´ºÓêÖ±²¥app

FEATURES

Japanese

Print

AI tools allow diverse ideas, opinions to become visible ¡ª and heard

October 17, 2025

Adobe image
©Adobe Stock


Artificial intelligence (AI) is rapidly becoming an integral part of daily life and business. Yukino Baba, an associate professor at the Graduate School of Arts and Sciences, is a computer science specialist who has been conducting research on human-AI collaboration. She has developed AI tools that support decision-making processes, like identifying reliable opinions among the views expressed in group discussions, or making fair and unbiased evaluations. ¡¡

“At the core of my research is the idea of realizing more efficient decision-making,” said Baba, who joined the University of Tokyo in 2022. She was selected as a University of Tokyo Excellent Young Researcher the following year.

In January 2025, Baba released the beta version of a web application named Illumidea, a portmanteau formed by combining “illuminate” and “idea.” The app uses ChatGPT, an AI chatbot, to automatically sort and categorize anonymous opinions, making them clear and easy to understand. The tool adds visibility to important or innovative ideas that are otherwise overlooked because the participant was not assertive enough to voice their opinion or struggled to express themselves verbally in a meeting or discussion setting. It can change the general direction of discussions that tend to be swayed by the majority opinion or loud, dominant voices.

“I want every member of society to be able to take part in activities that make our world better,” Baba said. “That requires sharing opinions and ideas from many different people with everyone. That said, it’s difficult for humans to process a massive amount of information. That’s where AI can play a role.”

illumidea1
The submission form of the Illumidea app, released in January 2025. Users can submit their opinions anonymously.
illumidea2
By clicking the “close submissions” button after collecting opinions, the app begins classifying them — a process that takes only a few seconds.
illumidea3
The collected opinions are categorized by point of view, with similar responses consolidated into one. Because AI-driven classification is not always accurate, users can manually review and amend the results.
 

Programmer from an early age

Baba’s first experience with computers came when she was in elementary school. Back then, her father, who was a software engineer, owned a home computer (operating on the MSX open standard popular in Japan at the time), which he had placed in Baba’s room because he couldn’t find anywhere else to put it. So, she began playing with it. Using a beginner’s guide for the BASIC programming language, which she had borrowed from the library, she started writing simple programs and creating computer games. And after Microsoft’s Windows 95 operating system was released and internet use had spread, Baba began creating her own websites and chatting with strangers on online forums during her junior high and high school years. Looking back, she said she immersed herself in the digital world partly because she had difficulty interacting with her classmates at her all-girls school.

After graduating from the Department of Electrical Engineering at the Tokyo University of Science, Baba enrolled at the Graduate School of Information Science and Technology at ´ºÓêÖ±²¥app. There, she began studying AI.

“Back then, the internet was rapidly expanding and people were beginning to post all kinds of information online,” she said. “That gave rise to web mining – a technique to extract information, such as current trends, from the massive amount of data on the web. That’s what drew me to AI research.”

At graduate school, Baba studied web mining, focusing her research on the photo sharing platform Flickr. She used tags and location information attached to the photos posted there to deduce the knowledge people had gained subconsciously. She also conducted research linking the visual features of photos and their tags.

After earning a doctoral degree, Baba worked on developing AI technologies related to crowdsourcing. Among them is a technology to discern reliable evaluators. For example, when people are asked to rate a restaurant, some may evaluate half-heartedly. The tool can be used to find out who the trustworthy critics are.

She also developed an AI that detects valuable opinions — which eventually led to the creation of Illumidea. In group discussions, participants submit their views anonymously; then they vote on the opinions that were collected, also incognito. Based on the results, the tool highlights the opinions that are considered important.

“Rather than simply treating the most popular opinions as the most important, the tool assigns weight to minority perspectives and selects the meaningful opinions from among them,” Baba explained. “Based on a mathematical model of voting behaviors, it ascertains the underlying values of each participant.”

In a pilot study with first-year high school students, the freshmen discussed the problem of some classmates not actively engaging in group projects. In the study, the students broke up into two discussion groups: one equipped with the AI tool and the other without. In the latter group, without the AI tool, only those who actively participated in group work spoke up, and the discussion proceeded led by the opinions of the dominant members, who were larger in number. But in the group that used the tool, as AI highlighted and made visible the minority views, the direction of the discussion shifted significantly, leading to more inclusive and thoughtful dialogue, Baba said.

 

AI tool
In an experiment with first-year high school students, discussions in the group that did not use AI skewed toward one side and ended up reflecting only that side’s point of view.
 
AI tool
In the AI-assisted group, the minority opinion was made visible, which led to a discussion that incorporated the perspectives of both sides.
 

Partner AI for scientists

Baba-sensei
Yukino Baba, an associate professor at the Graduate School of Arts and Sciences
 

One of Baba’s current projects is the development of a partner AI for researchers. She envisions AI that suggests hypotheses, recommends readings, and assists in conducting experiments and writing papers. The joint project is part of the Moonshot Research and Development Programs of the Japan Science and Technology Agency, with the goal set for 2050.

Her team is tasked with encoding specialized knowledge into AI systems. They are currently working together with chemistry researchers. The challenge, she said, is capturing the intuitive “sense” researchers sometimes rely on.

“They say they sometimes can just sense whether a chemical compound is a strong prospect for a drug or not. But they can’t verbalize that intuition. When they try, it doesn’t quite match the actual sense they have,” Baba said. “We may need to look beyond language, to things like the gaze cast by their eyes, to capture that intuition.”

Looking ahead, Baba hopes to develop AI that facilitates communications, such as bouncing ideas off tools that mimic conversation partners and simulate how they might respond.

“I believe many conflicts stem from friction between people,” she said. “Friction happens even when both sides have no malicious intent. I hope to use AI to support communication to reduce such unfortunate situations where people regard things differently from what was truly intended.”

Access Map
Close
Kashiwa Campus
Close
Hongo Campus
Close
Komaba Campus
Close