve the level of customization that AI can. And once you take into account travel and coordination costs, it’s game over. You may still say that being taught by humans is better, but it is already no longer as efficient. That’s just a fact.

Why, just this past semester, I finished teaching a course in Principles of Economics to students at the Gokhale Institute. There were a hundred and fifty students in the class, and so I gave forty speeches across the semester. That is, there was absolutely no chance of in-depth conversations, detailed feedback and customized learning.

Why not split up the class instead, you ask? Because cost, of course. You save on the money you pay to the faculty by combining classes, quality be damned. From my own selfish perspective, combined classes are great, because saying the same thing twice is boring as hell – but from the point of view of the students, it is a whole other story.

So why not take the next logical step and save the cash that you have to pay for just the one combined class too? Why not have AI take the class instead? I’m quite serious, AI should be putting people like me out of my current job.

Note, however, that I’m saying it should be putting me out of my current job. When I write posts about the death of the classroom, what I am really hoping for is the death of universities structured around the concept of in-person classes the whole day, day after day, for two years.

AI solves the coordination problem, and takes away the need for in person classes. But at the same time, and as a direct consequence, it raises the need to have in person mentoring, and in person interactions with peers. The role of the in-person university with a physical campus goes up, not down, because of AI.

The classroom need not solve a coordination problem, but the university can (and should) become a coordinating point. It becomes a place where students interact with their mentors. It becomes a place where guest lectures take place. It becomes a place for in-person seminars, conferences, talks and walks. It becomes, in short, a place for ideas to blossom and bloom.

Cafeterias, amphitheatres, lawns and cafes will dominate such a campus, not classrooms. We will need Schelling points for discussions, as opposed to Schelling points for listening.

The truth, of course, lies somewhere in the middle. So all my rhetoric and flights of fancy aside, there will still be demand for (and therefore space for) classroom based learning. But it will be a supplement to other kinds of learning, as opposed to being the primary mode that it is today.

A university dedicated to the spread and discussion of ideas, as opposed to a university that drowns you in classes, day after dreary day.

My, what a revolutionary thought experiment.

The Economist on AI and Transforming Education

Almost three years ago to the day, I’d written a post on The Long, Slow, but Inevitable Death of the Classroom:

When the pandemic ends, whenever that may be, do we swing back to the other end of the spectrum? Does everybody sit in a classroom once again, and listens to a lecture being delivered in person (and therefore synchronously)?

Or does society begin to ask if we could retain some parts of virtual classrooms? Should the semester than be, say, 60% asynchronous, with the remainder being doubt solving sessions in classroom? Or some other ratio that may work itself out over time? Should the basic organizational unit of the educational institute still be a classroom? Does an educational institute still require the same number of in person professors, still delivering the same number of lectures?

In other words, in the post-pandemic world…

How long before online learning starts to show up in the learning statistics?

https://econforeverybody.com/2021/01/28/the-long-slow-but-inevitable-death-of-the-classroom/

And three years later, we have our answer, from the Economist:

The sector remains a digital laggard: American schools and universities spend around 2% and 5% of their budgets, respectively, on technology, compared with 8% for the average American company. Techies have long coveted a bigger share of the $6trn the world spends each year on education

https://www.economist.com/business/2024/01/11/ai-can-transform-education-for-the-better

Higher education is a bundle, of course. When you enrol with a university, you are purchasing an education, a degree and the ability to build the kind of networks with your peers that you’re never ever going to be able to build again. And online education takes away the misery of having to listen to bad professors drone on in classrooms, sure, but does nothing to solve the problem of having a degree that very few other people have. And it is really, really bad at helping you build out good peer networks.

So the death of the physical classroom isn’t imminent just yet – not because we fell in love with bad professors and musty classrooms with “smart” boards after the pandemic, but because the degree continues to matter, and because nothing (nothing!) beats bunking classes with friends.

But the answer to the question implicity posed by The Economist article – why did classes resume much as before post the pandemic – is quite simple. Because in addition to the lure of being one among a select few who gets to clutch a degree from a hallowed university and the awesomeness that is hanging out with friends IRL, online education simply meant that you got to listen to the same bad professor, except it was online.

And that is worse! The prof is as boring, but you are listening to that boring prof in your PJ’s, in bed. Which is very welcome one day out of five, sure, but for two long years? Fuhgeddaboutit.

But a fun prof who gives you customized, tailored teaching and mentoring? A prof who customizes their teaching style, their pedagogy and their problems tailored to how well you seem to be learning? A prof whose lectures you can pause and resume, as needed, on a 24/7 basis – maybe that will work?

Two-fifths of undergraduates surveyed last year by Chegg reported using an AI chatbot to help them with their studies, with half of those using it daily. Indeed, the technology’s popularity has raised awkward questions for companies like Chegg, whose share price plunged last May after Dan Rosensweig, its chief executive, told investors it was losing customers to ChatGPT.

https://www.economist.com/business/2024/01/11/ai-can-transform-education-for-the-better

The Economist article goes on to point out how education specialists might end up doing a better job than plain vanilla GPT. They argue how education specialists, such as Chegg’s and their like know the ins and outs of the education business, and will therefore likely do a better job at customizing and deploying AI in education. This is, the Economist says, because of the following reasons:

  1. Pearson, McGraw Hill and some other publishers haven’t made their data available to ChatGPT, and are instead incorporating AI into their own products
  2. Chegg’s and friends are following a similar approach, and have years of mentoring related data ready to deploy.
  3. Firms in this sector have “an in” with educational institutes already, and that will make their pitches about deploying AI more palatable to educational institutes.

Maybe so, and I honestly don’t know how this will play out. Maybe ChatGPT will get better, especially with the launch of their store. Maybe the competition will be definitively better than ChatGPT.

But us boring ol’ profs have competition, and lots of it. As The Economist mentions, we may have to “shift our attention to motivating students and instructing them on how to best work with AI tools”.

That last bit I agree with most passionately. The job of educators in the age of AI isn’t to teach, but to mentor. Our job is to help students learn, not teach them. This statement is banal to the point of being a platitude in education, but with AI, there may well be an “iota of truth” in there now.

Along with – for now – an iota of inevitability.

On True Roles

Arnold Kling has a nice post up on his Substack about “The True Role of the Central Bank“.

He was asked recently about three ideas that he (Arnold Kling) is known for, and these are the three that he came up with:

  1. Subsidize demand and restrict supply: Government intervention might well be thought of as the subsidization or provisioning of what the market fails to provide. But in practice, he says, the political process tends to be controlled by incumbent producers or owners. These guys are going to lobby the government for subsidized demand and restricted supply. He cites the example of housing: subsidize demand (subsidized home loans, for example), and restrict supply (zoning restrictions). Might higher education be another? Do we end up restricting the supply of quality higher education, and subsidizing demand for it by giving subsidized student loans?
    If you distrust the private sector, now would be a good time to say, “Hah! See? Big bad incumbents derail well-meaning government”
    If you distrust the government, now would be a good time to say, ”Hah! See? Government is as corrupt as big bad incumbents”
    If you agree that incentives matter, you might want to think about how to redesign various systems with better incentive design front and center. But that’s boring work. Saying “Hah! See?” is much easier.
  2. His second idea is that “price discrimination explains everything”. Given high fixed costs in so many different industries (especially Internet-based businesses), the marginal cost of serving an additional consumer is zero. So if marginal-cost pricing makes no sense, what to do? Announce Big Billion Day sales, for example. How’s that price discrimination, you say? That’s just low prices all around for all goods, you say? Well, what about the rest of the days in the year, when the Big Billion Day sale isn’t around? Who do you think is buying then? Are those folks paying the same price for the same good? And that’s just one example. You could teach a semester’s worth of micro by using nothing more than the Amazon app on your phone!
  3. And his third idea is to do with the true role of the central banks. In the 2008 financial crisis, he says, the main concern of the Fed wasn’t forestalling a recession, but rather to focus on the health of the primary dealers. Or more simply put (although you should read the whole blog, as always), the central bank’s top priority is “always going to be enabling the government to borrow more money”. His last paragraph is worth quoting in full:
    “The Fed’s job is to make sure that the Treasury can market its debt. For that purpose, it has to be much more concerned with keeping banks healthy than with hitting a target for inflation, unemployment, nominal GDP, or any other supposed goal.”

Fascinating ideas, all of them. But the last one in particular mad me think about the phrase “the true role” more generally.

What is the true role of:

  1. Educational institutes?
  2. Hospitals?
  3. The Patenting System?
  4. YouTube?
  5. Students?

Ask yourself, and the people around you, these questions. And add to the list! Try answering them yourself. See if the answers differ, and ask yourselves what that reveals about the items on these lists, and about the respondents.

My Random Question to My First Year Students

I just wrapped up a semester of teaching at the Gokhale Institute. It is my favorite course (Principles of Economics) to teach, at my favorite place, so a bittersweet moment of sorts.

And the last class was an extended five random questions session, with lots of fun questions coming my way. One of which, it turns out, was a request for me to ask them a random question. Fun request, and here is what I have asked them:

“You get to redesign higher education from the ground up. All higher ed institutions are scrapped, and society, industry and academia will go along with the institutions, culture and regulations that you choose to construct/create to make higher education as good as it can possibly be – good itself being defined howsoever you like.
What will you do, and why?”

They have all the time in the world to answer, and of course it is not mandatory to do so. But should you choose to answer, I would love to hear it! So please, do let me know the how, the what and they why of your proposal to change higher education in India for the better.

Happy Diwali, everyone!

(I hope to post everyday next week, but am very much on leave. We’ll see!)

Ajay Shah and Nitin Pai on Higher Education in India

The IIT JEE (Indian Institute of Technology joint entrance exam) is revered as the arbiter of merit. With industrialised coaching classes, it is less clear how the JEE selects the right people to attend an IIT. Simplistic measurement of marks in an exam is not how the entry barriers into most sensible institutions work. The high-powered incentive — attending an IIT — is damaging the learning process. We propose a two-part mechanism: A broad exam that filters for sound capability, and then randomised allocation. The overall impact of such a mechanism would be positive. Test preparation has corroded Indian education. Across India, children no longer attend just high school. They are enrolled in coaching classes. Here, it is not necessary to study the subject and understand concepts. All they need to learn is the finite list of multiple-choice questions (MCQs) that are likely to be asked in entrance examinations.

https://www.business-standard.com/opinion/columns/doing-better-than-a-competitive-exam-123082000552_1.html

Two of my favorite columnists writing about a topic that I am most passionate about – the conduct of examinations in India. What’s not to like, and how can I possibly not write about it?

  1. As Nitin and Ajay mention, “test prep is not education”. Indeed. Test prep is just that: test prep. You can get very good at test prep, but that doesn’t mean that you will be educated.
  2. They have their own definition of education, and while I don’t disagree with it one bit, I have a shorter one.
    Education is about learning to ask (and then answer) really good questions.
    My problem with our education system, and with exam test prep, is that we simply don’t teach our students the art of asking really good questions.
  3. “The tendrils of curiosity, dissent, imagination, creativity, and risk-taking are likely to be crushed in these years. We are creating followers, not leaders.”
    If I nodded any harder, I could probably power a wind turbine all by myself.
  4. “Upgrading high-school syllabi and implementing the National Education Policy 2020 are sometimes proposed as the answer. This is insufficient because they do not fundamentally change students’ and their parents’ incentives. Real change will come only if a seat in an elite academic institution is no longer seen as the sole objective of going to school.”
    Two first-pass responses to this issue would be to either increase supply or reduce demand. That is, increase the supply of high quality education seats (easier said than done), or reduce the demand for them (not gonna happen). Therefore their solution: lottery system.
    “The price of an IIT lottery ticket can be reduced to zero. Seats can be randomly allocated to applicants who meet basic requirements. Specifically, we could envision a first-level exam, which is not about the things that Google knows. Out of that the top 200,000 ranks are shortlisted. At the second stage, a random list of 20,000 would be chosen to attend the IITs. Such an approach, we contend, outperforms the current method on several important dimensions.”
  5. It takes away the incentive of the mad entrance test world that we live in, sure. But I worry that “universities and engineering colleges around the country will feel customer pressure to upgrade their standards because they will now encounter candidates with higher expectations and ambitions” won’t necessarily work out in practice. Students enrolled in an institute have *massive* switching costs, and colleges know this all too well. Trust me on this one: they don’t feel no customer pressure.
  6. My biggest problem with their solution isn’t with their solution per se, but with the fact that it ultimately boils down to whether we can scale up the supply of high quality higher education in our country quickly enough. I happen to think the answer is no, and I hope to god I’m wrong.

A chat with Pi about blended learning in Indian higher education

Arnold Kling:

I tried Personal Intelligence (Pi) from Inflection AI. As a chatbot companion, it charms you by offering encouraging reactions to what you tell it. After commenting on what you have to say, it always asks an interesting question. Think of it as a very skillful and probing interviewer. Yes, it’s only software playing a game with you, but it plays it well.
To get an idea of where a conversation with Pi can go, see part of my chat with Pi. The excerpt I posted starts with its message after I’d told it about my Marginal Revolution is Dead post. I predict that you’ll be impressed by it.

https://arnoldkling.substack.com/p/gptllm-links-dfa

I did go and see Arnold’s chat with Pi, and yup, I was impressed.

Impressed enough to have a conversation with Pi myself. What topic would you guess I chose? The one closest to my heart in a professional context, of course:

I am convinced that classroom education in higher education in India is inefficient, takes too much time, leads to sub-optimal learning and rote memorization for examinations. This is because of a lot of different factors, most of which are interlinked with each other in many different ways. But long story short, young people in India, even in the very best universities, do not learn as well as they could. And since best practices trickle down to other universities, we end up creating a culture of learnng that is sub-optimal at all levels.

https://pi.ai/s/Z3QUR9V69aAmNgc2ek1rS

If you, like me, are convinced that classroom learning is overrated, please do go and read my conversation with Pi. If you, unlike me, are not convinced that classroom learning is overrated, definitely go and read my conversation with Pi, and please do tell me where I’m wrong.

Three points that I would like to highlight:

  1. Far too much of student’s time is spent in passive listening (and that for hours on end). Reduce it dramatically, and even the bit that remains should be online. If your choice is between packing a hundred and fifty students into a classroom like sardines or allowing students to learn online, go online. Is online bad? Well it’s not perfect, sure. But relative to what alternative? If the alternative is the sardines-in-a-can approach, then why not?
  2. To me, the job of a professor in higher education is to mentor, not to teach. This is not a binary variable, and the truth lies somewhere in the middle, but more mentoring than classroom teaching, that much is for sure. So if anything, the workload for a professor will go up in my proposal, not down. But more personalized teaching/mentoring. Leave the large scale classroom to online education. What else is it for?
  3. AI in education is coming. You may not like it, you may resist it and you may say (as a professor) “but what are we here for then?”. But read the rest of Arnold’s post and ask yourself if the median professor in your university is better or worse than AI tutoring. Then ask yourself how many students this professor can mentor/tutor. Again, AI in education is coming. But the answer to the question “but what are we here for?” lies in learning to think of ourselves as complements to AI. Ask what AI can’t yet provide, and provide that. What the “that” will be changes based on a variety of factors, but in my specific case, I would think it is working on projects with students. And that is what I am focussing on this year.

I sent my conversation with Pi to two friends of mine, who gave me extremely thoughtful responses.

Samrudha Surana highlighted the fact that I should also be thinking about the question “what is college for?”. Different students want different things from the same course. Some may wish to become professors, while many more may wish to join the corporate world. This is as it should be: higher education is not a replication machine, whose sole job is to produce more professors over time. But we need both (future professors and future employees in the corporate world), and people trained in many other professions besides. Blended learning, and AI’s introduction allows for more customization, and that is a good thing.

He also pointed out that we should carefully think through how the online courses will be chosen by the students, and to what end. What he means by this is how much of a say a student should have in choosing their course(s), and how much of it should be the decision of the professor. Not just the choice of the courses, but also the choice of the project/assignment/paper for which the course is being taken. He favors more autonomy for the student, and while I’m inclined to agree, the magnitude will be tricky to set as a rule. In general, a higher degree of autonomy in later semesters, I would think.

Much more discussions are needed in our classrooms. Much, much more. We professors need to be challenged in class, our assumptions and claims scrutinized, our premises questioned and our conclusions critiqued. Learning is best achieved through Socratic discourse (in my opinion). But our classrooms are more about proclamations by the professor rather than any of the above. Smaller class sizes will help, as will more seminars, discussion groups and workshops. That matters, and is rendered more probable under such an arrangement.

Undergraduate courses, finally, might involve much more of classroom learning in the initial semesters. Although under the new four year undergraduate programme (and the likelihood of them more or less replacing Masters programmes altogether), even here you would want to shift to more of blended learning towards the end of the degree.

We need to teach students in higher education better, of that I am convinced. What I have suggested here is worth further discussion, I’m fairly sure. Whether you agree with me or otherwise (and I hope it is otherwise), please tell me why 🙂

Put Me Out of a Job – 2

Let’s begin with the second class today. Your outline mentions the topic “Time Management and Opportunity Cost”. Before we begin the class, outline a definition of both terms, and explain their importance to my life. When you focus on the importance of time management to my life, make sure that you remember I am an eighteen year old. Why should I bother with managing my time? I have my entire life ahead of me – time isn’t a constraint, surely? When then do I need to manage it?
Once you explain your answer to these questions, proceed with the outline as discussed, as per the format I have asked for.

You might think this (time management and opportunity cost) to be a weird topic for a second class in a course called “Principles of Economics”. You would certainly think it to be unconventional. Not the latter half of the topic – opportunity costs – but the first one. What does time management have to do with economics? Well, think of it this way – if you are an Indian student who has learnt economics, you have almost certainly come across Lionel Robbins’ definition, and have most likely memorized it back then.

Here it is: the science which studies human behaviour as a relationship between ends and scarce means which have alternative uses.

What is more scarce than time? We all have a limited amount of time, and we all have ends to achieve. The ends we would like to achieve in our lives are much more than the time that is afforded to us, and so we must choose which of these ends to pursue, and which to sacrifice. That’s a pretty good, and if you ask me, useful way to understand opportunity costs. The weirdness, or the unconventional choice (to some) of the topic, is a direct consequence of my request to ChatGPT from yesterday’s post:

Optimize for helping me understand the principles and their applicability to various aspects of my life (professional, personal etc), and it is fine if the outline doesn’t match what is usually taught in introductory economics textbooks. Given this, would you like to change your outline in any way? If yes, please show me the new outline. If no, please begin with the first class.

I am seeking to learn economics not to write an examination at the end of the semester in a college or university with this course. I am seeking, instead, to learn economics in order to apply it to various aspects of my life. Of course, as a student enrolled in a university, you may seek to optimize your learning geared towards doing well in a examination at the end of the semester, and that’s fine. Just let ChatGPT know accordingly, that’s all:

A minor tangent before we get back to our course. What if I asked you to change your course outline for these thirty classes to help me write an examination in an Indian undergraduate course? What would the new outline look like? In your answer, show me both outlines together, per lecture. That is, the first line in the description of lecture 1 (out of 30) should be your original lecture, and the second line should be the course tweaked for doing well in an examination, rather than applicability to various aspects of my life. Ditto for the second course, and so on and so forth for all thirty.

I won’t show you the whole output, but simply how the second lecture’s outline has changed:

Here’s a revised outline for the thirty lectures, with the original lectures designed for applicability to various aspects of life (Lecture A) and the modified lectures tailored to help you succeed in an Indian undergraduate economics course (Lecture B):
Lecture A: Time Management and Opportunity Cost
Lecture B: Opportunity Cost: Definition, Types, and Examples

ChatGPT-4’s Output (in part)

I know which one I prefer, and why. The good news, as a student, is that you can do both! Learn in order to score well in an examination, and also learn in order to figure out how to apply economics better in the case of your own life. Why should the two be different? Ah, some questions you should reflect upon, rather than ask ChatGPT.

Anyway, back to our lecture du jour. I asked ChatGPT to explain why it chose time management, and I do not think I would have asked that question as an eighteen year old. The older you get, the more aware you are of how limited your time is. And at least in my own case, the converse is also true. I count this as a mark in my favor – that while a good prompt may get a student going, said student will still need help and advice on an ongoing basis.

So far, at any rate.

Further proof of that fact that I’m not out of a job, just yet, is below. The context is that I read the answer, and felt it to be incomplete. So I prodded it a bit, and then just a little bit more:

I’ve often read the phrase “all costs are opportunity costs”. Please explain what this means. Remember that I know no economics, and as usual, give me one example from the Mahabharata, and one from a real-life situation

My professor wondered why the word “TINSTAAFL” hasn’t come up in your answers yet. I didn’t even know this was a word! Is he joking, or is this word relevant to what you’re telling me right now?

To be clear, it isn’t so much about the phrase TINSTAAFL, as it was about the fact that I felt its explanation to be incomplete. This prompted me (no pun intended) to ask it to be more thorough:

I have an ongoing request. I’m looking to make my professors job as easy as possible, or even make him, in some sense, redundant. Optimize your answer for thoroughness, and if you think you can’t fit all of what you want to say in a single output, end with a line that says, “I can tell you more, please let me know if you’re interested to know even more” This will always be applicable in our conversations.
Now, back to the second lecture’s outline. Expand upon the three sub-points from the broad contours. I am particularly interested in the third one, so give more details, explanations and background in the case of “recognizing and evaluating opportunity costs in decision making”

This is an important lesson in and of itself. Feel free to tell ChatGPT to give more (or less) detail, or ask it to modify how it gives you the answer (more examples | simpler language | write like person X | show your output as a debate between person X and person Y). Get your “teacher” to be the kind of teacher that you like to learn from!

With regard to your explanation of “Recognizing and evaluating opportunity costs in decision-making”, I’m confused about how to think about short term and long term factors while making my choices, and the short term and long term consequences of my choices. How should I think about this, what framework should I use, and is there an underlying principle at work here that I should know about?

I count this as a pretty important miss on ChatGPT’s part. My personal opinion is that you haven’t fully explained opportunity costs without talking about the importance of how your evaluation of opportunity costs changes given different time horizons. Time matters! ChatGPT actually agrees with me (see below), but only after prodding. And this after making explicit the fact that I was interested in learning about time horizons! And so I asked it again:

Is it useful to think of time preference as a separate principle of economics? More broadly speaking, how should a student of economics think about time preferences? Give me answers from a theoretical perspective, but also from an application perspective.

I’m two days in, where I’m the “student” and ChatGPT my teacher. Today’s class wasn’t great. I don’t think ChatGPT’s output was good enough to stand on its own, and it needed additional prompts to deliver what I would consider to be a good introduction to the concept of opportunity costs, its many nuances and its many applications. It wasn’t bad, but it was far from being good, in my opinion.

Should I take this as a sign that I need to get better at writing prompts, or should I take this as a sign that AI isn’t good enough to replace me yet? How should I change my mental model about whether the average student in a typical college can learn better from AI?

If you are a regular reader of EFE, you know what’s coming next: the truth always lies somewhere in the middle.I need to get better at writing prompts, yes, but also AI isn’t good enough to replace me yet. Both of these things will change over time, of course, but for the moment, less than ten percent into the course, I am inclined to think that I am not out of a job, just yet.

And even better, the complements over substitutes argument just got stronger – I’ll be a much better teacher of a course such as this the next time I get to teach it. Tomorrow we tackle “Supply and Demand: Basics and Market Equilibrium”.

I’ll see you in class tomorrow!

Put Me Out of a Job – 1

Let’s say you’re a student who is going to start learning economics in the coming semester (starting July 2023). Let’s assume that you’ve never learnt economics in a classroom before, save for a brief introduction to it in high school. If you chose to learn from an LLM instead, how should you go about it?

Leave aside for the moment the question of whether you should be doing so or not. The question I seek to answer over many blog posts is whether you can do so or not. Whether or not this is a good idea for you depends in part on my abilities to add to the value that an LLM generates for you from such a course. And once these thirty (yes, thirty) blog posts are written out, I’ll write about my thoughts about whether a student still needs me in a classroom or not.

My current thinking is that I would still be needed. How much of this is hope, and how much dispassionate analysis is difficult to say right now. For that reason, I would like to tackle this problem at the end of this exercise. For the moment, I want to focus on helping you learn economics by teaching you how to learn it yourself, without the need for a human teacher (online or offline).

In each post, I’ll give you a series of prompts for that particular class. I will not always give you the output of these prompts – feel free to run them as they are, word for word, or tweak them as per your likes, fancies and hobbies.

My motivation in this series is twofold. One, to find out for myself just how much better ChatGPT is than me at teaching you principles of economics. Second, to help all of you realize that you ought to hold all your professors (myself included!) to a higher standard in the coming year. We have to do a better job than AI alone can, along all dimensions – let’s find out if we can.

Buckle up, here we go.

Here’s my first prompt:

Remember, LLM’s work best when you give really detailed prompts. Note the following:

  1. I began by giving some information about myself – my limitations as regards economics, where in the world I come from, and what my interests/hobbies/passions are.
  2. I specified what I’m looking to learn from the LLM.
  3. I specified the quantum of output required (thirty classes).
  4. I specified how broad the output should be.
  5. I specified how I would like the answer to be customized for me
    • I would like to learn about economics by relating it to what I like to read about in any case (use examples from the Mahabharata)
    • I would like to learn about economics by relating it to real life situations.
    • It is amazing to me, regardless of how many times I experience it, that it “gets” what I really mean in spite of having phrased my question using really bad grammar.
    • The specific examples aren’t the point, the idea is the point. Learn calculus by relating it to mandala art, for example. Learn history by relating it to dance forms. Learn geography by relating it to food from different parts of the world. A teacher in a classroom cannot possibly do this for all the students, because of the size of the class, and because a teacher cannot possibly know your hobby in as much detail as you can. Make good use of AI!
  6. Should the examples from the Mahabharata be chosen for how prominent the examples were in the text, or should they be chosen for their relevance to economics? My preference is for the latter, and I made sure the LLM knows this. Ditto for the real life examples.
  7. I ended with a meta-prompt, that will stay true for the next thirty (or more questions) – ask if I need to learn more, and only then proceed with the next class.

Should you copy this prompt, word for word? Of course not! For one, you may not want to learn economics, but rather a different subject. The underlying principles still holds. You may not like to read about the Mahabharata, for another. You may want only ten lectures, not thirty. Or you may want two hundred! Feel free to tweak the prompt to suit your requirements, but it helps to “get” how to go about thinking about the structure of the prompts. That’s the point.

I took a look at the outline of the thirty course lecture series it prepared for me, and it was not bad at all. But I had a follow-up request:

Now, you might think that you need to know economics in order to judge the output, and tweak your request. And sure, you’re right that it will help. But regardless, even if you cannot judge the quality of the output, surely you know enough about what and how you want to learn. My apologies for going all meta on you, but if you don’t know enough about the supply side of the market, surely you know what you would like as a consumer – at least in part. So feel free to help the LLM become a better teacher by telling it more about you.

It went ahead and gave me the refined output, and also the broad contours of the first class. Here are the broad contours of the first class:

Again, note that I am quite excited about how this class is shaping up, because if economics is, indeed, the study of how to get the most out of life, Arjuna’s choice to fight in the Kurukshetra war is an awesome way to get some really thought-provoking questions in for discussion. But this may not be your cup of tea – so feel free to brew your own cuppa of econ, by customizing it to what you like the most (Avengers? Cricket? RRR? Bharatnatyam? Junk food? Anime? Go for it!)

I did have follow-up questions:

And based upon its answer to this prompt, I had yet another clarificatory question:

Note that your conversation will be (I would go so far as to say should be) different. You will have different questions, different prompts, different things that make you curious. And that’s not just fine, that is the whole point. Depending on how carefully you read its output, and depending on how probing and detailed your questions are, you can keep just this first class going for a long, long time. How long? That’s up to you!

Here are two examples:

You can, of course, ask it to answer any (or all) of these five questions. Ask it to create ten (or twenty, or a hundred) instead – and as a student, assume that this is how us professors might well be “coming up” with questions for your tests, assignments and exams.

Here are more, and note how they get wilder (more random?) with each passing question:

In each of these cases, you don’t have to have trust in, or agree with, the answer given by the LLM. Treat the output as a way to get you to think more deeply, to challenge what has been said, to verify that the answers are correct, and to have further discussions with your peers and with your (human) teachers, whoever they may be.

Note to myself (and to other teachers of an introductory course about the principles of economics):

  1. How can we do a better job than this in the classroom…
    • Without using AI (we’re substitutes)?
    • By using AI (we’re complements)?
  2. What is missing from the LLM’s output (this is assuming you’ve tried these prompts or their variants)?
  3. What stops us from recommending that students do this in class on their own devices, and we observe, nudge and discuss some of the more interesting output with everybody? That is, how does teaching change in the coming semester?

Feedback is always welcome, but in the case of the next thirty posts, I think it is especially important. So please, do let me know what you think!

The End of the College Submission (Thank God)

This blog post is a riff on Seth’s post from the other day, titled “The End of the High School Essay“:

New York City schools are trying to ban GPT3 because it’s so good at writing superficial essays that it undermines the command structure of the essay as a sorting tool. An easy thing to assign (and a hard thing to grade) just became an easy task to hack.
High school essays had a huge range of problems, and banning the greatest essay device since Danny Dunn and his Homework Machine is not the answer. In fact, it’s a great opportunity to find a better way forward.
The first challenge of the essay was the asymmetrical difficulty in giving useful feedback. 30 essays, 5 minutes each, do the math. It doesn’t scale, and five minutes isn’t even close to enough time to honor the two hours you asked a student to put into the work.

https://seths.blog/2023/01/the-end-of-the-high-school-essay/

Exams in almost all of the colleges and universities I have taught at don’t mean a thing. The students know this, the faculty knows this, the examination department knows this, but we all keep up the charade that Meaningful Work Is Being Done through the conduct of examinations.

Newsflash: there is no meaningful work being done. It is a complete farce.

Some universities choose to not pay faculty members for correcting papers at the end of the semester. Let’s assume a college is paying a visiting faculty member two thousand rupees per hour to teach a class. They might slip in a line towards the end: this also includes examination duties. In English, this means that if you teach a thirty hour course, you will be paid sixty thousand rupees for those thirty hours. So far, so good. But “also includes examination duties” means that for a batch of (say) a hundred and twenty students, you are also expected to design question papers (a set of two, usually) and correct a hundred and twenty answer sheets.

Even if you assume that one is able to correct paper after paper without taking a break, with five minutes being the time taken per paper, that still means that at least ten hours worth of work. Which means, of course, that you are not being paid two thousand rupees per hour, but rather fifteen hundred. Accounting is a subject that may well be taught at universities – that doesn’t necessarily mean that it is practised at universities.

Some other universities offer to pay forty rupees per answer sheet corrected. Which is better than zero, admittedly, but we then run into the problem of incentives. If you’re paid two thousand rupees to teach, and forty rupees per paper to correct answer sheets, how many answer sheets should you correct in an hour to “make” the same wage? And if fifty answer sheets being corrected in an hour is clearly far too many, how do you expect this incentive to work? Or do we teach our students that incentives matter, but ignore this point ourselves?

Students know the farcical nature of examinations all too well. The pandemic took away that last remaining fig leaf of dignity that surrounds examinations, and the ostrich-in-the-sand approach that most universities have adopted post-pandemic is that of closed-book, no-internet-access examinations. Quite how this pen-and-paper examination is supposed to prepare students for what they will do in the real world is a question nobody wants to raise, let alone answer.

And so students quite reasonably ask for “the pattern of the paper”, or the “important questions” or the “important topics” before an examination. They are, in other words, seeking to minimize efforts in order to maximize marks scored in an examination. The tragedy lies in the fact that academia is supposed to be about maximizing learning. But on and on we go, in our mad headlong rush to maximize NAAC scores, difficult and uncomfortable questions about examinations be damned.

But all that these pen-and-paper examinations do is to train students to produce mediocre output that AI can already produce – and of a much better quality than these scribbled answers in answer sheets will ever produce. That’s not a knock against students; it is praise for how good AI has already gotten.

Think about it, for this is a point that bears repetition. Our examination system is geared towards training students to do a worse job than AI, by definition. And for this, we take money from students and their families, and we call it “an education”. Pah.

Now, I’m well aware of the fact that this is not applicable in all cases. There are some subjects/courses in the social sciences where these kind of examinations are entirely justified. And medical and engineering fields is a whole separate story. But I’m not arguing for an extreme solution – I’m saying that the pendulum has swung far too much over into Luddite territory when it comes to examinations and submissions. We need to wake up and smell the AI, and adjust accordingly.

But how? Well, the easy thing to do is to say that’s a difficult answer to give in a blogpost, but here’s Seth Godin again:

The answer is simple but difficult: Switch to the Sal Khan model. Lectures at home, classes are for homework.

When we’re on our own, our job is to watch the best lecture on the topic, on YouTube or at Khan Academy. And in the magic of the live classroom, we do our homework together.

In a school that’s privileged enough to have decent class sizes and devices in the classroom, challenge the students to actually discuss what they’ve read or learned. In real-time, teach them to not only create arguments but to get confident enough to refute them. Not only can the teacher ask a student questions, but groups of students can ask each other questions. Sure, they can use GPT or other tools to formulate where they begin, but the actual work is in figuring out something better than that.
At first, this is harder work for the teacher, but in fact, it’s what teachers actually signed up to do when they become teachers.

This is far less cohesive and controllable than the industrial model of straight rows and boring lectures. It will be a difficult transition indeed. But it’s simple to think about: If we want to train people to take initiative, to question the arguments of others, to do the reading and to create, perhaps the best way to do that is to have them do that.

We’ll never again need to hire someone to write a pretty good press release, a pretty good medical report or a pretty good investor deck. Those are instant, free and the base level of mediocre. The opportunity going forward remains the same: Bringing insight and guts to interesting problems.

https://seths.blog/2023/01/the-end-of-the-high-school-essay/

Kill our current mode of examinations, and help build a world in which we have passionate teachers who help students create. Not a world in which we minimize soul, and maximize those stupid, accursed “marks”.

But on and on we go. Pah.