We all know that AI models are becoming increasingly commonplace, with hundreds of software companies now offering AI-powered tools, and dozens of iterations of ChatGPT and similar LLMs popping up every year. While some bemoan AI tools, they’ve also been lauded as powerful and helpful tools that we ought to be using every day.

Regardless of where you stand on the AI-usage spectrum, one thing is certain: AI models are reshaping education, even at the highest levels. Economics is no exception. Even for PhD students like me, AI tools are becoming a feature of student life — and when used smartly, can be a great asset to learning.
 

Below I present three main reasons why it is becoming increasingly difficult to imagine postgraduate education without the use of AI. Whether you’re a student or a lecturer, knowing how to use AI effectively — or just being aware of how students are already making use of it — will help you succeed.

1) AI offers better searches for niche information

Many of the common AI models can be used as a search tool, making them the first truly new innovation enabling Internet searches since the emergence of sites like Yahoo! and Ask.com - and of course google - in the second half of the 1990s. This search feature makes it a great study and research partner, beyond what a typical user can accomplish with a search engine.

One of the weaknesses of traditional search engines is that it can be hard to find the specific information you need, especially if you’re searching for something niche, or if the language used for that topic is generic. For example, consider a scenario where your favorite bakery is a small mom-and-pop shop named “Croissant”. Or, for a more economics-relevant example, consider searching for content about the inverse elasticity of substitution in (specifically) the neoclassical growth model.

In both cases, it can be extremely difficult to find the right information you need. In the first case it’s hard to craft a good search with such limited language, since “croissant” will return far too many irrelevant results. And in the latter case, the need to be extremely specific (thus using so many words) means that a traditional search will end up getting lots of irrelevant content as well. Even using advanced search functions might be ineffective, or at least time-consuming.

But AI has the ability to sidestep a lot of these problems. That’s partly because if an LLM answers with something you weren’t looking for, you can just tell it that, and it will pivot to your intended meaning instead. This makes LLMs ideal tools to begin a search for esoteric information — although the ever-present chance of hallucination must be kept in mind (more on that later!).

2) AI models are great as study tools

AI models can go far beyond being search engines — they can explain their output and interact with the reader. This is where one of their core use cases lies: as a study tool, or even a supplementary TA.

AI models are frequently used by economics postgraduate students, with some using them on a daily basis to aid studies. AI can parse even the most complicated equations and break down complicated or long-winded paragraphs into more digestible bits. This helps students learn new concepts faster than if they’d spent time puzzling over arcane lecture notes, reading a textbook, finding a helpful YouTube video, or waiting to attend office hours. If someone is puzzled over how to get started with a difficult problem, asking an AI model can help them know where to begin.

Still, there are limitations. There’s no shortcut to true understanding, and sometimes AI output isn’t sufficient. And of course, LLMs suffer from the hallucination problem, so AI output must be treated with suspicion and tested rigorously to ensure it’s accurate. In the context of an economics course, though, ensuring output is accurate can be a great way to learn!

In my personal experience as a PhD student, fellow economics postgrads have agreed that AI is best used to either help with starting a difficult problem or help understand the steps taken in a sparsely-explained proof, problem, or model setup. Often, research papers, textbooks and lecture notes skip several important steps at a time when deriving or showing a concept, much to the ire of students. AI can help bridge the gap by explaining the omitted steps. This is something that AI models have grown quite adept at — given a specific enough prompt. In this way, AI takes on the role of a miniature TA, available at any time a student may be studying.

Further, AI can even be used to generate practice problems and mock exam questions, making them good tools to ensure understanding. But these aren’t the only things AI can be helpful for. They can also help economics students parse long stretches of code…

3) AI is an ideal partner for coding

Economics is an increasingly broad and cross-disciplinary field. As such, knowledge of mathematics and social sciences are important, but increasingly, data science competence and coding skills are becoming invaluable for economists. This is where LLMs are particularly useful, especially for postgraduate students who don’t have as much practical experience yet.

One of the first things that a professor recommended to me during my economics PhD was to use AI to help me write code. Economists are increasingly required to use statistical software programs to do their jobs and their research, and even limited coding skills go a very long way.

Even so, the vast majority of economics courses don’t teach software skills — often, there’s only time for an introductory course into one program, and students end up having to learn the rest on their own. This is a gap in the economists’ education that, for better or worse, degree programs often don’t have enough time to cover.

Nobody should rely fully upon AI to code everything for them, and my professor wasn’t telling us to do so. Rather, as coding is an important but secondary skill for an economist, my professor recommended that we use AI to help us learn while doing. And, as he pointed out, using AI in this manner makes managing the hallucination problem quite easy: simply test the code the AI gave you. If it works, great! If not, clearly the AI made a mistake.

Of course, this process works best for small pieces of code — asking the AI to change the color of a graph, or create a formula to update a matrix, for example. Asking it to write an entire research paper, or compile code to clean a whole database, is a recipe for disaster (and, as I argued before in an article justifying why I did not use AI for my PhD application, is only robbing yourself of the ability to learn).

Conclusion: AI can be great, but remember: it’s only a tool

There are, of course, pitfalls to avoid with AI. Relying on its output too much is a mistake, and students who over-use it may fail to learn as much as they thought. This is important because economics exams are typically held in person, without the aid of computational devices (outside of perhaps a small calculator).

Thus, when grades are on the line, AI often won’t be available. For this reason alone, it’s imperative that students learn course content and how to think about and solve economics problems without relying on AI. In fact, the advent of AI makes it even more likely that, at least in the short run, economics degree programs will continue to rely on in-person exams to assess students.

Yet aspiring economists shouldn’t fear this. Any economist worth their salt will know their field, and won’t need to rely on AI to do their job. Yet AI tools can — as with so many other industries — greatly increase the productivity of an already capable economist.

Likewise, economics instructors shouldn’t be afraid of AI tools. Embracing them and training students to use them properly can actually increase engagement and learning. Even if you as a lecturer don’t feel confident teaching how to use AI, it’s imperative that you are aware of AI’s capabilities. This applies when designing any assessments, projects, or homework that isn’t done in person, as students could make use of AI while doing them.

How has AI been used in your economics degree program? Have you experienced professors or students recommending it, or actively incorporating it into the syllabus? Share your experiences in the comments below!

---

_{^{Header Image Credits: halfpoint, via Canva Pro License / Daniil Komov from Pexels, via Canva Pro License / metamorworks from Getty Images, via Canva Pro License}}