Core Summary of Global AI Ethics Guidelines: A Complete Guide to Research Ethics Controversies in the AI Era

Why do we need to talk about AI research ethics right now?

Today, with a cup of freshly brewed coffee by my side, there is a serious yet comfortable conversation I want to share with you all. It is about the essential survival knowledge in the AI era brought upon us by generative AI like ChatGPT and Gemini: research ethics.

Many of you probably use AI tools these days when writing papers or analyzing data, right? I, too, receive unimaginably great help when reviewing vast amounts of literature or writing complex system code. However, behind this overwhelming convenience hides a very dark shadow. The boundary between what is a human researcher's unique achievement and what is an AI's creation is breaking down.

Recently, unfortunate news has been frequently heard in global academia. There is a surge of cases where entire papers are retracted because researchers blindly cited fake data or non-existent fake papers plausibly fabricated by AI, believing them to be real. Furthermore, instances of researchers facing disciplinary action for violating research ethics after using models trained without permission on data entangled in others' intellectual property rights are increasing. It is a precarious time where years of effort poured in day and night by a researcher can vanish into thin air due to a momentary mistake.

Because of this, we absolutely must familiarize ourselves with the standards being set by governments and authoritative international organizations around the world, known as global AI ethics guidelines. Today, I will break down those hundreds of pages of complex and rigid legal documents and explain them to you very easily and clearly.

How is the world trying to control AI right now?

The first things to look at are the European Union's EU AI Act, which serves as a textbook for AI regulation worldwide, and UNESCO's Recommendation on the Ethics of Artificial Intelligence, which presents comprehensive ethical standards. Both carry strong legal and declarative meanings, but the message they send to us conducting research in the field is very clear.

1. European Union (EU) AI Act: Strict Management Based on Risk Levels

The core of the EU AI Act, which recently passed the European Parliament and drew global attention, is simple. It aims to thoroughly manage artificial intelligence by dividing the risk it can pose to humans and society into four levels: unacceptable risk, high risk, limited risk, and minimal risk.

The part we researchers need to be most tense about and pay attention to is the high-risk level. If you are researching an AI model that analyzes medical data to predict diseases or handles human biometric and behavioral information, there is a very high probability that it will be classified as a high-risk system. In this case, a legal obligation arises to transparently prove that the data used is not biased toward a specific race or gender, and that a human can explicitly trace back and explain the process of the conclusion drawn by the AI.

Also, when researching technologies that interact directly with humans, like chatbots or deepfake technology, you must clearly inform the experiment participants that they are currently communicating with an AI. The word we must absolutely engrave in our minds here is simply transparency. Revealing the entire process without hiding anything is the most important core demanded by the EU.

2. UNESCO AI Ethics Recommendation: Human-Centeredness and Respect for Diversity

On the other hand, UNESCO's guidelines take a slightly more philosophical and macroscopic perspective. They establish a grand principle that the dazzling development of technology must never flow in a direction that infringes upon fundamental human rights or destroys the environment.

How about applying this principle to the actual research process? It means we must constantly ask ourselves whether discriminatory views toward specific regions or social minorities are mixed into the datasets we collect and refine. UNESCO raises its voice that AI technology should contribute to the public good and peace of all humanity. Therefore, we must keep in mind that developing algorithms where someone's rights are ignored merely in pursuit of simple performance improvement or cost efficiency is an action that directly violates global research ethics.

Three Major Ethical Controversies Occurring in Actual Research Fields

So, how are these seemingly grandiose guidelines applied to the everyday lives of our researchers? I have summarized three actual issues that are currently clashing most hotly in global academia.

First, can AI proudly put its name up as a co-author of a paper?

This is probably the part you are most curious about. To give you the refreshing conclusion first, the world's most prestigious academic journals, including Nature and Science, strictly deny AI authorship.

The position of an author of a paper should not merely belong to someone who writes well, but to an entity capable of bearing full moral and legal responsibility for the research results. However, artificial intelligence is not a personhood that can take responsibility for itself when a fatal error occurs. Instead, if AI was used as an auxiliary tool for writing a paper or analyzing data, it has now become mandatory, not optional, to transparently disclose what AI model was used and what prompts were inputted to obtain what results in the Methods or Acknowledgements section of the paper.

Second, the Trap of Data Copyright and Unauthorized Crawling

To train a high-performing AI model, an enormous amount of data is required. However, just because information is scattered openly on the internet for anyone to see does not mean researchers can indiscriminately scrape (crawl) and use it for training.

Recently, AI models trained with data collected without authorization from original authors or companies are successively getting swept up in massive copyright lawsuits. Even if it is pure research for academic purposes, clearly recording the exact source of the data and meticulously checking whether the license of that data allows for external AI training has now become the absolute basics of research.

Third, Hallucination Phenomena and Undermining Research Integrity

AI does not honestly admit when it does not know something; instead, it sometimes concocts very plausible lies. This is called a hallucination phenomenon. As mentioned earlier, it disguises non-existent fake papers as references or minutely manipulates simulation figures.

If a researcher pressed for a paper deadline rashly includes this without cross-verification, it can be considered serious research manipulation, potentially leading to permanent expulsion from academia. No matter how excellent and smooth a draft the AI writes for you, the entity making the judgment to conduct the final fact-check and stamp the result must absolutely be a meticulous human researcher.

Comparative Summary Table of Global AI Guidelines

To help your intuitive understanding, I have summarized the contents explained earlier into a table so you can grasp them at a glance.

In Closing: May You Become a Researcher Who Fully Masters the Tools

So far, we have thoroughly reviewed the core of global AI ethics guidelines and the controversies currently unfolding in academia. The content might have been a bit long and heavy, but since these are essential issues that any researcher living in the AI era must seriously ponder at least once, I tried to explain them as detailed and easily as possible.

Artificial intelligence is an excellent assistant that will explosively boost our research speed. However, we must never forget the fact that the person tightly holding the steering wheel of the car called research is ultimately ourselves. The ability to sharply verify the flashy results generated by AI without uncritically accepting them—I believe that is precisely the most competitive and respected essential virtue of a researcher in the future academia that is approaching.

I sincerely cheer that the stories we shared today with a cup of coffee will serve as a safe and reliable compass, however small, on your long and arduous research journey.

Thank you for reading this long post.