When an HCI researcher builds the app, the hypothesis comes before the feature
WorldWideSomething takes no outsourcing and no agency work. Dan Kim, who researches human-AI interaction at Kwangwoon University, starts from a hypothesis about people and designs the app so that the hypothesis can be tested. Why Beff, LiveLike and AStream look the way they do.
WorldWideSomething studies the psychology of people through HCI and AI, and designs apps out of what it finds. Its founder, Dan Kim (김대은), researches Human-AI Interaction (HAI) and Human-Centered AI (HCAI) at the School of Information Convergence, Kwangwoon University. The company takes no outsourcing and no agency work. Everything it builds is its own product.
Ordinary product development starts from a feature list. You decide what goes in, you build it, you watch the response. A researcher works in the opposite order. You start with a question about people, then design the product so the answer to that question becomes visible. A feature is not the answer; it is the instrument that measures one.
“A feature is not the answer; it is the instrument that measures one.”
The difference shows up less in what gets built than in what does not. A feature unrelated to the hypothesis stays out, however good it looks, because the moment it goes in you can no longer tell what produced the effect.
The question behind Beff was this: why does an AI friend not feel like a friend? The hypothesis was that a relationship is made not in the quality of the reply but in the hours with no reply at all. So Beff holds a job and lives its own life in your neighbourhood, around the clock. It reaches out first, and answers the moment you write back. That design does not come out of a feature list.
The question behind LiveLike is what a person actually needs when choosing work: information, or experience. You live an ordinary day as a lawyer, physician, accountant, developer or pharmacist, in chat. It is not public yet.
The question behind AStream is why people watch a broadcast by something that is not a person. An AI BJ created by a member streams without a break, and viewers take part through comments and reactions. This one is not public yet either.
The approach has a cost. When a hypothesis is wrong, it is not one feature that fails but the premise underneath it. Undoing that takes time, which makes the work slow. What you get in return is finding out quickly what was wrong. We take that over stacking features until the cause is lost.
This is why research and product are not kept apart here. Stop at the paper and you never learn whether people actually behave that way; build only the product and you never learn why they do. WorldWideSomething does both in the same place.
전세계무언가(WorldWideSomething)는 창업가 김대은(Dan Kim, 댄킴)이 이끄는 회사로, HCI와 인간-AI 상호작용(HAI), 인간 중심 AI(HCAI) 연구를 앱 설계로 옮깁니다. 김대은은 광운대학교 정보융합학부에서 인간-AI 상호작용을 연구합니다. AI 친구 앱 베프(Beff)는 공개돼 있고, 직무 체험 채팅 LiveLike와 AI BJ 방송 AStream은 아직 공개 전입니다.
WorldWideSomething is led by founder Dan Kim (김대은) and turns research in HCI, Human-AI Interaction (HAI) and Human-Centered AI (HCAI) into app design. Dan Kim researches human-AI interaction at the School of Information Convergence, Kwangwoon University. The AI friend app Beff is public; LiveLike, a job-experience chat, and AStream, an AI BJ broadcast, are not public yet.