Event Report: NTT Communication Science Laboratories Open House 2025

Shiro Kumano, Naomi Watanabe, Yusuke Tanaka,
Takuya Komura, Ryohei Shibue, and Hiroyuki Fujinaga

Abstract

NTT Communication Science Laboratories hosted Open House 2025 from 20 to 22 May 2025 at QUINTBRIDGE and the adjoining PRISM facility on NTT WEST’s i-CAMPUS in Kyobashi, Osaka. Operated entirely as an on-site event with advance reservations, the Open House welcomed 1078 visitors—about a 20% increase over the previous year. This article reports on the event’s highlights.

Keywords: information science, human science, AI

1. Overview of Open House

Since its founding, NTT Communication Science Laboratories (CS Labs) has pursued forward-looking basic research aimed at enabling heart-touching communication, whether between people or between people and computers. As an event where visitors can “see, touch, and feel” our latest achievements, CS Labs has held the Open House every year from late May to early June at the NTT Keihanna Building (Seika-cho, Kyoto).

From 2020 to 2022, due to COVID-19 countermeasures, the on-site event was suspended and instead held online by publishing lecture and exhibit videos on a special website. In 2023, on-site exhibits resumed for the first time in four years at QUINTBRIDGE, NTT WEST’s open-innovation facility in Osaka, which offers convenient access from major train stations in the Kansai region. The following year, 2024, the adjacent PRISM facility was also used, enabling the first full on-site event since 2019, including invited and research talks. Continuing this approach, Open House 2025 was again held on-site at QUINTBRIDGE and PRISM for three days, 20–22 May.

Under the theme “Orchestrating our future through the symphony of knowledge,” the 2025 program featured 6 talks (11 sessions in total, including 1 invited talk) at QUINTBRIDGE and 20 research exhibits at PRISM. As in the previous year, a slot-based advance-reservation system was adopted, and total attendance reached 1078 visitors—about 20% more than the previous year. As a new attempt, each research talk was presented twice; together with newly introduced open seating, this attracted a total of 1037 attendees (94 per talk on average).

2. Invited talk

The invited talk was delivered by Professor Emeritus Mutsumi Imai of Keio University, Director of Mutsumi Imai Educational Research Institute, under the title “How humans construct knowledge systems: Symbol grounding and abduction” (Photo 1). Facing the fundamental question, “How will artificial intelligence (AI)—becoming part of our daily infrastructure via smartphones and cloud services—affect human intelligence?”, Professor Imai centered her discussion on the differences between AI and human thinking/learning, drawing on findings from language-acquisition research. The first key phrase of the talk was “symbol-grounding problem.” Proposed in the 1990s by Canadian cognitive scientist Stevan Harnad, this problem points out that AI without a body cannot tie language symbols to external objects, therefore fails to reach true semantic understanding. Whereas human children ground words like milk or bread in multi-sensory experience, AI can define them only as chains of other symbols and continues to drift in a sea of language—this contrast was highlighted.

Photo 1. Invited talk by Professor Emeritus Mutsumi Imai.

The second key phrase was “abductive inference (hypothesis-forming inference).” Children locate newly acquired words within the entire lexicon and infer their scope while building a semantic system. This process parallels a scientist’s refinement of hypotheses through repeated experimentation and correction and was explained as the source of human-knowledge creation. In contrast, current neural-network AI, which inductively extracts patterns from large amounts of data, lacks abduction that assumes causal mechanisms, making discovery of new knowledge and conceptual reframing difficult.

Professor Imai further stated that the inevitable “errors” within abductive inference and the process of correcting them support human flexible and creative learning. She concluded that resolving the symbol-grounding problem and overcoming the “frame problem,” which concerns deciding which knowledge to use when, are crucial for achieving general, autonomous AI learning, and constitute core issues for both human understanding and AI development. The talk presented points of intersection between language/cognition research and AI technology in the age of generative AI, prompting lively questions from diverse participants. The talk video, like those of the research talks, is now available on the Open House website (in Japanese only) [1] until approximately June 2026.

3. Research talks

From CS Labs’ latest achievements, five high-profile topics were selected for the research talks. Each presentation was followed by a lively question and answer (Q&A) session, and visitor surveys showed an average satisfaction rate above 90%.

In the talk entitled “From the study of embodied empathy to supporting family wellbeing—Understanding embodied empathy and connecting distant families via bodily information transfer—,” Dr. Aiko Murata (Human Information Science Laboratory) focused on affective empathy, which arises as bodily responses align subconsciously. She introduced a new online visitation system that enables parents to perceive their hospitalized baby’s heartbeat as vibrotactile sensations while observing them, as if holding their baby (Photo 2), a technology expected to support bonding.

Photo 2. Research talk by Dr. Aiko Murata.

In the talk entitled “Techniques for ‘reading the room’ in attentive conversational AI—Understanding dialogue context through multimodal information and incremental response generation—,” Dr. Yuya Chiba (Innovative Communication Laboratory) presented a dialogue agent that infers atmosphere and interpersonal relationships from multimodal cues, i.e., language, vocal tone, and facial expression, and generates timely, considerate responses (Photo 3). The work aims to build dialogue systems that grasp everyday situations and support people naturally.

Photo 3. Research talk by Dr. Yuya Chiba.

In the talk entitled “AI that learns to listen on its own—Advancing self-supervised audio representation toward cutting-edge sound understanding with large language models—,” Daisuke Niizumi (Media Information Laboratory) presented audio representation learning technologies that enable AI to interpret the diverse sounds in our environment (Photo 4). Technologies based on self-supervised learning methods are further evolving to enable AI with a linguistic understanding of sounds with the help of large language models. Potential applications include services such as health monitoring based on everyday sounds.

Photo 4. Research talk by Daisuke Niizumi.

In the talk entitled “Discovery of hidden knowledge in data relationships—Prospects for reliable healthcare through infinite-hypothesis AI models that interpret biological phenomena—,” Masahiro Nakano (Media Information Laboratory) introduced Bayesian nonparametric methods that integrate biological data and uncover previously unknown causal structures (Photo 5). He showed three applications, i.e., relational data analysis, phylogenetic analysis, and trajectory inference, that pave the way for trustworthy healthcare.

Photo 5. Research talk by Masahiro Nakano.

In the talk entitled “Children perceive minds in robots—Learning companion robots for the future of early childhood education—,” Dr. Yuko Okumura (Innovative Communication Laboratory) presented experiments showing that social interactions with robots can enhance children’s prosocial behavior and lead them to perceive the robots as having a mind (Photo 6). These findings suggest that social robots may serve effectively as learning companions that support children’s development.

Photo 6. Research talk by Dr. Yuko Okumura.

For more details on the five research talks, read their respective feature articles in the current issue.

4. Research exhibits

This year’s Open House presented 20 exhibits across four categories, with interactive explanations by researchers on-site; 12 exhibits featured demos for more intuitive understanding (Photo 7). The summaries of these exhibits are available on the Open House website [2], and the titles in each category are listed below.

Photo 7. Research exhibit.

4.1 Science of Machine Learning

Collaborative learning with multiple thoughts—Transparent prediction based on decision tree superposition
Accurate spatial prediction with limited data—Meta-learning based on neural Gaussian processes
AI can adapt to new environments without labels—Theoretical understanding of source-free domain adaptation
Toward the realization of low-power optical AI—Training of optical neural networks with special structure
A mathematical link for light-matter interaction—New unification through non-commutative harmonic oscillators
Crafting noises in quantum computing!—Reshaping noises to improve accuracy of quantum computation

4.2 Science of Media Information

Touch experience without contact—Non-contact rendering of texture by focused ultrasound
Cleaning-up speech from noisy, reverberant recordings—Ensemble of multi-stream diffusion model enhances speech
Live streaming with real-time voice conversion—Real-time voice conversion with high quality and low latency
Enhancing general data compression—Code-tree sets for efficient versatile lossless encoding

4.3 Science of Communication and Computation

Children behave well in front of a social robot—Interactive robots promote children’s prosocial behavior
Secondborns’ lower verbal skills improve in school-age—The effect of older siblings on child language development
Faithful translation without excess or deficiency—Preference optimization for LLM-based translation
Capturing temporal relationship changes on SNS—Social orbits: Linking SNS logs and sociological concepts

4.4 Science of Human

Let me lead you through the city, empowered by AI—Toward a real-world deployment of tactile gadget Buru-Navi4
Supporting family bonds with hospitalized babies—An embodied online visitation for NICU newborns and families
Seeing the essence of baseball batting—Winning visual strategies of professional baseball players
The “Batting eye” in action—Motor control underlies rapid decisions
Unseen light that enhances cognitive task performance—Interventions in mental states via ipRGC-modulating light
Is oversleeping on free days really a problem!?—Effect of chronic sleep debt on cognitive task performance

5. After the Open House

In 2025, by once again adopting an advance-reservation system, refreshing the venue layout, increasing the number of talk sessions, and introducing open seating—balancing continuity and new challenges—the Open House welcomed more visitors while focusing on safety and comfort. For CS Labs researchers, direct conversations and free, lively discussions with visitors of diverse backgrounds were extremely meaningful. The layout of the Main Stage at QUINTBRIDGE, which made full use of the entire space, enabled audiences to feel close to the speakers during the Q&A sessions. At PRISM, compactly grouping 20 exhibits allowed for efficient viewing. We extend our heartfelt appreciation to everyone who supported the event.

References

[1]	Website of NTT Communication Science Laboratories Open House 2025 (in Japanese), https://www.kecl.ntt.co.jp/openhouse/2025/index.html
[2]	Website of NTT Communication Science Laboratories Open House 2025 (in English), https://www.kecl.ntt.co.jp/openhouse/2025/archive_exhibition_en.html

	Shiro Kumano Senior Research Scientist, Human Information Laboratory, NTT Communication Science Laboratories. He received a Ph.D. in information science and technology from the Graduate School of Information Science and Technology, the University of Tokyo, in 2009. He joined NTT in the same year. His research interests include affective computing, particularly in the areas of interpersonal behavior measurement and emotion estimation. He is a member of the Institute of Electrical and Electronics Engineers (IEEE), the Institute of Electronics, Information and Communication Engineers (IEICE), the Japanese Society for Artificial Intelligence (JSAI), the Japan Neuroscience Society (JNS) and Association for the Advancement of Affective Computing (AAAC).
	Naomi Watanabe Senior Research Scientist, Innovative Communication Laboratory, NTT Communication Science Laboratories. She received a Ph.D. in developmental psychology from George Mason University in USA in 2015. Since joining NTT Communication Science Laboratories in 2016, she has conducted research focusing on various aspects of child development, particularly social-emotional learning, emotional development, and the influences of childcare, parenting, and cultural factors. She is affiliated with the Japan Society of Developmental Psychology and the Japanese Association of Educational Psychology. Currently, she is working in the Corporate Strategy Planning Department at NTT.
	Yusuke Tanaka Research Scientist, Innovative Communication Laboratory, NTT Communication Science Laboratories. He received a Ph.D. in informatics from Kyoto University in 2020. He joined NTT in 2013. His research interests include scientific machine learning, especially data-driven approaches to physical modeling and simulation. He is a member of IEICE and JSAI.
	Takuya Komura Research Scientist, Human Information Science Laboratory, NTT Communication Science Laboratories. He received a Ph.D. in arts and sciences from the University of Tokyo in 2016 and joined NTT the same year. His research interests include the emergence of psychological functions and computational neuroscience. He is a member of the Japanese Neural Network Society and the Acoustical Society of Japan.
	Ryohei Shibue Research Scientist, Media Information Laboratory, NTT Communication Science Laboratories. He received an M.S. in mathematical informatics from the University of Tokyo in 2017 and joined NTT Communication Science Laboratories the same year. His research interests include statistical modeling of biological time series, such as heart sounds, and neural data analysis, particularly spike train analysis from brain recordings. He is a member of the Japan Statistical Society.
	Hiroyuki Fujinaga Assistant Section Manager, Research and Planning Section, NTT Communication Science Laboratories. He received a B.A. in business administration from Saga University in 1998 and joined NTT the same year, where he was mainly engaged in sales and business support. He joined NTT Science and Core Technology Laboratory Group in 2013, worked at NTT Information Network Laboratory Group, and has been working at NTT Communication Science Laboratories since 2021.

↑ TOP