fNIRS Hyperscanning and Generative AI: The GAI Partner Effect in Science Learning

A BrainLatam2026 reading based on the abstract and appendices, on TSI, GSI, metacognition and educational Jiwasa

Before talking about generative AI in education, we need to talk about presence.

Learning is not only receiving the correct answer. Learning involves attention, doubt, error, reformulation, metacognition and connection. That is why the study “The GAI Partner Effect: A comparative study of interaction types on students’ learning engagement and academic performance in science education”, by Wang, Ren and colleagues, is highly relevant for BrainLatam2026.

The article compares two modes of science learning:

TSI — Teacher-Student Interaction, interaction between teacher and student;
GSI — GAI-Student Interaction, interaction between student and generative AI.

The central question is:

what changes in engagement, academic performance and neural dynamics when students learn with a human teacher or with a generative AI partner?

What the study showed

Based on the available abstract, the study used a within-subjects design and applied fNIRS hyperscanning to investigate cognitive and neural mechanisms associated with both paradigms.

The results indicate that interaction with generative AI (GSI) increased cognitive engagement more strongly. Teacher-student interaction (TSI) more effectively supported behavioural, emotional and social engagement.

This is very important. AI may require more processing, more self-regulation and more metacognitive effort. The human teacher, on the other hand, better sustains social presence, bonding, shared emotion and participation.

The study also shows that teacher-student brain synchrony was positively associated with social engagement, suggesting that human learning remains deeply relational.

Recognition of the scientific question

The merit of the article lies in not treating AI as a simple substitute for the teacher. The question is more intelligent: which dimensions of learning are strengthened by each type of interaction?

This design is valuable because it brings together education, generative AI, metacognition, engagement and fNIRS hyperscanning. It helps us understand AI not only as an answer tool, but as a cognitive partner that can modify how students think, monitor and reorganize their understanding.

Equipment and limits of this reading

The material available for this reading includes the abstract and appendices, but it does not provide complete details on the brand and model of the fNIRS system, number of participants, optode montage, measured brain regions, channels, acquisition software or full statistics.

What is explicit is the use of functional near-infrared spectroscopy — fNIRS hyperscanning and a pedagogical system in which ChatGPT was configured as an intelligent tutor. The prompt instructed the AI not to provide direct answers, but to work through questions, feedback, follow-up questions and summaries, supporting reasoning and metacognition.

BrainLatam2026 reading

From the BrainLatam2026 perspective, generative AI can support Zone 2 when it helps students think about their own thinking. It can ask, provoke, request explanation and support revision of reasoning.

But if it is used only to deliver fast answers, it may push students toward cognitive passivity.

The human teacher remains essential because they sustain educational Jiwasa: presence, bonding, listening, affective adjustment, bodily reading and belonging. AI can expand cognition, but the teacher organizes the human field where learning happens.

From the article’s question to a BrainLatam2026 design

The article asked:

how do TSI and GSI affect engagement, performance and neural mechanisms in science learning?

BrainLatam2026 can expand the question:

how can teacher, generative AI, body, metacognition and belonging form a Zone 2 educational ecosystem?

A future design could combine:

fNIRS hyperscanning + EEG/ERP + eye-tracking + HRV/RMSSD + respiration + GSR + language analysis + academic performance + belonging measures.

fNIRS hyperscanning would measure teacher-student and student-student synchrony.
EEG/ERP could capture error, surprise, semantic conflict and conceptual updating.
Eye-tracking would show how students distribute attention between teacher, screen, AI and learning material.
HRV/RMSSD, respiration and GSR would help measure effort, bodily safety and emotional regulation.
Language analysis would indicate whether students are merely repeating answers or truly reorganizing concepts.

A generous decolonial critique

In Latin America, the question cannot be only whether AI improves performance. The question must be:

under which social, affective, technological and pedagogical conditions does generative AI improve learning without weakening teacher, body and belonging?

If AI is used to replace teachers and reduce costs, it may weaken educational Jiwasa. If it is used to support teachers, expand questions, personalize feedback and strengthen metacognition, it can become a Zone 2 technology.

DREX Cidadão and education

Generative AI does not solve inequality by itself. To learn, students need internet, food, safety, sleep, time, valued teachers and living schools.

DREX Cidadão, as a minimum economic metabolism distributed to the social body, enters here as a foundation for belonging. It can reduce material urgencies that hijack attention and make learning harder.

Technology without belonging becomes cold automation.
Belonging without technology may lose power.
Public policy needs to unite both.

Closing

The study by Wang, Ren and colleagues shows that the future of education does not depend only on smarter machines, but on the quality of interactions.

AI can expand cognitive engagement.
The teacher can sustain social and emotional engagement.
fNIRS hyperscanning can show when learning becomes synchrony.
And Jiwasa reminds us that learning was never only about accumulating information.

The great question is not whether AI will replace the teacher.

The question is:

how can teacher, student, AI, body and territory form a learning ecosystem with more consciousness, belonging and cognitive freedom?

Reference

Wang, J., Ren, J., Liu, W., Rong, Z., Shi, Z., Zhao, Y., Zou, S., & Gao, S. (2026). The GAI Partner Effect: A comparative study of interaction types on students’ learning engagement and academic performance in science education. Journal of Science Education and Technology. https://doi.org/10.1007/s10956-026-10311-x