The GASA team demonstrates its educational AI robot at the Tsinghua–Penn forum.
On October 16–17, 2026, GASA founders Lijie Chen and Peng Lin visited Tsinghua University in Beijing to discuss the latest development of the Murmur educational AI system and the GASA educational robot. They also participated in the “Education·Technology·Innovation” Forum hosted by Tsinghua University’s School of Education and the University of Pennsylvania Graduate School of Education (Penn GSE).
Against the backdrop of generative AI’s impact on independent learning, the team raised a question worth continued exploration: now that AI can generate answers quickly, can it do more to accompany students as they think and help them understand their own learning processes, rather than think for them?
From the Tsinghua Lab to Real-World Learning Environments
On October 16, the GASA team visited Professor Liu Lin’s lab at Tsinghua University’s School of Software, where they discussed the latest development of Murmur with the research team and shared updates on the GASA educational robot. Key topics included natural voice interaction, observation of learning behaviors, analysis of learning psychology, and user experiences across different educational settings.
In GASA and Murmur’s product vision, learning analytics should go beyond answer accuracy and final scores. How students express their ideas, why they change their judgments, when they need a prompt, and how they reflect on their learning strategies are also worth understanding. Through natural conversation, thoughtful follow-up questions, and records of the learning process, the team hopes to explore a more continuous form of support that interrupts learning less.
The GASA team discusses the latest development of Murmur with the R&D team at Tsinghua University’s School of Software.
On October 17, the GASA team joined Professor Liu Lin at the Tsinghua - Penn GSE “Education·Technology·Innovation” Forum, where they demonstrated the GASA educational robot and the Murmur educational AI system. The robot supports natural voice conversations, English speaking practice, and interactive learning activities, while Murmur explores combining learning behavior signals, educational psychology models, and intelligent feedback to provide learners with more tailored support.
The demonstration attracted the attention of scholars attending the forum. Professor Shi Zhongying, Dean of Tsinghua University’s School of Education, expressed interest in further discussion after the event. Professor Michael Nakkula of Penn GSE also found the direction interesting and continued the conversation with the team.
The forum jointly hosted by Tsinghua University’s School of Education and the University of Pennsylvania Graduate School of Education.
The GASA educational robot and learning interface on display at the forum.
Can AI Help Without Thinking for Students?
Generative AI has drastically lowered the barrier to finding answers and completing tasks, prompting educators to pay closer attention to the potential risks it poses to students’ ability to think independently. GASA aims to explore a different role for AI: shifting it from an “answer generator” to a “companion and guide throughout the learning process.”
In natural conversations, AI can ask questions such as “Why do you think that?”, “What evidence supports your view?”, and “Could there be another explanation?” to encourage students to examine their judgments and learning methods. This kind of interaction is closely related to metacognition in educational psychology—the awareness, monitoring, and adjustment of one’s own thinking and learning processes.
With respect for privacy and a commitment to scientific validation, the team hopes to explore how meaningful evidence of the learning process can be recorded to help teachers better understand students and help students gradually develop habits of reflection and self-regulation.
Must Educational Responsibility and Commercial Value Be at Odds?
The conversation with Professor Michael Nakkula extended to the long-term responsibilities of educational AI. He noted that commercial objectives can put pressure on the effectiveness and responsibilities of educational products: when growth, efficiency, and user experience become key metrics, can a product still stay true to its original purpose of supporting students’ long-term development?
The GASA team proposed another possibility: if an AI product can help students build better learning habits, give teachers more valuable evidence of the learning process, and help parents understand genuine growth, could educational outcomes themselves become the foundation for a sustainable product?
Both parties engaged actively in an exchange of views on this idea. For GASA, the business model worth validating is not one that simply seeks to increase how often students use AI, but one that continually improves AI’s real value to students’ growth.
After the forum, the GASA team continued discussing educational AI research and applications with Professor Michael Nakkula.
From Dialogue to Future Interdisciplinary Collaboration
After the forum, Bruce Zou, a senior research associate at the University of Pennsylvania, held further discussions with the GASA team. He expressed interest in potential applications of GASA’s educational robot in integrating course content and pedagogical approaches from Penn GSE’s Global Possibility Network. Both sides hope to explore specific approaches during subsequent discussions. These conversations remain at an exploratory stage.
Educational AI is not simply a matter of connecting a language model to a course. Learning sciences, developmental psychology, curriculum design, human–computer interaction, and AI engineering all need to be tested together in real-world settings. For GASA, this exchange in Beijing further strengthened its commitment to this interdisciplinary exploration.
The GASA team and Professor Michael Nakkula after the forum.
From “Answer Intelligence” to “Process Intelligence”
As more AI systems improve the speed and quality of answers, GASA and Murmur aim to address a different set of questions: How did a student arrive at an answer? Which prompts supported understanding? What kind of feedback can help the student perform better independently next time?
From discussions on R&D at Tsinghua University to product demonstrations and academic dialogue at the Tsinghua–Penn forum, this evolving small robot offers a tangible starting point for exploring a new relationship with learning. Going forward, GASA will continue developing its products and research collaborations around natural interaction, learning-process analytics, metacognitive support, and verifiable educational outcomes.
We are not building AI to think for students. We are building AI that helps students see how they think.
About GASA | GASA brings together artificial intelligence, learning sciences, and real-world educational settings to explore innovative applications in educational AI, learning-process analytics, voice interaction, research training, and project-based learning.