When AI Classifies: What Counts as Public Administration?
This study demonstrates that different scholarly representation systems, including AI-assisted methods, produce fundamentally distinct and non-overlapping classifications of public administration and AI-related research, revealing that algorithmic knowledge organization is a non-neutral, interpretative force that shapes disciplinary boundaries and underscores the enduring necessity of human judgment.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are trying to find every book ever written about "Public Administration" (how governments and public services work) and specifically those that talk about "Artificial Intelligence" (AI).
This paper asks a simple but tricky question: Does it matter how you look for these books?
The authors, Shaoming Cheng and Laurie Schintler, decided to test this by using two different "libraries" (databases) to find the same topic. They wanted to see if the tools we use to organize knowledge actually change what knowledge we find.
Here is the story of their discovery, explained with some everyday analogies.
The Two Librarians: The Human vs. The Robot
To understand the study, imagine two different librarians trying to build a reading list for you.
- Librarian A (Web of Science): This librarian is a traditionalist. They rely on a strict, human-made filing system. If a book is published in a journal called "Public Administration," it goes in that bin. If it's in a "Computer Science" journal, it goes in a different bin. They trust the labels the authors and publishers put on the books.
- Librarian B (OpenAlex): This librarian is a high-tech robot. Instead of looking at the book's label, the robot reads the entire text of millions of books, looks at who cited whom, and uses AI to guess what the book is about. It creates its own categories based on patterns it finds in the text.
The Experiment: Finding "AI in Public Administration"
The researchers asked both librarians to find all the writing about AI in Public Administration. They expected the robot librarian (OpenAlex) to find more books because it scans a much larger, open library that includes drafts, student papers, and conference notes that the traditional librarian misses.
The Surprise Result:
The traditional librarian (Web of Science) actually found more relevant books about AI in Public Administration than the robot librarian did.
Why? Because the robot librarian got confused by the "interdisciplinary" nature of the topic.
- The Human Librarian looked at the journal title. If it said "Public Administration," it counted it.
- The Robot Librarian looked at the content. It saw a paper about AI in government, but because the paper used heavy computer science jargon, the robot decided, "Ah, this is clearly a Computer Science paper," and filed it there. It didn't count it as "Public Administration" at all.
The "Ghost Town" of Overlap
The most shocking part of the study was the lack of overlap.
- Imagine the Human Librarian gives you a list of 100 books.
- The Robot Librarian gives you a list of 50 books.
- Only 2 books appear on both lists.
It's as if they were searching for the same treasure, but one found a chest of gold coins, and the other found a chest of silver coins, and they were in completely different rooms. They weren't just finding different amounts of the same thing; they were finding completely different things.
The "Time Travel" Effect
The two librarians also told different stories about when this research started:
- The Robot's Story: It found papers going back to 1980. It suggested that AI in government has been slowly growing for decades.
- The Human's Story: It found that almost all the relevant papers were written very recently (since 2023). It suggested that AI in government is a brand-new, exploding trend.
The robot saw the "slow diffusion" of ideas across many fields, while the human saw the "explosive moment" when the Public Administration community finally started talking about it openly.
The "Outfit" Analogy
Think of a scholar writing a paper on AI in government.
- If they publish in a Public Administration journal, the Human Librarian sees them as a "Public Administrator."
- If they publish in a Social Work journal (because they are helping social workers use AI), the Robot Librarian sees them as a "Social Worker."
- If they publish in a Computer Science conference, the Robot Librarian sees them as a "Programmer."
The paper argues that who you are (as a field of study) depends entirely on who is looking at you and what tool they are using to look.
The Big Takeaway
The main point of this paper is that AI is not a neutral tool.
When we use AI to organize our knowledge (like sorting books or papers), we aren't just finding facts; we are creating reality.
- If an AI system decides that "AI in Government" belongs in the "Computer Science" bin, then "Public Administration" scholars might never see it.
- If they never see it, they won't cite it, teach it, or fund it.
- Eventually, the field of Public Administration might actually stop growing in that area, not because the research isn't there, but because the robot librarian hid it.
In short: The paper warns us that we cannot just let robots decide what counts as a specific field of study. We need human experts to look over the robot's shoulder and say, "Wait, that paper about AI in social work is actually part of Public Administration too." Without that human judgment, we risk losing the very connections that make our knowledge rich and interconnected.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.