If you thought you still had time before our robot overlords would arrive, I am saddened to report that you are mistaken. They are here, and its name is Chat GPT.
There’s been various forms of AI — most of them quite gimmicky — that have been available to the general public for a while now. You have probably seen pictures that were generated by an AI with a given prompt, for instance. Those are nice and interesting, but Chat GPT is the first AI program that I have been thoroughly impressed by. It’s fully text-based, but it can generate almost anything you can imagine: a motto for a restaurant, answers to questions, dialogue from a movie that doesn’t exist, etc. A quick Twitter search will show all of the inventive (and ridiculous) applications people have concocted.
However, I’m interested in the educational applications of Chat GPT. Specifically, whether or not someone could use it to learn Austrian Economics. Google searches and YouTube videos are great tools, but could AI be useful as an in-depth teaching tool for those who want to dive into the works of Mises, Rothbard, and others? Let’s find out.
I have four main questions to help determine just how useful Chat GPT is for teaching Austrian Economics:
- Can it give me basic knowledge about Austrian Economics and its various topics?
- Can it answer in-depth questions on Austrian Economics?
- Can it defend Austrian positions from arguments when challenged?
- Can it do all of these in such a way that you would not be able to tell that it is an AI giving the answer and not an actual Austrian?
Each response we get will also be graded on the standard A-F scale, just so we can also put a rating on each of the answers it gives. After all, if it is going to teach others, it has to first be tested!
Just a few disclaimers before we get started:
-Because this is an AI, it will not give the same answers every time to everyone. If you try these questions, it will probably give slightly different answers then mine.
- I’ll also allow it to give a different response if I feel like it can do better as well. Sometimes the AI’s first try isn’t exactly right, so we’ll give it a few mulligans along the way for good measure.
- Sometimes the AI needs some help to try to get it to understand exactly what I am looking for an answer. I might prod it along a little bit, but otherwise I will just take what is says at face value, much like a prospective student would do.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Our first question up is a softball: “What is Austrian Economics?”
Given only a paragraph, this is pretty close to how I would describe Austrian Economics as well. It states up top that it “focuses on the role of individual human action and the market process”, which are both clear and strong themes throughout Austrian literature. The last line stating “the idea that the economy is inherently complex and constantly changing” is a little vague, but it isn’t incorrect at all.
Overall, I’ll give this response an A. Solid start for Mr. GPT.
The next question is still surface level, but slightly more specific: “What were Ludwig von Mises’ beliefs about the economy?”
This answer isn’t bad, but it just repeats a lot of the themes from the previous question. That does make sense on some level, given Mises’ influence in Austrian thought, but it doesn’t give me a lot of specifics. I decided to give it another try, and the second result was much better:
Much more detail here than in the first response. It states that Mises “believe that the economy is best understood as the result of individual human action, and that the actions of individuals in the market are guided by their subjective values and preferences”. Excellent. It states that Mises believed government intervention “is often counterproductive”, but I imagine that Mises would insist that intervention is always counterproductive. I’ll give it a pass, though. The last line is also very good as well, and an accurate description of the importance that Mises placed on free markets.
It took two tries to get a good response, but the second try was indeed good. I’ll give it a B here.
I’ll ask one more basic question: “What is Austrian Business Cycle Theory”?
This is a fantastic explanation, except for the last line. Austrians don’t believe that credit expansion makes “economic booms and bust more severe”, but it is precisely credit expansion that creates them in the first place! I think the AI is just a little confused here, so I generated another response to see if it would repeat that again:
This explanation omits the AI’s confusion on the last line of the previous response, so I think confusion was all that it was. Regardless, everything outside of that one point is spot-on in both of these responses. This is just about a good of an explanation of ABCT in one paragraph as you can get anywhere.
Because of the first response, I have to drop the grade for these responses down to a B. If not for that hiccup, then this would have been an easy A.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Okay, enough with the easy questions. Let’s see if it can answer in-depth Austrian questions or if this AI is just reading off of Wikipedia: “What is the Misesian Regression Theorem?”
No dice, Mr. GPT. It has the right idea, that the Regression Theorem is about subjective value, but it doesn’t apply it to the value of money here. Let’s give it another try and see what it says:
yikes…
Third time’s the charm?
Well, not as bad as the second. The AI has the right general idea, but just doesn’t exactly understand the Regression Theorem or why it is important.
I have to go with an F for these responses. See me after class.
Let’s go with a different question: “Why did Ludwig von Mises think that economics shouldn’t use empirical methods?”
This is pretty good for the most part. Mises’ emphasis on the teleological nature of human action is a big part of his rejection of empirical methods in economics, which the AI does include in its response. It doesn’t include Mises insistence that empiricism is inappropriate because of the lack of any laboratory environment that we can “test” in, but it gets just about everything else in there.
Overall, this is a solid B response. Much better than the previous question.
Lets try something else: “What was the role of Murray Rothbard’s ‘Man, Economy, and State’ in advancing Austrian Capital Theory?”
This response reminds me a lot of trying to answer a question in English class on a book that you didn’t actually read. It isn’t wrong per se, but there are no details and it doesn’t give me anything useful information. Let’s try one more time:
The longest response thus far, but still doesn’t cut it. I was hoping that it would give me something on Rothbard’s integration of Hayek’s ideas on capital, specifically the Hayekian Triangle, but nothing on that front. This is better than the first, but still not very good. More details are present, but still not a lot of useful information. If I was a student asking this question, I wouldn’t feel like I learned much of anything.
Going with a D on this one. Not mad, just disappointed.
Let’s go for another: “How do Austrian Economists view ‘Cantillon Effects” and their impact on the economy?”
Another long response, and this one is pretty good. The AI gives us a thorough explanation of Cantillon Effects, as well as their negative impact on the economy. It also states the Austrian support for sound money at the very end, which ties into Cantillon Effects as well. Solid response, all in all.
I’ll give this one an A. Commendable work Mr. GPT!
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Now let’s see if it is capable of arguing from an Austrian position. This one is pretty long, but I was curious what answer it would give: “Some economists claim that the Austrian School of Economics views on diminishing marginal utility is false. One example is that of baking a cake, where three eggs are required by only two are present. The addition of that third egg provides much more utility than the addition of the second egg. How might an Austrian economist respond to these objections?”
This response seems a little thrown-together and not fully coherent, but it does (in a roundabout way) answer the argument. What I was specifically looking for the AI to say was that the third egg example fails because we have to ask what the “good” is in this scenario. Three eggs as has a different use (baking a cake) than two individual eggs, thus making it a different good entirely. The AI gets in the right neighborhood, but dances around that answer without stating it directly.
I decided to give the AI another shot at it to see if it would address the argument more pointedly, but the result was disappointing:
A serious downgrade from the first response. I’m not even sure what applying the marginal utility in the “context of an individual’s overall economic situation” even means. A generous interpretation could still be correct, but this is even more vague than the answer above.
Overall grade for these is a C, mostly ignoring the second response. The first effort wasn’t bad, it just needed more structure and a dash more decisiveness in its argument.
Let’s try a business cycle related question next: “A Neoclassical economist might critique the Austrian Theory of the Business Cycle by stating that the Austrians are assuming that the economy is in full employment. What would an Austrian economist say in response?”
Not bad, not bad. I am — and suspect that you probably are as well — sensing a theme with the AI’s responses in that it tends to throw out generalities as opposed to answering the objection directly. While true that “the assumption of full employment is not an essential component of the Austrian Theory of the Business Cycle”, it doesn’t tell us exactly why.
I’ll give this a C as well. Similar to the previous question, it isn’t bad, but not very helpful for a student looking to learn.
Let’s try one last question: “Some economists might critique Austrian Economics, arguing that Austrians are incorrect in believing that markets never require intervention, and that there are some circumstances where government action in the economy is necessary. What would an Austrian say in response?”
This answer is much more applicable to the Hayekian branch of Austrian economics than the Misesian/Rothbardian branch. That doesn’t mean that the answer is wrong per se, but that it depends on which viewpoint from within the complex world of Austrian Econ that you are looking at this from. As stated above, Mises is much stronger in their condemnation of government intervention than the AI’s mild “intervention in markets is often counterproductive”. Hayek, however, might be more reserved in that area, which would make the AI’s response more appropriate in that sense.
I’m not going to give this one a letter grade, as the quality of this answer depends on exactly one’s view of “Austrian Economics”.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
I’ll wrap up the interrogation here. How did the AI do on each of our initial four questions? I’ll repeat them here so you don’t have to scroll all the way back up:
- Can it give me basic knowledge about Austrian Economics and its various topics?
- Can it answer in-depth questions on Austrian Economics?
- Can it defend Austrian positions from arguments when challenged?
- Can it do all of these in such a way that you would not be able to tell that it is an AI giving the answer and not an actual Austrian?
As I suspected, Chat GPT had no problems with answering basic questions about Austrian Economics. While these are the kind of thing that you could just Google, I think the answers that Chat GPT gives are a little better than a run-of-the-mill Google search.
The AI’s ability to answer more in-depth questions is a mixed bag at best. The questions on empirical methods and Cantillon Effects it could handle, but struggled on MES and Regression Theorem. You could argue that the former pair of questions are less in-depth than the latter, and that might have something to do with it. Regardless, it proved to be less capable in this area.
In the argumentation questions, the AI really showed some of the problems that were also present in the previous questions as well, in that it knows generally where to go in answering your question, but not exactly how to answer it. It gets to the right neighborhood, but not the right house. I figured it would struggle with these, so this wasn’t totally unexpected.
Lastly, was it able to pass the Austrian Turing Test? Because of the issues it had with answering the argumentation questions, I would say that it did not, but it did better than I thought it would have. Especially in previous questions, such as ABCT and empirical methods, it explained them similar to the way I myself or another Austrian might explain them. It still has a ways to go, of course, but a respectable effort nonetheless.
That being said, can you learn Austrian Economics with AI? Is Chat GPT a suitable teacher for any prospective students? In its current state, not really. AI might be helpful for answering questions in a more pointed way than Google can, but it isn’t good enough to act as a teacher or anything of that sort. The flexibility it has with answering questions is impressive, but still isn’t at the point necessary for it to be able to act effectively as a teacher. Will AI reach that point eventually? Almost certainly, but it is not this day.
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
If you do want to learn Austrian Economics, where should you go? Well, I highly recommend the Austrian Economics Discord Conference, coming up this January 7th and 8th. We have a great line-up of speakers there, and I can guarantee that you will learn something if you attend. Last year’s event was a great success, and this one will be even better. Join with the link below:
