Tohoku University Cognitive Neuroscience Application Center

Generative AI has rapidly become part of second language (L2) learning, yet little is known about how AI-generated feedback affects learners’ emotions and brain activity. In particular, the neurocognitive impact of explicit versus implicit corrective feedback from AI remains unexplored.
In this study, we used a newly developed real-time fMRI–ChatGPT paradigm that allows natural turn-by-turn conversation with an AI during scanning. Results showed that learners with higher communication anxiety or fear of negative evaluation exhibited stronger amygdala responses to both explicit and implicit feedback, highlighting that AI corrections can carry social-emotional salience. 
This work was presented at the Australasian Cognitive Neuroscience Society (ACNS 2025) in Melbourne, under the title “How explicit and implicit feedback shape L2 learners’ anxiety in real-time communication: An fMRI study.” Although many sessions focused on clinical topics, the use of AI in cognitive neuroscience attracted considerable interest, and the discussions emphasized the growing international attention toward AI–brain research.
During the conference, I also had the opportunity to exchange ideas with Dr. Wei-Peng Teo’s team from Nanyang Technological University. Their group uses diverse methodologies—including EEG, fNIRS, and TMS—to investigate cognitive mechanisms related to learning. I found strong relevance between our research themes, and the exchange was highly meaningful.(Chunlin)

Chunlin Liu, Shoma Tanaka, Satsuki Kurokawa, Taira Nakajima, Takumi Uchihara, Hyeonjeong Jeong, Motoaki Sugiura
How explicit and implicit feedback shape L2 learners’ anxiety in real-time communication: An fMRI study
Australasian Cognitive Neuroscience Society 2025,Melbourne, Australia, November 25-28,2025 【 Poster 】
https://www.acns.org.au/acns-conference-2025/