Learner: classify a point学习者：对一个点进行分类
Change k, predict the label, and explain which neighbours changed the result.改变 k，预测类别，并解释哪些近邻改变了结果。
Open the KNN experiment →打开 KNN 实验 →Start in five minutes五分钟开始
Choose one question, make a prediction, then inspect what the algorithm does.选择一个问题，先做出预测，再观察算法如何运行。
AI Playgrounds is a bilingual, offline-ready suite for learning foundational AI through controlled experiments rather than passive animation. No account, installation, or student data submission is required.AI Playgrounds 是一套支持中英双语和离线使用的基础人工智能交互工具。学习者通过可控实验而非被动观看动画来理解算法；无需账户、安装软件或提交学生数据。
- 12 inspectable applets12 个可检查的交互工具
- English + Simplified Chinese英语 + 简体中文
- 45 algorithm checks45 项算法检查
- 63 browser cases63 项浏览器测试
- MIT licensedMIT 许可证
Educator: compare search choices教师：比较不同搜索策略
Use a predict–run–explain cycle as a five-minute demonstration or a twenty-minute lab.采用“预测 - 运行 - 解释”循环，可用于五分钟演示或二十分钟课堂实验。
Open the search experiment →打开搜索实验 →Researcher or developer: audit the model研究者或开发者：审查模型
Inspect assumptions, limitations, deterministic checks, architecture, and release evidence before reusing the suite.复用前可检查假设、局限、确定性测试、架构与发布证据。
Open a probability experiment →打开概率实验 →What the release evidence establishes发布证据能够说明什么
Deterministic algorithm behaviour, bounded browser rendering, static release integrity, and reproducible deployment for the tested cases.在已测试情形下，证据支持确定性的算法行为、限定范围内的浏览器呈现、静态发布完整性与可复现部署。
What it does not establish发布证据不能说明什么
Learning gains, classroom adoption, accessibility conformance, or superiority to another teaching method. Those require separate empirical evidence.它不能证明学习成效、课堂采用情况、无障碍标准符合性或相较其他教学方法的优越性；这些结论需要独立的实证研究。
Privacy and uptake measurement隐私与使用情况测量
Aggregate use signals without learner profiles.只统计汇总使用信号，不建立学习者画像。
On the public GitHub Pages site only, the project counts page views, coarse referral sources, and a first substantive interaction per applet session. It does not send experiment values, student work, names, email addresses, exact location, or persistent cross-site identifiers. Offline and local copies send nothing. Global Privacy Control, Do Not Track, and the opt-out below are respected.仅在公开的 GitHub Pages 网站上，本项目统计页面浏览、粗粒度来源以及每次交互工具会话中的首次实质性操作。系统不会发送实验数值、学生作业、姓名、电子邮件地址、精确位置或持久的跨站标识符。离线或本地副本不会发送任何分析数据，并尊重全局隐私控制、Do Not Track 与下方的退出选项。
Explore the twelve applets
Start with a featured experiment or filter by the idea you need to teach.
Use one in class
The applets are designed for a short demonstration, a full inquiry cycle, or independent practice. Student work stays on the learner's device until they submit it through your normal classroom system.
A ten-minute comparison
Open a featured experiment, ask for a prediction, run the comparison, and collect one explanation.
Open an exampleA complete lesson
Use the scenario gallery, visual explanation, and response packet as one Predict, Observe, Explain, Transfer sequence.
Open the Teacher PackA four-lesson sequence
Move from search and logic to probability, model evaluation, and reinforcement learning.
Open the curriculum mapWhat makes the suite different
The contribution is not one new algorithm visualization. It is a coherent classroom packaging pattern across the foundational AI canon.
One concept at a timeEach applet opens on a concrete behavior rather than a menu or setup screen.
English and ChineseThe interface, teaching prompts, and state descriptions support both languages.
Portable by designEach applet is one HTML file that can run offline without an account or backend.
Misconception and fidelity notesTeacher guidance names the error the applet targets and what the model leaves out.
Classroom evidence flowScenario comparisons and local response packets keep student reasoning visible.
Inspectable qualityAlgorithmic regression tests and browser QA are included in the repository.
Use it, adapt it, or help others find it
AI Playgrounds is an open teaching project. Use it in class, reuse it under MIT, or support discovery with a GitHub star or a shared link.
Research, source, and reuse
Use the tagged release as the artifact of record, cite the project, inspect the implementation, or adapt one applet for your own course.
For researchers and reviewers
The research companion record, citation metadata, architecture notes, and exact release information connect the public artifact to its scholarly and reproducibility context.
For developers and teachers
The source is intentionally plain HTML, CSS, and JavaScript. Fork the project, take one applet, translate it, or change the scenarios without adopting a build system.