1. What struck me most, and inspired me most, in the Redbook
What struck me most is that the Redbook makes the capacity for cognitive autonomy the overarching goal of AI general education. This framing captures the essence of AI general education. When frontline teachers teach AI, it is all too easy to end up with tool operation, technology experiences or showcasing student products. The Redbook helped me see more clearly that AI general education is not only about teaching students to "use AI" but to "use AI well": when AI takes part in how information is gathered, how judgments are formed and how decisions are made, students should be able to actively regulate sources, evidence, boundaries, trust and responsibility.
This fits closely with the shift in my own teaching over recent years: from "making a product" to "seeing the real society", from a product orientation to an orientation towards educating the person. Whether in my interdisciplinary AI teaching cases "When AI Enters a Mock Public Hearing" and "Feeling the Pulse of 'China' in Bronze Vessels with AI", or in "Creating My Own AI Table-Tennis Coach", technology is never merely a tool; it is a vehicle through which students understand real problems, form judgments and take responsibility.
2. How the Redbook's ideas guide teaching and curriculum practice
The six dimensions, the four capacities, the five-part content, and "risk, ethics and responsibility" running through the whole, give me a framework and a yardstick for assessment when designing interdisciplinary lessons.
In the "Creating My Own AI Table-Tennis Coach" lesson, for example, students collect stance data, train a model, test for misjudgments and refine their data, going through the full "data–model–output–evaluation" loop. That is exactly a training process in induction, generalization, judgment, and self-awareness. The debate on "Can AI replace a table-tennis coach?" then embeds responsibility judgment and trust calibration in real practice, rather than in empty preaching.
In the AI + Morality and Law lesson "When AI Enters a Mock Public Hearing", I set up a "mock public hearing" on the Zhiqi platform. The "Mock Hearing Assistant" agent I built is a mock-hearing and debate-sparring tool designed around the issue of "adding elevators to old apartment buildings". What is new about it is its "two lines, two directions" structure of dialogue and inquiry. The first line is human–machine debate. Students take the affirmative or the negative side and argue back and forth with the agent, practising making claims, answering challenges and rebutting the other side. They can also switch on AI self-play: the agent generates the arguments of both sides on the same point of contention, while students can continue, pause, switch to another point or stop at any time, and, as observers and reviewers, judge whether either side shows insufficient evidence, conceptual confusion, bias in position or gaps in reasoning.
The second line is the mock hearing itself. Students play roles such as upper-floor residents, lower-floor residents, a sub-district representative, a technical expert and a legal expert; they prepare statements and take questions. They can switch positions at any time and temporarily stand in the other side's shoes, so that they do not see only one point of view.
The "two directions" lie in breaking the points of contention down round by round. Daily mobility needs, lighting and noise, privacy, cost sharing, compensation mechanisms, construction safety… each concrete issue has to be gone through from both sides. The expression-training segment helps students distinguish facts, opinions and legal grounds, and organize their own statements, rebuttals and summaries.
In class, students have to question sources, compare evidence, identify positions and weigh the rights of different parties, which corresponds to what the Redbook calls source vigilance, evidence sensitivity, boundary awareness, trust calibration and responsibility judgment.
In the AI + History lesson "Feeling the Pulse of 'China' in Bronze Vessels with AI", AI text-to-image generation, Metaso AI, a 3D model of the He Zun vessel and an AI platform for bronzes are used to train evidence-based historical reasoning and source checking: generated content must not be taken directly as historical fact.
The Redbook's positioning of lower secondary school as "building systems and a global perspective" also helped me lift my AI lessons from technology experiences to an understanding of mechanisms and social judgment, avoiding "using AI for the sake of using it".
3. Comments and suggestions for revising the Redbook
The Redbook is heading in exactly the right direction, and it gives us frontline teachers academic leadership in moving from "technology as a tool" to "the goal of educating people". Drawing on my experience of developing interdisciplinary lessons in information technology, morality and law, history, physical education, labour education and biology at the High School Affiliated to Renmin University of China, I would like to offer a few suggestions:
1. I hope that, within its existing framework and assessment system, the Redbook will include more outstanding cases from frontline basic education, especially in subjects such as morality and law, history, Chinese and physical education. At present, when many teachers hear "AI general education", their first reaction is still that it is "the information technology teacher's job", or that it simply means programming and robotics. In fact, interdisciplinary lessons like mine can all carry training in cognitive autonomy. I hope the Redbook will use such cases to show everyone that AI general education does not belong only to information technology and the sciences: other subjects can do it too, and do it in a way that truly educates.
2. I hope the observation points and assessment tools for cognitive autonomy can be refined further, ideally into templates that frontline teachers can pick up and use, such as an AI-use log, an evidence-checking sheet, a data-refinement record, a responsibility statement for student work and a debate-performance rubric. Doing all this is hard, but with such tools teachers could see in class whether students have checked their sources, stated what AI contributed, refined their data in response to misjudgments, and reflected on their own trust and reliance.
3. For specific scenarios such as generative AI, AI digital humans and camera-based data collection, I hope the Redbook will give boundary guidance that is more specific to each school stage and each scenario: how to use them in primary school and in lower secondary school, what can and cannot enter the classroom, what requires parents to be informed, what data must not be uploaded, how to handle copyright and deep-synthesis labelling, and when ethical review should take place. If the Redbook could set out clear "boundaries of use" and "operational advice", frontline teachers would feel much more confident.
In short, the Redbook is heading in exactly the right direction and gives us frontline teachers academic leadership in moving from "technology as a tool" to "the goal of educating people". I look forward to future editions becoming a handbook that teachers can consult when preparing lessons, use in class and refer to when reviewing lessons. The Redbook was released just in time for Teachers' Day 2026; my thanks to the Center for Artificial Intelligence General Education, Department of Computer Science and Technology, Tsinghua University, for this great gift to teachers across the country.