The AI tool has been circulating ever since news about ChatGPT-3.5 encouraged people to create OpenAI accounts in early 2023. I remember sitting in a coffee shop, sipping a latte and holding a half-dozen “conversations” with newly released models.
I asked a series of mathematics questions. I asked about politics and poetry. I asked him to write a short story. I asked about niche topics like Catan Settlers and checked whether I “understand” the strategy in the context of the game. And of course, I asked about economics.
However, as a member of INOMICS editorial board, I found that AI tools lacked the ability at the time to write thorough, detailed, engaging articles on economic topics such as the pros and cons of corruption focusing on economic growth and the “mesoeconomics.” I have decided not to use AI tools in the article creation or editing process.
Similarly, I decided not to use AI tools at all when I was applying for a PHD program. That’s not to think that these models will not be useful. I found the AILLM model to be extremely useful in particular in one thing: generating ideas. However, I am seriously considering becoming a PhD economist and have decided to submit the best of my writing for my application. Furthermore, there is always concern that using AI tools is considered a form of plagiarism.
I think it was the right choice to have no AI tools in my application. I think that would be the right choice for you too. Furthermore, even if the AI model becomes a highly capable “economist” in itself, I think avoiding them is the right choice in your application process.
Don’t struggle any more, explore the reasons.
First: Hallucinations delay you from building your own expertise by adding extra work and relying on AI
AI models are known to have hallucinations. So, checking the AI output for accuracy is very important when using them for important purposes, but that adds work that I didn’t think is worth it for my application. I felt that the effort required to study each aspect of the output of the AI model (to ensure it is accurate) was inefficient.
Instead, I spent my time reading economics literature myself, seeing what interests me, and reading about how other economists in the universe constructed their research. This was extremely helpful, especially as I began to outline my doctoral research proposals.
Think of it this way: If you want to be a successful doctoral student, you need to understand most of your subfield literature anyway. You can also start during the application process! This will greatly help you with research suggestions you have to write and the conversations you have with potential supervisors (or admission interviews!).
Instead of getting the potentially incorrect output of the AI model and starting from there, I searched Google Scholar and searched for Repec. I read the literature to see if there were any gaps that my ideas could fulfill. I have put together the methodologies I knew I knew, using econometric methods that I was sure I understood. I also found a public dataset that I knew was accessible and relevant to make sure that realistically usable data was available and demonstrated in my application. Finally, I wrote my ideas for future research and studied the research interests of faculty members at the institution I was applying to.
Sometimes doing “hardway” things is really the best in the long run. After all, there’s no free lunch.
Second: Personal essays require personal touch
Some schools have topics that you need specific essay questions or you would like them to broach in a personal statement. For this reason, it is highly recommended to reaffirm the expectations of each institution for personal statements (if this is, it is usually found or referenced in the online portal of the school that uploads the statement). Each school had slightly different needs, so several different personal statements had to be written.
Obviously, these statements require a personal touch. If you outsource your work to a machine, the admissions committee is preventing you from hearing your voice and preventing you from reading your story as you tell it. This clashes with the overall point of a personal statement. Your application must show the Admissions Committee who you are and why you want to study a PhD in Economics.
This may sound like a “fluffy” or stupid concern, but the truth is that admissions committees receive many applications from highly qualified students. Sometimes you can decide who the committee will provide for less measurable factors, such as how unsuccessful it is to succeed in their own facility or to fit the faculty’s research interests. A good personal statement will help you convince them that you are a good candidate and pursue a PhD for the right reasons.
Apart from this, many admissions officers can either identify works written by AI, or at least claim that they can. Whether it’s overuse of stock phrases, inconsistent registration, or “generality” of style or discussion, if ultimately an admissions officer or potential supervisor suspects that the submission is written, at best treat your application with skepticism, or at worst throw it away.
Overall, you should give your admissions team a greater sense of who you are as much as possible. Express your personality and show them your enthusiasm. ai can’t do it, not just you.
Third: AI output is not your job, this limits your growth
Various people may find that using AI tools different, or to what extent, constitutes plagiarism. However, no matter which camp you personally fall into, the following statement is true: The AI model produces output that did not write itself. And if you’re aiming to be the best economist you can be, this is a problem.
This relates to the first point regarding the issue of hallucination, but deeper. AI tools provide cognitive shortcuts. However, if you are constantly taking shortcuts, starting an idea and starting work is depriving you of your ability to learn how to successfully start the research process. I realized this trend in myself long before the popular AI models emerged, so I know this is likely to be true to me. Let me explain.
When I studied mathematics in graduate school, I often felt confident that my teacher understood the example of a teacher when he was working in class. But when they ask me to leave the chalkboard and solve similar problems myself, my mind will often be blank. Suddenly, it was much more difficult to get back the correct procedure and understand exactly what it meant.
I think this is an accessible experience for many people. I think relying on AI tools is the same as relying on a teacher to start solving math problems. It seems harmless when you are still using the help of a teacher. Limitations or integrals – or the beginning of the econometric theorem – all seem very simple when your teacher sets them for you.
But once aid is removed, it can suddenly lead to panic because our brains aren’t actually learning how to start properly. Remember, your brain is an expert in finding shortcuts, and that means you can’t learn the way you think you did!
Apart from this, it has been shown that excessive reliance on AI tools can potentially worsen, which can exacerbate hallucination detection and allow people to stop thinking about serious issues like plagiarism.
As Zhai, Wibowo and Li included in their 2023 paper1:
“Despite the undeniable benefits of AI dialogue systems in streamlining research processes and improving academic efficiency, our analysis reveals the trend of potential erosion of critical cognitive skills due to ethical challenges such as misinformation, algorithm bias, plagiarism, privacy invasion, and transparency issues.”
Therefore, I am confident that by overrely relying on AI tools, for example, to start a review of your research proposal literature, you will limit your growth as an economist, even if you are not aware of it.
If it doesn’t take long to actually read and process economics papers yourself, you’re missing out on a lot of knowledge. Even simple meta-knowledge about what exists in the body of economics literature is important to your career. It is embarrassing to not know famous or original papers in economics, and if you don’t know literature in your own niche subfield, it certainly won’t look good.
Fourth: AI cannot output to the same quality as you or me
Honestly, the idea of using AI in my PhD applications never really crossed my mind. That’s because through Inomics writing and editing, we continuously test new AI models to see how good it is to write thoughtful economics articles.
I have not yet found an AI model that allows me to write useful, informative, analytical articles, as I can, like Hannah Safno, Dr. Sahar Milani, Dr. Tom Mackenzie, Dr. Gomez Pineda, or I can. The AI model has yet to demonstrate its ability to present general discourse on topics and teach readers something new or present new light.
We also know that if the AI model has posted to the INOMICS website but has not met the standards, then we have not met the admissions committee standards for essays and research proposals for PHD applications.
Those who are seriously interested in pursuing a PhD in Economics are sure to be far brighter than today’s most advanced LLM and neural networks. Perhaps it will change the day in the future, but even if that is the case, working on your own will develop you into a much better economist than relying on AI output.
Conclusion
I learned a lot through the process of applying for a doctorate in economics. I gained a solid, preliminary understanding of topical areas that are extremely useful when I began writing my first published paper. Furthermore, after examining literature myself, I am confident that the research I want to pursue will advance the field in ways that have not been done before. I don’t think AI has a place in that process.
And at the end of the day, my hard work landed me a very generous offer of admission to the wonderful university I am excited about.
So, my fellow economist hoper, take it from me: don’t rely on AI models to write your application for you. You owe yourself to put effort into it. After all, there is no free lunch in economics or life.
References:
1: Zhai, C., Wibowo, S. & Li, Ld. The effect of overreliance on AI dialogue systems on student cognitive abilities: a systematic review. Smart Learn. environment. 11, 28 (2024). https://doi.org/10.1186/S40561-024-00316-7
Notes on AI use with INOMICS
Here at INOMICS, we are not opposed to the use of new technologies. On the contrary, our website has useful AI tools, such as the ability to generate questions (and answers) for AZ articles, or the AI Summary button on the article page. However, quality and accuracy are top priority when it comes to writing articles and resources.
So, while I sometimes use AI tools to brainstorm ideas, suggest phrasing, translate existing text, use articles, and generate header images that are interesting to me, when I created longer articles using AI, I found that there is always a shortage, not to mention sharing the many concerns expressed in this article. So I have not published partially or fully written advice articles, blog posts, or economics terminology/lectors (yet) using AI.
Image credits: Inomics, Using Canva Magic Studio AI Image Generator
