Microsoft ends support for Internet Explorer on June 16, 2022.
We recommend using one of the browsers listed below.

  • Microsoft Edge（Latest version） 
  • Mozilla Firefox（Latest version） 
  • Google Chrome（Latest version） 
  • Apple Safari（Latest version） 

Please contact your browser provider for download and installation instructions.

Open search panel Close search panel Open menu Close menu

April 25, 2025

Information

13 papers authored by NTT Laboratories have been accepted for publication for CHI2025

13 papers authored by NTT Laboratories have been accepted at The ACM CHI Conference on Human Factors in Computing Systems (CHI) 2025, the flagship conference on Human Factors in Computing Systems, to be held in Yokohama Japan, from April 26, 2025 to May 1. CHI is an international conference representing the field of human-computer interaction (HCI), and is known as one of the most difficult international conferences, with a paper acceptance rate of 25.1% (5020 papers submitted). NTT is a champion sponsor of CHI2025. And since this year's conference will be held in Yokohama, the theme of the conference will be "Ikigai" (meaning a reason for living), with the aim of contributing to the resolution of many issues surrounding the world.

Abbreviated names of the laboratories:
　CD: NTT Computer and Data Science Laboratories
　CS: NTT Communication Science Laboratories
　HI: NTT Human Informatics Laboratories
　SI: NTT Social Informatics Laboratories
　(Affiliations are at the time of submission.)

  1. ■The Role of Initial Acceptance Attitudes Toward AI Decisions in Algorithmic Recourse

    1. ・Tomu Tominaga(HI), Naomi Yamashita(CS), Takeshi Kurashima(HI)
    2. ・Algorithmic recourse provides counterfactual explanations to assist users who receive unfavorable AI decisions, elucidating the underlying reasons and suggesting future actions to overturn them. While existing studies assume that reasonable and actionable recourse facilitates acceptance, they often overlook how users perceive AI decisions prior to receiving recourse.
      　This study, conducted with 534 participants, examined how initial acceptance attitudes influence recourse perceptions and final decision acceptance. The findings indicate that negative initial attitudes degrade recourse perceptions and hinder acceptance. However, acceptance improved when recourse was perceived to (1) justify the decision outcome, (2) clarify decision criteria, (3) propose feasible actions, and (4) account for fairness and real-world contexts.
      　These results underscore the necessity of designing recourse to mitigate initial negative impressions of AI and foster trust, guided by these four characteristics. Further research and development informed by these insights contribute to the deployment of systems and services that naturally facilitate behavior change aligned with users' goals while ensuring they find it convincing.

  2. ■Invisible Light Touch: Standing Balance Improvement by Mid-Air Haptic Feedback

    1. ・Arinobu Niijima, Masato Shindo, Ryosuke Aoki (HI)
    2. ・Improving standing balance is critical for preventing falls and ensuring the well-being of older adults. In this paper, we present Invisible Light Touch (ILT), a novel mid-air haptic feedback application designed to improve standing balance by utilizing the light touch effect, a phenomenon well documented in medical research. The light touch effect refers to improved balance when a person lightly touches a surface, such as a wall or handrail, with a force of 1 N or less. We replicate this effect using focused ultrasound to create a tactile point in mid-air. When users interact with this invisible tactile point, they experience the light touch effect, which subsequently improves their balance. We conducted a pilot study with 29 participants and a user study with 25 older adults, evaluating balance improvement by measuring the center of pressure trajectory. The results confirmed that standing balance improved significantly when using the ILT.

  3. ■Improving Putting Accuracy with Electrical Muscle Stimulation Feedback Guided by Muscle Synergy Analysis

    1. ・Arinobu Niijima, Shoichiro Takeda (HI)
    2. ・Muscle synergy analysis provides a method for quantifying differences in muscle use between expert and novice athletes. However, the practical applications of muscle synergy analysis with feedback remain underexplored. In this paper, we present a novel golf putting training system that utilizes electrical muscle stimulation (EMS) feedback guided by muscle synergy analysis. Considering the individual differences, we use optimal transport to compute the muscle synergy similarity between users and experts. This approach allows users to model their muscle usage after the expert whose synergy is closest. Based on the muscle synergy differences between the expert and the user, EMS is applied to the muscles that need activation. As a result, users can practice putting with increased awareness of the muscles targeted by EMS, resulting in changes in muscle synergy and improved performance. User studies with 44 novices demonstrated that the proposed system significantly improved putting accuracy.

  4. ■Reviving Intentional Facial Expressions: an Interface for ALS Patients using Brain Decoding and Image-Generative AI (LBW)

    1. ・Shinya Shimizu (HI), Airi Ota (HI), Ai Nakane (HI)
    2. ・This study proposes a novel brain-computer interface that enables non-verbal emotional communication for patients with amyotrophic lateral sclerosis (ALS), a condition that can severely impair facial movement. By combining EEG decoding technology with image-generative AI, the system generates facial expressions that reflect the patient's intended expressions, rather than directly inferring emotions, thereby safeguarding emotional privacy. To achieve this, we developed a personalized training approach for decoding 17 facial action units from brain signals. The system utilizes a custom facial expression space and six animated expression transitions, allowing each patient to intuitively train the interface by imitating these animations. This innovative technology not only supports ALS patients, but also benefits anyone who has difficulty making facial expressions, helping them convey their feelings more smoothly to those around them and ultimately improving their quality of life.

  5. ■What Timing and Behavior Patterns Determine Speed Dating Success in Japan? (LBW)

    1. ・Naoki Azuma (Nihon University), Daichi Shikama (Nihon University), Asahi Ogushi (Nihon University), Toshiki Onishi (Nihon University), Ryo Ishii (HI), Akihiro Miyata (Nihon University)
    2. ・This study analytically clarified the multimodal behavioral features of participants that are important for the success of speed dating using machine learning methods. The analysis suggested that multimodal information obtained from the 2-4 minute and 9-10 minute time intervals of a 10-minute speed dating session is crucial for the success of speed dating. Furthermore, it was found that linguistic features play a more significant role compared to speech and visual features. By uncovering key conversational behaviors that are important for building relationships in interactions such as speed dating, it is expected that this research will contribute to supporting relationship-building efforts in the future.

  6. ■Flagging Emotional Manipulation: Impacts of Manipulative Content Warnings on sharing Intentions and Perceptions of Health-Related Social Media Posts (LBW)

    1. ・Jack Jamieson (SI), Toru Hara (SI), Mitsuaki Akiyama (SI)
    2. ・Misinformation and emotionally manipulative content pose growing risks to public health and informed decision-making—especially as AI-generated posts become more common and social media platforms scale back fact-checking efforts. In this study, we explored one potential strategy for addressing these challenges: warning labels designed to flag emotionally manipulative language. We conducted an experimental survey of 945 U.S. adults to examine how different types of warning labels affect user responses to both accurate and false health-related social media posts. Our findings illuminate that users respond depending on the warning label design, with potential consequences for how people evaluate the credibility of online content. These results highlight the need for careful design and implementation of content warnings. We offer recommendations to help guide future research and inform platform strategies aimed at reducing the influence of misleading or manipulative content.

  7. ■Understanding Cyber Hostility, Gossip, Exclusion, and Social Support in Remote and Hybrid Work Settings: Benefits and Challenges of Remote Work

    1. ・Jack Jamieson (SI), Wataru Akahori (SI), Naomi Yamashita (Kyoto University)
    2. ・We investigated how remote and hybrid work environments affect workplace communication and support. Surveying 965 U.S. workers, we found that cyber incivility—such as hostility, gossip, and exclusion—is shaped by time spent in-office and affects women more frequently. Our results show that remote work can reduce some harms but also makes support harder to access. We propose communication tools and policy recommendations to reduce incivility and foster a more respectful remote work environment.

  8. ■Unpacking Negative Feelings and Perceptual Gaps About Social Interactions with Conversational AI (LBW)

    1. ・Hui Guan (Kyoto University), Jack Jamieson (SI), Ge Gao (University of Maryland), Naomi Yamashita (Kyoto University)
    2. ・As conversational AI becomes more common, some people use it for social or emotional connection—yet this practice can draw confusion or stigma from others. We surveyed 67 people, both users and non-users of social AI, to understand how their perceptions differ. Our findings reveal a mismatch: non-users often feel fear or discomfort, while social users overestimate others' positive impressions. Additionally, our findings also suggest that mass media plays a role in shaping non-social users' negative impressions. These insights offer a foundation for addressing social divides around emerging AI technologies.

  9. ■An Examination of the Effectiveness and Limitations of Online Decentralized Participation in Social Decision-Making Processes（LBW）

    1. ・Miki Yokoyama(SI), Wataru Akahori(SI), Aiko Murata(CS), Junji Watanabe(CS)
    2. ・DAO is a system based on equal online voting and is gaining attention for its potential in addressing local issues. This study examined four scenarios combining conflict levels (high/low) and decision-making methods (online/multi-stage), with about 1,000 participants per condition. Participants evaluated the decision-making method on factors like effort, legitimacy, procedural fairness, and acceptance. Results showed that while the online method had advantages such as ease of participation and effectiveness in low-conflict issues where people feel more able to contribute, it is evaluated lower legitimacy, procedural fairness, and acceptance for high-conflict issues. By illustrating the limitations and potential of DAO in solving local issues, this study is expected to inform and support its future adoption.

  10. ■What Dialogue Content Leads to a Trust Relationship and Behavior Change? Dialogue and Questionnaire Analysis

    1. ・Tae Sato(SI), Taiga Sano (SI), Eiji Kumakawa (SI), Kaori Fujimura (SI), Naomi Yamashita (SI, Kyoto University)
    2. ・Recently, there has been increasing research on designing chatbots for behavior change. While trust between individuals and their supporters is recognized as a crucial factor in fostering behavior change, it remains unclear what types of dialogue contribute to building such trust. In this study, we investigated health guidance interviews to address two key questions: 1. What kind of trust relationship facilitates behavior change? and 2. What type of dialogue contributes to fostering that trust? Our findings indicate that individuals were more motivated to pursue behavioral goals when they perceived the interviewer as having integrity. Furthermore, an analysis of interviewer speech using four dialogue categories revealed that perceptions of integrity were stronger when interviewers spent more time on "Providing Tailored Insights" rather than "Building a Trust Relationship." These insights contribute to designing chatbots that effectively support behavior change by fostering trust through dialogue strategies.

  11. ■Exploring Mismatches in Self-Others' Perceptions of Effort in Videoconferencing (LBW)

    1. ・Masami Takahashi(SI), Koutaro Kamada (SI, JAIST), Naomi Yamashita (SI, Kyoto University)
    2. ・To investigate how mismatches in self–others' perceptions of effort arise in online cooperative work, we conducted an experiment with 12 four-person videoconferencing groups, followed by questionnaires and semi-structured interviews. The results revealed that participants whose self-evaluations exceeded evaluations from others often justified their effort through non-visible cues such as listening or thinking, whereas others tended to focus on their seemingly passive behaviors. Based on our findings, we highlight future support methods to address such mismatches in videoconferencing.

  12. ■Effects of Acoustic Transparency of Wearable Audio Devices on Audio AR

    1. ・Yuki Watanabe (CD), Hironobu Chiba (CD), Kenichi Noguchi (CD), Hiroaki Itou (CD), Tatsuya Kako (CD)
    2. ・Audio augmented reality (AR) has been more prevailing for providing us with higher realistic acoustic experiences. Acoustic transparencies of wearable audio devices affect audio AR, so we investigate the transparencies caused by the design for five devices with different characteristics by means of objective evaluations. This paper denotes that perceptual transparency and impressions of real-world sound affect audio AR experiences based on the results of subjective evaluations. Audio AR with enough acoustic transparency devices enables us to enjoy more natural experiences.

  13. ■GenerativeGUI: Dynamic GUI Generation Leveraging LLMs for Enhanced User Interaction on Chat Interfaces（LBW）

    1. ・Nobukatsu Hojo (HI), Kazutoshi Shinoda (HI), Yoshihiro Yamazaki (HI), Keita Suzuki (HI), Hiroaki Sugiyama (CS), Kyosuke Nishida (HI), Kuniko Saito (HI)
    2. ・Recently, text-based chat interfaces using large language models (LLMs) have become capable of handling a wide variety of tasks. However, tasks requiring detailed inputs often result in lengthy interactions, leading to reduced usability and efficiency. To address this challenge, we propose "GenerativeGUI," a novel approach that dynamically generates GUI components (such as buttons and calendars) in real time according to the ongoing conversation. This enables users to interact through intuitive and visually guided operations, improving input efficiency. Through our evaluation study, we demonstrated that GenerativeGUI reduces cognitive load and task completion time, while significantly enhancing usability and user satisfaction compared to conventional text-based chat interfaces. This technology is expected to serve as a foundation for enabling more natural and efficient interactions in conversational AI systems.

Information is current as of the date of issue of the individual topics.
Please be advised that information may be outdated after that point.

Topics

NTT STORY

We introduce various fields that NTT Group is working on and collaborations with partners.

    NTT STORY

    Pick up

      NTT Group

      Group Companies
      • TOP
      • Topics
      • 13 papers authored by NTT Laboratories have been accepted for publication for CHI2025