I work on AI Alignment & Safety at Sony Computer Science Laboratories (Sony CSL), focusing on Model Psychology, Artificial Metacognition (especially in large language models), and Consciousness Science. My work includes studies on LLMs' metacognition (Yoshizawa et al., 2026) and the relationship between Metacognitive Closure and consciousness in LLMs (Yoshizawa & Mogi, 2026).

I also lead the Metacognition & Hallucination Team at Shiba AI, a Tokyo-based organization working on AI safety, alignment, and governance. 

Alongside my AI research, I am an undergraduate physics student at Tokai University, where I conduct research in Quantum Thermodynamics under the supervision of Dr. Eiki Iyoda.

I serve as an Expert Consultant in AI evaluation and safety at Duco, a company based in San Francisco, USA.

CV

Papers(Journal/Proceedings) 

  
[1] Yoshizawa S, Onzo A, Nozawa S, Takano T, Ishikawa T and Mogi K (2026) Metacognition of ChatGPT in confidence judgements. Front. Artif. Intell. 9:1694192. doi: 10.3389/frai.2026.1694192

[2] Yoshizawa, S., & Mogi, K. (2026). Metacognitive Closure and Consciousness in Large Language Models. Proceedings of the AAAI Symposium Series, 8(1), 380–390. https://doi.org/10.1609/aaaiss.v8i1.42569

International Conferences

[3] Yoshizawa, S. and Mogi, K.(2026), “Do Large Language Models Have a Metacognitive Axis?”, Society for Neuroscience annual meeting, November, Washington, D.C., USA. (accepted)
[4] Inoue, K., Tamori, Y., Yoshizawa, S., Mogi, K.(2026), “Persona Trajectory across Multi-Turn LLM Dyadic Conversations”, Society for Neuroscience annual meeting, November, Washington, D.C., USA. (accepted)
[5] Yamada, M., Yoshizawa, S., Inoue, K., Tamori, Y., Mogi, K. "The Impact of Perceived Affective Alignment (PAA) on University Students’ Academic and Cognitive Performance", Society for Neuroscience annual meeting, November, Washington, D.C., USA. (accepted)
[6] Ariga,  K., Yoshizawa, S., Mogi, K. "Survival Pressure as a Necessary Condition for Semantic Grounding in Coupled Sensorimotor-Linguistic Systems" , Society for Neuroscience annual meeting, November, Washington, D.C., USA. (accepted)
[7] Horie, T., Tamori, Y., Yoshizawa, S., Mogi, K. "Cross-domain analogy as lossy compression: a graph-based model of useful abstraction and culturally stabilized false belief dynamics", Society for Neuroscience annual meeting, November, Washington, D.C., USA. (accepted)

[8] Yoshizawa, S. and Mogi, K., (2025), “Metacognition in Large Language Models and Humans in Post-Decision Wager”, Society for Neuroscience annual meeting, November, SanDiego, USA 

[9] Ito, Y., Onzo, A., Takano, T., Tamori, Y., Yoshizawa, S., & Mogi, K. (2025). “Aspects of uncertainty cognition in AGI/ASI capabilities and their Implications for AI regulation.”, Society for Neuroscience annual meeting, San Diego, USA 

[10] Inoue, K.,Yoshizawa, S., Tamori, Y., Mogi, K., (2025), “Characterization of LLMs with the Big Five factors of personality”, Society for Neuroscience annual meeting, November, SanDiego, USA 

[11] Yoshizawa, S. and Mogi, K., (2024), “Consciousness and Intelligence in Large Language Models”, Association for the Scientific Study of Consciousness annual meeting, July, The University of Tokyo, Japan

[12] Yoshizawa, S. and Mogi, K., (2024), “Symmetry and asymmetry in human-AI cooperation”, Society for Neuroscience annual meeting, October, Chicago, USA

[13] Yoshizawa, S. and Mogi, K., (2023), “Natural and artificial metacognition in humans and Large Language Models”, Society for Neuroscience annual meeting, November, Washington, D.C., USA

Domestic Conferences

[14] Yoshizawa, S. and Mogi, K., (2025), “Cooperative behaviour in Large Language Models and Humans”, The Japanese Society for Artificial Intelligence annual meeting, July, Osaka, Japan

[15] Tamori, Y. and Yoshizawa, S., Mogi, K.,(2024), “Towards building an AI alignment paradigm”, The Japanese Society for Artificial Intelligence annual meeting, July, Shizuoka, Japan

[16] Yoshizawa, S. and Mogi, K., (2024), “Metacognition in Large Language Models”, The Japanese Society for Artificial Intelligence annual meeting, July, Shizuoka, Japan

Hackathon

• Meader, A., Che, M., Yoshizawa, S., Li, Y. LLM Military Decision-Making Under Uncertainty: A Simulation Study, Women in AI Safety Hackathon, Preprint at Apart Research, (2025)

•  Metacognitive Closure and Consciousness in Large Language Models-AAAI 2026 Spring Symposium on Machine Consciousness in Burlingame, CA, USA. April 7–9, 2026. (Oral).

• Metacognition in Large Language Models
  -Future day 2025, online, Feb 28th – March 2nd, 2025. The Youtube Video

• Excited to share that I’ve been selected for the Introduction to Digital Minds course by Cambridge Digital Minds, as one of just 12% of applicants selected from this year’s applicant pool!

• My talk at the AAAI Spring Symposium is now available online!

• I gave a lightning talk at the AAAI 2026 Spring Symposium on Machine Consciousness (April 7–9, 2026) in Burlingame, CA, USA.

• I am pleased to share that my first-author paper has been accepted for publication in the AI for Human Learning and Behavior Change section of Frontiers in Artificial Intelligence.

• I did a poster presentation at the 39th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI2025) in Osaka.  Cooperative Behavior in Large Language Models and humans

• I completed the Center for AI Safety course on AI Safety, Ethics and Society (AISES) by

Dan Hendrycks. This is the certificate.

• 2025/4/4 : We had a great time discussing AI Safety and Governance with former Prime Minister Yukio Hatoyama on The Way of AI(AIの道) podcast at UT-LAB. 

• 2025/3/12 : I launched "The Way of AI (AIの道)", a podcast with Dr Ken Mogi and lawyer Yoshie Ito at the UT-lab near the University of Tokyo in Japan. This podcast covers important topics in Alignment, Safety, Governance and Economics related to AI.

• 2025/3/7-9: I participated in the "Woman in AI Safety Research Hackathon". With the great teamwork of Ashley Meader, Mingli (Alondra) Che and Yingquan Li, we submitted our paper on "LLM Military".

• 2025/3/1: I presented at "Future Day 2025" on Metacognition in LLMs. This is my first public talk. Thank you for a wonderful opportunity! The YouTube Video was published.