Extending Teacher Capabilities with Generative AI

In recent years, generative artificial intelligence has transitioned from conceptual discussions to practical applications in education, presenting both opportunities and challenges for teachers. In November 2025, the Expert Guidance Committee for Teacher Workforce Development of the Ministry of Education released the “Guidelines for the Application of Generative AI by Teachers (First Edition),” which serves as the first framework for educators to effectively utilize generative AI in their teaching practices. For frontline teachers, the challenge lies in translating macro scenarios into daily practices, requiring them to explore how to adapt advanced technology to meet educational needs and connect technology with the essence of education, thereby extending teachers’ capabilities.

How Teachers Can Use Generative AI to Achieve Learning Transformation

The core of learning transformation is to shift students from being passive “knowledge receivers” to active meaning-makers and explorers. Due to the convenience and accessibility of AI, many teachers initially view generative AI simply as an “advanced search engine” or “answering machine,” allowing students to ask questions and receive conclusions directly. This approach not only proves unhelpful but also stifles students’ thinking processes. True transformation lies in using generative AI to create open-ended problem situations, guiding students to “discover” knowledge through inquiry, rebuttal, and diverse perspectives.

Strategy 1: Dialogue-Based Learning

Utilize the dialogue-based learning function of generative AI to help students dynamically understand learning concepts. For example, in a biology class, a teacher can set a dialogue theme with generative AI like “Explore the survival strategies of the North China leopard alongside Darwin,” instead of having students memorize knowledge points like “factors affecting population size.” Students can assume the role of Darwin in 19th century China, observing a small population of North China leopards and engaging in dialogue with generative AI acting as an “ecologist assistant.” They would describe the characteristics and habitat of the leopards and propose initial hypotheses. The generative AI, responding in a scientific tone, would prompt students with questions about the impact of prey population fluctuations on “survival struggles” and the manifestation of “survival of the fittest” under human activities. This dialogue encourages students to reason and refine their models using concepts like food chains and interspecies relationships, fostering deeper connections rather than isolated memorization.

Strategy 2: One-on-One Deep Interaction

Use AI’s one-on-one deep interaction capabilities to cultivate students’ thinking skills. In a lesson on “genetic engineering,” teachers can have students role-play as “research engineers in a biotechnology company,” tasked with improving a crop using gene editing technology, while generative AI acts as a project consultant and interacts as an “ethics review committee.” For instance, students might ask the virtual consultant, “I want to design an insect-resistant corn; which target gene should I choose? What vector and tool enzyme are most suitable?” Instead of providing direct solutions, the consultant would counter with questions like, “Considering that target pests may evolve resistance, is your single-gene strategy sustainable? What are the conversion efficiency and technical challenges of a multi-gene stacking strategy?” After formulating initial plans, students must submit their proposals for review by the ethics committee, with generative AI raising questions from perspectives of biological safety, ecological impact, and social ethics.

This process transforms cold technical steps into real projects filled with trade-offs and decision-making. Knowledge acquisition becomes problem-solving, and memory burdens shift to design challenges. More importantly, generative AI provides deep one-on-one interactions that traditional classrooms cannot offer, allowing each student to learn at their own pace.

In summary, when using generative AI to improve student learning methods, three key aspects must be considered: 1) Role-based design, where the intelligent agent is assigned a clear role (such as scientist, historical figure, debate opponent, or project consultant) to enhance immersion and purpose in human-computer interaction. If precise settings are challenging, teachers can input general ideas for the system to refine. 2) Focus on problem chain-driven learning, where teachers design guiding question chains in advance, incorporating generative AI dialogues into structured inquiry processes to avoid aimless chatting. 3) Maintain process records, requiring students to save key dialogue records with generative AI and explain their thought evolution as a basis for process evaluation.

Enhancing Teaching Quality with Generative AI

The key to improving teaching quality lies in transitioning from vague, experience-based judgments to precise, data-driven interventions based on student learning conditions. Generative AI can assist teachers in data-enhanced decision-making and intelligent resource generation, moving lesson preparation and delivery from a “handicraft” era to an “intelligent assistance” era. However, it is crucial to avoid using generative AI to produce lengthy, polished content that diverges from student learning conditions and teaching objectives. Teachers’ core responsibilities should involve inputting precise instructions (learning conditions, goals, scenarios) and professionally screening, adapting, and integrating the content generated by generative AI.

Strategy 1: Efficient Assignment Design

Utilize precise instructions to save time on assignment design for targeted student support. For high school teachers, efficiently addressing students’ rapid improvement needs within limited time is critical. During the senior year review phase, students often show significant differentiation, requiring teachers to prepare tiered exercises. Previously, teachers would spend considerable time sifting through numerous exercise books, which was labor-intensive and lacked specificity. AI can effectively alleviate this challenge. For instance, after each unit test, teachers can input detailed student performance data into office software, using its built-in generative AI analysis feature to issue composite instructions: 1) Calculate the error rates for three core concepts across the class. 2) Generate five basic consolidation questions for students with error rates above 50% (focusing on concept differentiation); five ability enhancement questions for those with error rates between 20% and 50% (focusing on comprehensive application); and two exploratory questions for students with error rates below 20% (linking to current events). 3) Each question should include a brief explanation and knowledge point source (textbook chapter).

Generative AI can produce a clear, structured draft of tiered exercises in minutes. Teachers can then replace one or two question scenarios with recent class discussions to make the questions more relatable. This significantly enhances teachers’ productivity, freeing up time for analyzing individual students’ unique error causes and providing face-to-face guidance.

Strategy 2: Transforming Memorization Tasks

Assign students to turn rote memorization tasks into active knowledge construction tasks. When teaching complex concepts, teachers can assign weekend homework where students create “visual knowledge maps” to make their thought structures visible. For instance, students might receive a starting instruction: “Construct a mind map centered on the relationship between cellular aging, apoptosis, cancer, and individual health.” They would also be informed about how to use AI tools to generate mind maps: input preliminary ideas into a generative AI model, generate an initial framework in Markdown format, and import it into software like Xmind to create visual mind maps. During this process, teachers and generative AI act as “logic reviewers,” checking for inaccuracies in concept relationships, missing key connections, or chaotic hierarchies and providing modification suggestions. Students can adjust their mind maps in real-time based on feedback, and they can share their generated mind maps in class groups. This approach results in each student creating a unique knowledge structure map, refined through their own thinking and AI assistance, which is far more effective than rote memorization of pre-existing frameworks, transforming the knowledge construction process into a deep understanding of the material.

In AI-assisted teaching improvements, three principles must be followed: 1) Data-driven student learning conditions. Transforming students’ assignments and test data into structured information for generative AI analysis is a prerequisite for precise teaching. 2) Instruction precision. The more specific the instructions given to generative AI (target, scenario, format, requirements), the more aligned the generated content will be with teaching realities. 3) Emphasize teacher re-creation. The output from generative AI is a “rough draft” that requires teachers to contextualize, evaluate, and pedagogically process it, which is an irreplaceable professional step.

Enhancing Evaluation Efficiency with Generative AI

Traditional evaluations are time-consuming and often limited to scores and correctness. Generative AI can quickly process large volumes of text and data, enabling process analysis, qualitative feedback, and attribution suggestions, deeply analyzing learning processes and providing personalized improvement paths. Teachers should not merely list incorrect knowledge points but guide generative AI to focus on deeper dimensions such as thought processes, argumentative logic, and expression structure to assist in effective student evaluation.

Strategy 1: Identify Common and Individual Issues

Use generative AI to identify common and individual issues within the class and generate diagnostic reports to support precise tutoring and teaching improvements. When conducting academic diagnostics, teachers can systematically leverage generative AI tools. First, input detailed, targeted evaluation criteria and instructions, such as, “As a biology teaching assistant, analyze this lab report: point out its strengths and weaknesses in ‘problem and hypothesis formulation,’ ‘rigor of design,’ ‘data recording and processing,’ and ’logical conclusion derivation.’” Next, for areas where students commonly make mistakes, issue further improvement instructions, such as, “Provide a specific modification suggestion segment for the weaknesses in the ’logical conclusion derivation’ section” or “Based on the student’s overall report performance, infer potential habitual weaknesses in scientific inquiry thinking and suggest a follow-up practice.”

In this case, generative AI’s feedback not only highlights superficial issues like “the conclusion lacks validation through repeated experiments” but also analyzes deeper problems: “The student jumped directly to conclusions when analyzing data, lacking elaboration on the logical chain. It is suggested to supplement: …” Such feedback addresses the essence of scientific thinking, achieving guidance that traditional evaluations of “right” or “wrong” cannot provide.

Strategy 2: Focus on Higher-Order Skills

Emphasize students’ thinking levels and creative expression in evaluations, conducting in-depth reviews and guidance to achieve human-AI collaborative feedback. By utilizing AI, teachers can generate personalized model essays for deeper guidance. In the past, teachers had limited and non-specific model essays; now, generative AI can produce model essays that precisely match students’ original intentions and expression logic, retaining their writing characteristics while demonstrating how to deepen themes and optimize structures. This “instruction based on writing” intelligent feedback allows students to see improvement opportunities through comparison rather than facing wholesale rejection of their original ideas. Furthermore, when generative AI-generated model essays closely align with students’ original texts, students are more likely to resonate and be inspired. This precise, personalized guidance is transforming the traditional challenges of writing instruction, where feedback is often superficial and delayed, into a process of cognitive enhancement.

One key aspect of AI-assisted evaluation is structuring evaluation dimensions, converting core competencies into actionable generative AI analysis instructions. For example, breaking down macro competency requirements like “language construction and application” and “thinking development and enhancement” into specific, analyzable dimensions such as “topic understanding and intention,” “structural logic,” “language expression,” “material application,” and “thought depth,” and designing precise generative AI prompts for each dimension ensures that intelligent grading is evidence-based and results are structured and comparable. Additionally, feedback must be actionable, shifting from “judging right or wrong” to “providing modification paths.” For instance, instructing generative AI to avoid vague comments like “content is hollow” or “structure is chaotic” and instead request specific modification suggestions like, “Please supplement a relevant factual argument based on the original viewpoint” allows generative AI to provide concrete improvement paths. The core goal is to utilize generative AI to generate personalized upgraded model essays based on students’ original texts, enabling students to visually see the specific paths for improvement through comparison. Furthermore, accumulating and tracking data will establish dynamic development records of students’ writing abilities. By using spreadsheet tools to automatically archive structured data from each generative AI grading session, including scores across dimensions, prominent problem types, and modification suggestion keywords under each student’s name, a long-term accumulation forms a “writing ability development curve” for precise individual tutoring and group teaching adjustments, providing students with visual growth feedback.

Generative AI requires teachers to deeply understand the essence of their subjects, keenly observe students’ thinking, and meticulously design teaching processes. Only when teachers can effectively harness technology can they sow infinite futures for every child in the limited field of education.