TO HOME

What Is Human-Centered AI?

인간 중심 AI란? ― What Is Human-Centered AI?Why the UN and UNESCO Have Begun Talking About “Human Dignity” in the Age of AI ― Until just a few years ago, AI had a fairly simple image.It was seen as a “convenient technology” that could translate, generate images, or automate work.But as generative AI has advanced at extraordinary speed, the atmosphere has changed completely. AI is no longer understood as merely an issue for the IT industry. It is now increasingly seen as something that could affect democracy, international security, human dignity, and even world peace.In some ways, stories we once saw only in movies are slowly entering reality.In the film Her, AI becomes deeply involved in human emotions and relationships. In Mission: Impossible – Dead Reckoning, a superintelligent AI appears as a force capable of manipulating entire information systems and shaking the international order.These once felt like science fiction. But today, the international community is discussing these issues in very real terms.In fact, the United Nations and other international organizations no longer see AI simply as a technological innovation. They have begun to view it as a “civilizational transition” that could reshape the international order, much like nuclear weapons, the climate crisis, and the internet.This is why the UN has recently become actively involved in discussions on AI ethics and international norms.One particularly important shift is that the central question in AI discussions is moving from “How intelligent is it?” to “Who does it exist for?” ― “Human-Centered AI”The Phrase the International Community Is Using Most Often Today ― One of the most frequently used expressions in today’s international discussions on AI governance is “Human-Centered AI.”At first glance, it may sound obvious.But its actual meaning is quite powerful.The core idea is this:Human values must come before technology.No matter how efficient or intelligent AI becomes, it must not completely replace human freedom, responsibility, and judgment.It also means that human beings cannot be reduced to mere data points or algorithmic scores.One of the documents that most systematically sets out this philosophy is the UNESCO Recommendation on the Ethics of Artificial Intelligence.Adopted in 2021 by all 193 UNESCO Member States, it is the world’s first comprehensive international standard on AI ethics.What is particularly interesting is that the UNESCO document emphasizes terms such as human dignity, ethics, inclusion, and cultural diversity far more than technical performance.Why is that?Because the international community now sees AI as something that goes beyond a “convenient technology” and has begun to reshape the very structure of human society. Three Major Problems Caused by AI 1. The Weakening of Autonomous ChoiceYouTube and social media algorithms can shape what users see and consume, creating the risk that they distort citizens’ sense of making choices for themselves and control the information environment. 2. The Learning and Reproduction of Social BiasAccording to MIT’s Gender Shades Project, AI can learn from data that reflects the biases of human society and produce discriminatory outcomes based on race and gender. 3. The Risk of Digital Surveillance and Manipulationeal-time biometric surveillance and the spread of disinformation through deepfakes threaten the very foundation of democracy: citizens’ trust. If you would like to learn more about digital privacy, please refer to the following article. What is Digital Privacy? ― AI Is Far More Political Than We ThinkHow Can One Recommendation Algorithm Change the Mood of Society? ― Many people think of AI as a neutral technology.But in reality, depending on what AI learns and what it prioritizes, it can influence not only individual behavior but also the broader mood of society.A representative example is the recommendation algorithms used on social media.There are many analyses suggesting that algorithms on platforms such as YouTube, TikTok, and Instagram tend to recommend increasingly stimulating content in order to keep people’s attention for longer.MIT professor Joy Buolamwini once said:“AI systems are not neutral.”“AI systems are never neutral.”This statement is extremely important.Technology does not become fair on its own.It depends on the values on which it is designed.The problem is that this can intensify social division and extremism. In fact, several studies have found that content provoking anger and hatred tends to spread more rapidly.AI is no longer simply a technology that “analyzes user preferences.”It has become a structure that can influence what people become angry about  and what they come to believe.This is also why the international community has begun to see AI as a question of peace. If you would like to learn more about algorithmic information bias,please refer to the following article. What is a Filter Bubble? ― Why Are Deepfakes Dangerous?War May Begin with Information Before It Begins with Guns ― (UN_Global Digital Compact)In the past, there were limits to how far false information could be fabricated.But generative AI can now imitate the voices and faces of political leaders with near-perfect accuracy.In fact, in several countries recently, AI-generated fake political videos and audio clips have spread during election periods, causing major controversy.It is becoming increasingly difficult for ordinary citizens to distinguish what is real from what is fake.This may sound somewhat extreme, but imagine this:What if, one day, a video of a national leader “declaring war” spread instantly across social media?Even if it were later revealed to be fake, markets and society could already have fallen into serious chaos.That is why the United Nations recently presented, through the Global Digital Compact, a shared international goal of building a “safe, inclusive, and human-centered digital future.” ― Why Has “Human Dignity” Suddenly Become So Important?Because Human Beings Are Not Algorithmic Scores ― One of the international community’s greatest concerns in the age of AI is the “datafication of human beings.”For example, what happens if an AI hiring system automatically disadvantages applicants of a certain gender or race?In fact, there have already been controversies overseas over AI recruitment algorithms allegedly rating female applicants lower. This is because AI learns from past dataand in doing so, it can also learn the biases that already exist in society.There are other examples as well.In some countries, there has been controversy over attempts to analyze citizens’ consumption habits, travel patterns, and online behavior, and manage them through something like a “social credit score.”The danger is that people may come to be seen not as unique human beings, but as data to be managed.UNESCO is deeply cautious about this trend.In its official documents, UNESCO emphasizes that no human being or community should be socially, economically, or culturally subordinated by AI systems. (unesco.org)Ultimately, human-centered AI means not a society where humans adapt themselves to AI, but a society where AI is governed in order to protect human values. ― Why Does UNESCO Talk About “Cultural Diversity”?AI Learns Human Worldviews, Too ―Most AI models today are trained primarily on English-language data, and the global AI industry is largely driven by major U.S. tech companies.The problem is that this is not simply an issue of industrial structure.It ultimately becomes a question of which languages, cultures, and values are more deeply reflected within AI systems.For example, when we ask AI about history, religion, or social issues, if it has repeatedly learned only the perspectives of certain cultural regions, the world’s diverse viewpoints may gradually disappear from digital spaces.That is why UNESCO emphasizes “cultural diversity” as one of the core values of AI ethics.This goes beyond simply protecting culture.It is based on the idea that when different cultures and perspectives are able to coexist, conflicts and clashes can also be reduced in the long term. ― The Core Question in the Age of AI Is ThisWhat Kind of Human Society Do We Want to Build with AI? ― We could move toward a society that pursues only efficiency and automation.Or we could move toward a future that protects human dignity and freedom, even amid rapid technological advancement.The values repeatedly emphasized by UNESCO and the United Nations human dignity, ethics, inclusion, and cultural diversity are not simply idealistic slogans.They are closer to the minimum safeguards needed to maintain a human-centered order of peace in the age of AI.Perhaps the most important competition in the AI era is not a race over technology itself,but a competition over the philosophy of how we choose to understand human beings.Written by Sharon ChoiDirector of PlanningSunhak Peace Prize Secretariat Further Reading : What is Digital Privacy?What is a Filter Bubble? References & Sources UN AI International Norms United Nations. Global Digital Compact — Principles for a human-centered digital future and global AI governance proposed by the United Nations. (United Nations) UNESCO AI Ethics UNESCO. Recommendation on the Ethics of Artificial Intelligence (2021) — The world’s first comprehensive international standard-setting document on AI ethics, presenting principles such as human dignity, human rights, inclusion, and cultural diversity. (UNESCO) UNESCO AI Overview UNESCO. Recommendation on the Ethics of Artificial Intelligence Overview — An overview explaining the core concepts of UNESCO’s AI ethics framework. (UNESCO) Cultural Diversity and AI UNESCO. Artificial Intelligence and Cultural Diversity — A discussion on the need to protect cultural diversity and languages in the age of AI. (UNESCO) Digital Governance United Nations. Global Digital Cooperation and Emerging Technologies — A discussion on international digital cooperation and global AI governance. (United Nations) AI and Human Rights OHCHR. The Right to Privacy in the Digital Age, 2021 — An analysis of AI surveillance and human rights issues. (OHCHR) AI Bias MIT Media Lab. Gender Shades Project, 2018 — A study on racial and gender-based error rates in facial recognition AI. (MIT Media Lab) AI Disinformation World Economic Forum. Global Risks Report 2024, 2024 — An analysis of AI-driven disinformation and global risks. (WEF)