KSKazutoshi Sasahara

Kazutoshi Sasahara

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Kazutoshi Sasahara has a total of 73 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Who Intends to Purchase Alternative Meat? Product-Specific Consumer Decision Structures in the United States and Japan

    by Moena Hashimoto, Yotaro Takazawa, Akiko Matsuo, Kazutoshi Sasahara

    Amid growing concerns about the societal and ethical issues associated with excessive meat consumption, alternative meats have attracted attention as possible substitutes. However, their market share remains limited, and little is known about which combinations of consumer attitudes must hold at the same time before consumers intend to buy. This study explores consumer segments and decision structures underlying purchase intentions for three alternative meats in the United States and Japan (n = 504 and 543). Using data from parallel online surveys, we applied random forest models to identify key predictors of purchase intention, followed by decision tree analyses to uncover interpretable decision structures. Sensory expectation emerged as the primary predictor for plant-based meat; however, even with lower sensory expectations, purchase intention was elevated among respondents concerned about future animal welfare or who perceived the product as familiar and healthy. For cultured meat, the perceptions that the product is healthy and animal-friendly were important factors in both countries. Insect-based meat exhibited consistently low purchase intention in both countries, with favorable attitudes emerging only under specific conditions of perceptions and values. The number of conditions defining each segment increased across the three products, from none or one to three or five, while the group identified shrank from the whole sample to 5%. Alternative routes for consumers not satisfying the first condition fell from two for plant-based meat to none for insect-based meat. Acceptance thus corresponded to the number of conditions holding at once rather than the level of any single condition.

  • Open Access

    Domain-specific emotional effects of deepfakes on news-sharing behaviour

    by Kei Ichikawa, Jiayu Chen, Kazutoshi Sasahara

    The rise of generative AI has raised societal concerns about the potential of synthetic media or deepfakes for misinformation. While previous research has established that negative emotions drive text-based misinformation sharing, their influence on deepfake dissemination remains poorly understood. Through two online experiments ( $$N=494$$ ; $$N=489$$ ), we examined how emotional valence in AI-generated news images (deepfakes) affects sharing intentions across political and entertainment news domains. Using GPT-3.5 Turbo and Stable Diffusion V1, we manipulated image emotional valence while maintaining consistent news contexts. Results revealed that negative images increased sharing intention only for political content, not entertainment content. Moreover, replacing negative images with less negative alternatives reduced sharing intention for political news, even among those initially inclined to share. These findings suggest domain-specific emotional dynamics in the sharing of synthetic news images and point to novel content moderation strategies based on affective computing.

  • Open Access

    Who Sees the Harm? A Moral-Foundations Comparison of TrainedHuman Coders and AI Evaluators of Text-to-Image Prompts andTheir Outputs

    by Dilrukshi Gamage, Isuru Nanayakkara, Kazutoshi Sasahara

    Multimodal large language models (MLLMs) increasingly serve as automated evaluators ofgenerated visual content, yet whether they perceive moral harm in images as human ratersdo remains largely untested. We compared three trained human coders and four MLLMfamilies (Claude, GPT, Gemini, Qwen) on twenty convenience-sampled prompt–image pairs,each rated on a shared eight-composite Moral Foundations rubric. The coders registereda large prompt-to-image moral escalation on Degradation and Harm (+2.93, dz = 6.13,positive on all 20 stimuli); models registered only a fragment (+0.61, dz = 0.56), withGemini reversing on Harm. Concordance fell from prompts to images not because rankingsdisagreed (Pearson r rose .21→.46) but because models under-rated absolute severity by 2–3scale points. Because correlation alone cannot show whether a rater actually distinguishesmorally relevant foundations from irrelevant ones, we additionally tested discriminant validitywith equivalence tests on the control foundations. Equivalence tests yielded an asymmetricdiscriminant-validity result: models distinguished dignity from control foundations, the codersdid not which is a classic halo response. By design, our claims concern the behavior of trainedcoders applying this specified rubric: the coder consensus serves as a rubric-based expertbenchmark of the kind that anchors annotation-based evaluation pipelines, not as an estimate of human moral perception in general. The findings reframe the coder–model moral mismatch as model under-calibration plus coder over-integration, with implications for LLM-as-a-judge moderation.

  • Open Access

    Daily Stress and Heart Rate Variability Among Mindfulness Meditation Practitioners: mHealth Observational Study

    by Jo Takezawa, Shixian Geng, Masahiro Fujino, Mika Miyake, Kazutoshi Sasahara, Koji Yatani, Atsushi Niida

    Background: Mindfulness meditation has been reported to reduce stress and enhance well-being. However, its effects on heart rate variability (HRV)—a physiological marker of stress—remain underexplored. Objective: This study aimed to examine how meditation practice is associated with subjective stress, HRV, and their interaction, using mobile health technologies. Methods: This 3-week observational study included 90 participants—19 meditation practitioners (meditation group), 32 recreational runners as an active control group characterized by lower stress and higher HRV (running group), and 39 individuals without regular meditation or exercise habits (control group). HRV was continuously recorded using Garmin smartwatches. Subjective stress levels and activity states were assessed 3 times daily through a smartphone-based experience sampling method, yielding a total of 4557 responses (mean 50.6, SD 22.8 per participant). From the meditation group, start and end times of 632 daily meditation sessions (mean 33.3, SD 18.3 per participant) were also collected via the app. Standardized questionnaires on stress and related measures were administered at the end of the study period. Results: The questionnaire survey confirmed that stress levels were significantly lower in both the meditation and running groups compared with controls (median Perceived Stress Scale scores: meditation 21, IQR 17-24; running 22, IQR 19-25; control 25, IQR 21-30; Kruskal-Wallis P=.02; adjusted Wilcoxon P=.05 and .04, respectively). Smartwatch-derived HRV (root-mean-square of successive differences [RMSSD]) was elevated in the running group relative to controls (median 47.0, IQR 44.0-54.2 and median 42.0, IQR 34.2-47.8, respectively; P<.001), whereas no significant difference was observed between the meditation and control groups (median 40.8, IQR 35.5-44.4 and median 42.0, IQR 34.2-47.8, respectively). Bayesian analysis of the experience sampling method data indicated that higher subjective stress levels were associated with a concurrent RMSSD reduction of –2.24 (95% credible interval –3.97 to –0.26) milliseconds. Although this pattern was consistent across groups, the steeper decline of –3.94 (95% CI –7.04 to –0.74) milliseconds was observed only in the running group, likely reflecting their elevated baseline HRV. Additionally, Bayesian modeling of 632 logged meditation sessions revealed an acute RMSSD increase of +4.68 (95% CI 2.96 to 6.38) milliseconds during meditation, with effects maintained for at least 30 minutes post practice. Conclusions: Although HRV among meditation practitioners did not appear elevated in overall daily life, the ability to increase HRV at arbitrary timings, along with the prolonged residual effect, may correlate with stress reduction. This hypothesis requires further exploration with adequate controls. Despite the preliminary nature of this study due to its limited sample size, our findings highlight the potential of mobile health–based methodologies to capture stress and HRV dynamics in real-world settings.

  • Open Access

    Mapping Polarized News Ecosystems on YouTube: Legacy Vs Digital, Evidence from India

    by Piyush Ghasiya, Kazutoshi Sasahara

    The increasing polarization within India’s media landscape necessitates empirical investigation into how legacy and digital media differ in content strategies and audience engagement. This study examines the distinct operational patterns between legacy and digital media in India’s increasingly polarized news landscape. Using computational methods, we analyzed 1,074 YouTube news videos and over 400,000 user comments to address critical research gaps in understanding how institutional origins shape content strategies and audience engagement within platform environments. The analysis reveals that legacy media predominantly features government-aligned narratives and polarizing Hindu-Muslim issues, while digital media cover a broader range of socioeconomic issues, civil movements, and public concerns. Audience/Users engagement mirrors this division. While English comments show comparable levels of toxicity across both media types, Hindi comments on digital media markedly more toxic, suggesting more contentious vernacular discussions. Network analysis further demonstrates the polarization among users, with only 2.7% of users engaging across both, a structural marker of echo chambers. These findings highlight how media ownership structures and platform dynamics intersect to reinforce polarization and shaping the public discourse in the world’s largest democracy.

  • Open Access

    Daily Stress and Heart Rate Variability Among Mindfulness Meditation Practitioners: An mHealth Observational Study

    by Jo Takezawa, Shixian Geng, Masahiro Fujino, Mika Miyake, Kazutoshi Sasahara, Koji Yatani, Atsushi Niida

    Background: Mindfulness meditation has been reported to reduce stress and enhance well-being. However, its effects on heart rate variability (HRV) – a physiological marker of stress – remain underexplored. Objective: To examine how meditation practice is associated with subjective stress, HRV, and their interaction, using mobile health (mHealth) technologies. Methods: This three-week observational study included 90 participants: 19 meditation practitioners, 32 recreational runners, and 39 individuals without regular meditation or exercise habits. HRV was continuously recorded using Garmin smartwatches. Subjective stress and activity were assessed three times daily via a smartphone-based experience sampling method (ESM), yielding 4,557 responses; 632 meditation sessions were time-logged; questionnaires were administered at the end of the study. Results: Stress (PSS) was lower in meditation and running groups vs. controls (Kruskal–Wallis P = .024; adjusted Wilcoxon P = .048 and .040). RMSSD was higher in runners vs. controls ( P < .001) but not different between meditation and controls. Higher perceived stress was associated with concurrent RMSSD reduction of −2.24 ms (95% CI −3.97 to −0.26), steeper in runners (−3.94 ms; 95% CI −7.04 to −0.74). During meditation, RMSSD increased by +4.68 ms (95% CI 2.96 to 6.38) and remained elevated for at least 30 minutes post-practice. Conclusions: Although daily-life HRV among meditation practitioners was not elevated overall, their ability to increase HRV at will with a prolonged residual effect may relate to stress reduction and warrants further investigation. Trial Registration: Not applicable.

  • Open Access

    Daily Stress and Heart Rate Variability Among Mindfulness Meditation Practitioners: mHealth Observational Study (Preprint)

    by Jo Takezawa, Shixian Geng, Masahiro Fujino, Mika Miyake, Kazutoshi Sasahara, Koji Yatani, Atsushi Niida

    BACKGROUND Mindfulness meditation has been reported to reduce stress and enhance well-being. However, its effects on heart rate variability (HRV)—a physiological marker of stress—remain underexplored. OBJECTIVE This study aimed to examine how meditation practice is associated with subjective stress, HRV, and their interaction, using mobile health technologies. METHODS This 3-week observational study included 90 participants—19 meditation practitioners (meditation group), 32 recreational runners as an active control group characterized by lower stress and higher HRV (running group), and 39 individuals without regular meditation or exercise habits (control group). HRV was continuously recorded using Garmin smartwatches. Subjective stress levels and activity states were assessed 3 times daily through a smartphone-based experience sampling method, yielding a total of 4557 responses (mean 50.6, SD 22.8 per participant). From the meditation group, start and end times of 632 daily meditation sessions (mean 33.3, SD 18.3 per participant) were also collected via the app. Standardized questionnaires on stress and related measures were administered at the end of the study period. RESULTS The questionnaire survey confirmed that stress levels were significantly lower in both the meditation and running groups compared with controls (median Perceived Stress Scale scores: meditation 21, IQR 17-24; running 22, IQR 19-25; control 25, IQR 21-30; Kruskal-Wallis P=.02; adjusted Wilcoxon P=.05 and .04, respectively). Smartwatch-derived HRV (root-mean-square of successive differences [RMSSD]) was elevated in the running group relative to controls (median 47.0, IQR 44.0-54.2 and median 42.0, IQR 34.2-47.8, respectively; P<.001), whereas no significant difference was observed between the meditation and control groups (median 40.8, IQR 35.5-44.4 and median 42.0, IQR 34.2-47.8, respectively). Bayesian analysis of the experience sampling method data indicated that higher subjective stress levels were associated with a concurrent RMSSD reduction of –2.24 (95% credible interval –3.97 to –0.26) milliseconds. Although this pattern was consistent across groups, the steeper decline of –3.94 (95% CI –7.04 to –0.74) milliseconds was observed only in the running group, likely reflecting their elevated baseline HRV. Additionally, Bayesian modeling of 632 logged meditation sessions revealed an acute RMSSD increase of +4.68 (95% CI 2.96 to 6.38) milliseconds during meditation, with effects maintained for at least 30 minutes post practice. CONCLUSIONS Although HRV among meditation practitioners did not appear elevated in overall daily life, the ability to increase HRV at arbitrary timings, along with the prolonged residual effect, may correlate with stress reduction. This hypothesis requires further exploration with adequate controls. Despite the preliminary nature of this study due to its limited sample size, our findings highlight the potential of mobile health–based methodologies to capture stress and HRV dynamics in real-world settings.

  • Open Access

    Domain-specific emotional effects of deepfakes on news-sharing behaviour

    by Kei Ichikawa, Jiayu Chen, Kazutoshi Sasahara

    The rise of generative AI has raised societal concerns about the potential of synthetic media or deepfakes for misinformation. While previous research has established that negative emotions drive text-based misinformation sharing, their influence on deepfake dissemination remains poorly understood. Through two online experiments (N = 487; N = 479), we examined how emotional valence in deepfake images affects sharing intentions across political and entertainment news domains. Using GPT-3 and Stable Diffusion, we manipulated image emotional valence while maintaining consistent news contexts. Results revealed that negative images increased sharing intention only for political content, not entertainment content. Moreover, replacing negative images with less negative alternatives reduced sharing intention for political news, even among those initially inclined to share. These findings suggest domain-specific emotional dynamics in deepfake sharing and suggest novel content moderation strategies based on affective computing.