[THE VIEW]What Matters Most in Artificial Intelligence

Rather Than Expecting AI to Provide the Right Answers,
We Should See It as a Tool for Human Decision-Making

With the rapid advancement of artificial intelligence (AI) technology, AI has become an essential element rather than an option in business models. Amid the flood of overwhelming data, service providers have been able to conduct more detailed and sophisticated user analyses through AI technology, enabling them to offer high-performance, customized services. As a result, consumers benefit from significantly reduced information search costs and more convenient access to information.

On the surface, AI may appear to play only a positive role for both service providers and consumers. However, current AI technology outputs results based on input data but operates as a black box, where the underlying mechanisms and processing steps are unknown. Due to this characteristic, different outputs may be generated even with the same data input, and interpreting the results becomes challenging.

To overcome these limitations, the most important and often overlooked aspects are “explainable AI (XAI: eXplainable AI)” and “AI transparency.” This means not simply accepting black-box outputs but understanding how data is processed and providing grounds for humans to comprehend and judge the results. Recently, the U.S. Federal Trade Commission (FTC) and the National Institute of Standards and Technology (NIST) have proposed legislation establishing rules on the transparency of training data, and regulations for the “proper” use of AI are continuously expanding worldwide.

Accordingly, research on explainable AI has been actively progressing in various directions, but there still appears to be a gap from the definitive answer?if such an answer exists. Moreover, even if AI technology’s outputs are explainable, endless debates are necessary regarding the evaluation of the accuracy of those explanations and the precise definition of an accurate explanation.

What I want to emphasize is that rather than expecting current AI to provide the correct answer or the right judgment, it is necessary to view AI as a tool that facilitates human decision-making. Although there may be differences across fields such as finance, healthcare, and education, the opinions of analysts or experts should still be critically considered in final decisions. Explainable AI holds value not as a technology close to the correct answer but as a means to help humans interpret situations and make appropriate judgments. This implies that human and AI technology collaboration can contribute to producing excellent outcomes.

The possibility of AI technology regressing or disappearing is very low; rather, I believe there will be even more rapid development. However, AI technology cannot be the solution to all problems. A comprehensive understanding of AI technology itself, an in-depth understanding of appropriate utilization methods, and continuous and systematic efforts for the “proper” use of AI technology are necessary.

Yunseok Son, Professor at the University of Notre Dame, USA

This content was produced with the assistance of AI translation services.