目前我国各地区在开展人工智能教育时，教学内容呈现出以下三个特点：以人工智能某个知识点或必修教材某一部分的内容为主，系统的课程较少；大部分学校开设的人工智能课程以人工智能基础为主，未涉及人工智能相关算法或模型；理论层面的探讨居多，教学实践较少且集中在少数地区的一线教师和硕士研究生。教学方法呈现以下两个特点：大部分教师采用项目式学习教学模式开展教学，少数教师基于STEM、创客等理念设计、实施课程。

在国家大力发展人工智能教育的背景之下，为满足本校学生对人工智能的学习需求，落实信息技术课程标准和省教育厅对信息技术选择性必修课程的要求，本研究设计出了一门基于项目式学习的人工智能课程，采用行动研究法，面向14名高一学生实施课程，对课程效果进行了验证，并在实施过程中不断优化课程设计。本研究的主要研究工作和研究发现包括以下五个部分：

第一，学习者分析。调查兰州市E中学271名高一学生对人工智能的学习兴趣和学习基础，发现学生对人工智能的学习兴趣较强，24%的调研对象从六门信息技术选择性必修课程中选择了“人工智能初步”这门课程；学生的人工智能学习基础整体十分薄弱：超过60%的学生在高一之前没有学习过任何一门程序设计语言，只有4名同学能正确说出至少一个人工智能算法或人工智能术语。

第二，课程设计。基于人教版《人工智能初步》教材设计35课时的人工智能项目式课程，分为32课时的教学和3课时的考试，包括10个小组项目。本研究中的项目式学习实施步骤分为以下六步：确定选题、制定计划、实施计划、成果交流、师生评价、完善改进。

第三，学习评价设计。学习评价主要包括小组项目汇报和个人课程考试两个方面。本研究面向14名高一学生开展了35课时的人工智能教育教学活动。课程中，5个小组均完成了“人脸识别”、“调查兰州的人工智能应用”、“自选项目”这三个项目。课程考试主要包括“人工智能技术基础算法的原理和应用”、“人工神经网络模型的原理、搭建和应用”、“人工智能产品应用与未来发展”、“Python程序设计基础”四部分，共25道题目。

综合课堂表现、小组项目汇报情况和考试结果，有7位同学基本达到了信息技术课程标准中对于选择性必修4模块的学业要求：计算思维方面，能够使用Keras开源框架搭建出简单的智能系统；数字化学习与创新方面，在课程中、考试中实现了多个人工智能应用的案例，解决了生活中的实际问题；信息意识和信息社会责任方面，学生能够意识到人工智能就在身边、发现身边的人工智能产品，客观认识人工智能技术对社会、法律等方面的影响。

第四，提高学生自主学习的教学策略设计。本研究开展的人工智能课程中学生自主学习情况总体较好。对9位学生进行半结构化访谈后发现：学习动机维度，学生总体上都可以接受课程的难度，在完成课堂任务和小组活动的时候学生的自我效能感较强，认为教师安排的案例具有实际意义和价值，这门课程对自己未来发展有较大的帮助。学生的学习兴趣浓厚，课程结束后同学们的收获较大。学习策略维度，每位同学都采取了“上网查找资料”、“看课程回放”、“编写并保存课堂程序”这三种学习策略。学生的学习目标不够明确或者没有设定学习目标，在完成项目的时候较少同学具有制定项目计划的意识和行为。学习情境维度，同学们普遍认为课程期间教师、组员、其他同学对自己学习和项目完成的帮助较大，接受小组合作完成项目的这种学习方式。

针对学生的自主学习情况，教师在开展基于项目式学习的人工智能课程时可以从以下三个方面提高学生的学习体验、培养学生的自主学习意识：1.录制课程，学生普遍有看回放的需求；2.设计丰富有趣的案例，包括课堂任务和小组项目，教师应为每个项目提供指导性材料，帮助学生完成项目；3.加强对学生制定个人学习计划和小组项目计划的引导。

第五，学习环境设计。本研究发现扣叮平台具有方便教师演示和学生编写程序、安装Python外部模块时速度较快、自带TensorFlow等人工智能模块、运行速度较快、支持小组在线协作等优势。扣叮平台也有以下两个缺点：1.每次退出之后需重新安装之前安装的模块；2.部分自带的Python模块不支持教材中的案例，例如人脸识别案例，而Lightly可以很好的完成人脸识别案例的编译运行。因此，对机房机器较为落后的学校或对部署人工智能开发和编译环境存在困难的教师，教学时可以使用IDLE、扣叮、Lightly等多种教学实验环境。

本研究的主要创新点是设计出了一门较为系统的人工智能课程；探究了人工智能项目式课程中学生的自主学习情况，为教师提供了能够提高人工智能项目式课程中学生自主学习能力的教学策略；通过教学实践总结出扣叮平台在开展人工智能课程教学时的优势和不足。本研究设计的人工智能项目式课程具有基础性、生活性、实践性等特点，可以为当地有意愿开展人工智能课程教学的一线教师提供参考。

At present, when AI education is carried out in various regions of China, the teaching content presents the following three characteristics: a certain knowledge point of AI or a part of the compulsory textbook is the main content, and there are fewer systematic courses; most of the AI courses offered by schools are based on the fundamentals of AI and do not involve AI-related algorithms or models; theoretical level is mostly discussed, and teaching practice is less and concentrated in a few regions of Theoretical discussions are predominant, and teaching practice is rare and concentrated in a few areas of front-line teachers and master's students. The teaching method shows the following two characteristics: most teachers use PBL teaching mode to carry out teaching, and a few teachers design and implement courses based on STEM, Maker and other concepts.

In the context of national efforts to develop artificial intelligence education, in order to meet the students' learning needs for artificial intelligence, implement the information technology curriculum standards and the requirements of the Provincial Department of Education for optional compulsory courses in information technology, this research designed an AI course based on PBL, and used the action research method to implement the course for 14 senior one students to verify the course effect, and constantly optimize the curriculum design in the implementation process. The main research work and findings of this study include the following five parts:

First, learner analysis. A survey of  271 senior high school students in Lanzhou E Middle School showed that they had a strong interest in learning AI, and 24% of the respondents chose the course "Preliminary AI" from six optional compulsory courses of information technology; The students' AI learning foundation is very weak as a whole: more than 60% of the students have not learned any programming language before the first year of senior high school, and only 4 students can correctly say at least one AI algorithm or AI term.

Second, course design. Based on the Preliminary AI textbook published by the People's Education Press, a 35-hour project-based-learning AI course is designed, which is divided into 32 hours of teaching and 3 hours of examination, including 10 group projects. The implementation steps of PBL in this study are divided into the following six steps: identifying a topic, developing a plan, implementing the plan, communicating the results, teacher and student evaluation, refining and improving.

Third, learning evaluation design. Learning evaluation mainly includes group project report and individual course examination. This study conducted 35-class hours of AI education and teaching activities for 14 students from Lanzhou E Middle School and Lanzhou Y Middle School. In the course, five groups completed three projects: "Face Recognition", "Investigation of Artificial Intelligence Applications in Lanzhou" and "Optional Project". The course examination mainly includes four parts: "Principle and Application of Basic Algorithms of Artificial Intelligence Technology", "Principle, Construction and Application of Artificial Neural Network Model", "Application and Future Development of Artificial Intelligence Products", and "Basis of Python Programming". There are 25 questions in total.

Based on the classroom performance, group project reports and examination results, seven students basically met the academic requirements for the four optional compulsory modules in the information technology curriculum standard: in terms of computational thinking, they can use Keras open-source framework to build simple intelligent systems; In terms of digital learning and innovation, many cases of artificial intelligence application have been realized in the course and exam, solving practical problems in life; In terms of information awareness and information social responsibility, students can realize that AI is around, find AI products around, and objectively understand the impact of AI technology on society, law, etc.

Fourth, improve the teaching strategy design of students' self-regulated learning. In this study, students' self-regulated learning in the AI course is generally good. After conducting semi-structured interviews with 9 students, it was found that: in the dimension of learning motivation, students can generally accept the difficulty of the course, students have a strong sense of self-efficacy when completing classroom tasks and group activities, and they believe that the cases arranged by teachers are of practical significance and value, and this course is of great help to their future development. Students have a strong interest in learning, and students have gained a lot after the course. In terms of learning strategies, each student adopted three learning strategies: "searching for materials online", "watching course playback", and "writing and saving classroom procedures". Students' learning goals are not clear enough or have not set learning goals, and few students have the awareness and behavior of making project plans when completing projects. As for the dimension of learning situation, students generally believed that teachers, team members and other students had a great help to their own learning and project completion during the course, and accepted the learning method of team cooperation to complete the project.

In view of students' self-regulated learning, teachers can improve students' learning experience and cultivate students' awareness of self-regulated learning from the following three aspects when launching AI courses based PBL: 1. Recording courses, students generally have the need to watch the course playback; 2. Design rich and interesting cases, including classroom tasks and group projects, and teachers should provide guidance materials for each project to help students complete the project; 3. Strengthen the guidance for students to formulate individual learning plans and group project plans.

Fifth, learning environment design. This study found that Tencent Coding platform has many advantages, including convenience for teachers to demonstrate and students to write programs, fast speed when installing Python external modules, its own artificial intelligence modules such as Tensorflow, fast running speed, and support for online collaboration of groups. Tencent Coding platform also has two disadvantages: 1. Reinstall the previously installed modules after each exit; 2. Some versions of built-in Python modules do not support cases in textbooks, such as the face recognition case, but Lightly can compile and run the face recognition case well. Therefore, for schools with relatively backward computer room machines or teachers who have difficulties in deploying AI development and compilation environments, IDLE, Tencent Coding, and Lightly can be used together in teaching.

The main innovations of this study include: designed a relatively systematic AI course; explored the situation of students' self-regulated learning in project-based-learning AI curriculum, and provided teachers with teaching strategies that can improve students' self-regulated learning ability in AI courses based PBL; summarized the advantages and disadvantages of the Tencent Coding platform in the teaching of artificial intelligence courses through teaching practice. The project-based-learning AI curriculum designed in this study has the characteristics of foundation, life and practicality, which can provide some reference for local front-line teachers who are willing to carry out AI courses.