Have you ever felt like saying "Good job!" to a food delivery machine bringing dishes in a restaurant? If you have felt a sense of "humanness" in a robot or AI that shouldn't get tired, Professor Osawa's HAI research might be involved in the background.

"A discipline that broadly deals with the relationship between humans and human-like characters or entities." Professor Osawa defines HAI research this way. Humanoid robots in human form, conversational AI, characters on screens, and video game characters—anything that humans perceive as an "other" becomes a subject of research.

When we hear the words robot or AI, we inevitably tend to associate them with sophisticated mechanical engineering or complex algorithms, but in Professor Osawa's case, the starting point of his ideas was surprisingly simple and creative.

"While doing robot research, I suddenly thought, 'My house is small; if I make a big robot, will it even fit?' When I left only the functionally necessary parts and stripped away the excess, I arrived at the anthropomorphism of home appliances."

Humans have a habit of perceiving objects as living things just because they have eye-like patterns. The first thing Professor Osawa worked on was an experiment attaching "eyes" and "hands" to home appliances like shredders and printers. Just by adding eyes to a simple mechanical appliance, humans begin to sense an "other." It is even more so if it emits a voice.

In experiments, when hearing a voice emitted by a machine, the level of understanding of the voice changed depending on the relationship with the human. "Even though the audio entering the ears is exactly the same, if you feel a sense of familiarity with the machine, the human's level of understanding and empathy changes dramatically. It's like the 'magic of anthropomorphism'," says Professor Osawa.

This "design of 'relationships'" is the essence of HAI.

While many researchers work on increasing the functionality of robots and AI, Professor Osawa focused on the "recipient." He believes that if robots and AI are designed by focusing on the "recipient's mind"—how humans perceive intelligence and how they react—they can function well within society. "HAI plays the role of that mediator," Professor Osawa thinks.

"We are designing agents, but there is a strong sense that we are designing relationships. You could say we are creating a kind of fantasy or magic."

For example, in an elementary school library, an initiative is being carried out where children entrust their impressions of books they've read to a robot, and the robot conveys the content to another child. By using the robot as a mediator, it was confirmed that even children who are not fond of books become interested, and communication between children becomes more active.

Furthermore, something unexpected happened: older students began teaching younger students how to use the robot.

"By placing a robot that people can intervene with, the spontaneous human sociality of 'wanting to help' or 'wanting to teach' is drawn out. The relationship of a human community changes based on the positioning of a single agent (AI or robot as an other)."

Why is Professor Osawa attracted to "intelligence" and "relationships"?

He reflects that as a child, he didn't fit in well with his surroundings, and "the situation of having to follow unreasonable school rules was also painful." In middle school, he wrote programs on a computer and made games.

"Whether a character moved or didn't move was my own fault for programming it. In the world of games, there was a freedom that no one could interfere with and an overwhelming sense of comfort."

He chose the path of AI research at the Keio University Faculty of Science and Technology because he thought, "If I can contribute to AI, it could be one of the reasons for being alive." Through games, he challenged research to evolve AI.

The research that made Professor Osawa's name known worldwide was the study of "AI that reads intentions" using a board game called "Hanabi."

Game research has greatly influenced the evolution of AI both domestically and internationally. "Perfect information games" where all pieces are visible, like chess or Go, are easy to analyze, and many researchers have worked on them. On the other hand, in "imperfect information games," even when players cannot see each other's hands in a cooperative board game, "inferring the opponent's invisible mind" while cooperating or conflicting determines the outcome of the game.

Through these games, Professor Osawa attempted to implement "Theory of Mind," as it is called in cognitive science, into AI.

In cooperative games, the ability to read the opponent's thoughts by considering "Why did this person tell me this information now?" becomes important. In cognitive science, it is thought that "humans have a function to estimate the minds of others (Theory of Mind)."

To have AI acquire this "Theory of Mind," Professor Osawa repeatedly conducted trials by giving the AI information based on the inference, "What if I were in the opponent's position?" Eventually, the AI was able to achieve results similar to when a human reads another human's intentions.

"Current AI is moving beyond simple probability calculations and is becoming able to back-calculate the opponent's intentions from their behavior and show consideration. If this is applied to business, truly kind services that anticipate and react to subtle user responses could become possible."

This research has been cited by world-class AI companies and research institutions such as Meta and DeepMind, and is attracting attention as a challenge that could become an important goal for next-generation AI.

There are voices of concern that AI, which is already demonstrating abilities surpassing humans in various fields, will become an entity that threatens humans in the near future.

However, the future Professor Osawa depicts is not one where AI rules humans, but a world where humans can fulfill their "individuality (dignity)" through AI. For example, a situation is possible where AI handles an elderly person with dementia when their behavior is unsettled, and they interact with family and friends when they are calm.

On the other hand, this might conflict with the value that "society should be one that accepts people, including their weakened states." Professor Osawa points out that even if technology progresses, there are problems that cannot be solved by that alone.

"We have entered a stage where everyone must discuss what kind of future is best. For that reason, engineers should not avoid points of debate regarding politics and social design."

Professor Osawa emphasizes the concept of "SF Prototyping." This is a methodology of concretely imagining a future society through the process of creating science fiction (SF), and then looking back to the present to examine ethics and systems.

The relationship between SF and AI is deep; Amazon founder Jeff Bezos and Tesla founder Elon Musk have publicly stated they were greatly influenced by SF novels and movies. Professor Osawa has also been familiar with SF since childhood, and during his university days, he belonged to both the robot technology research club and the SF research club. Furthermore, he preaches that it is important not just to read existing SF, but to know the methodology of SF.

Until now, science was something where "humans think and AI helps," but it is now shifting to a stage where "AI itself poses questions, experiments, and derives theories." The speed is fast; what takes a human a year, AI might finish in a few days. Professor Osawa points out that in this "period of upheaval" for scientific research, "the power to imagine a future that is not an extension of past data will be necessary."

By temporarily removing the constraints of "common sense was like this" and incorporating SF-like leaps, one can discover completely new "values" that are not bound by existing business models or social structures. With that in mind, he served as the president of the "Science Fiction and Fantasy Writers of Japan" for two years starting in 2022. He was the first non-professional writer to do so, and he worked on activities to fuse the imagination of SF writers with scientific knowledge such as AI research.

In 2024, he also became the Director of the "Keio University Science Fiction Research and Development/Implementation Center," a research base aiming for future social design and technological development through industry-academic collaboration utilizing the imagination of SF. As a platform where SF writers, AI researchers, humanists, and private companies mingle, they have begun SF prototyping projects.

When will the "singularity," the turning point where AI overtakes human intelligence, occur?

"There is no doubt that something extraordinary in human history is happening. We have been in a state for a long time where we don't know what will happen three months from now," says Professor Osawa. As countries and companies rich in capital and talent engage in an investment race for research, he cannot help but rethink his own role.

On the other hand, he is certain that only humans can answer such dilemmas.

"The answers provided by large-scale AI are like the median of the data; in principle, they do not provide answers that make 'oneself' happy."

Our relationship with AI will likely deepen further in the future. What will be needed then is the human imagination and thinking power that connects the two. Professor Osawa's challenge asks what "humanness" is in an era where AI can substitute for many things.

Composition: Toru Tamakawa, Editor-in-Chief of Asahi Shimbun GLOBE+

Interview/Text: Nasuka Yamamoto

Photography: Hidehiro Yamada