# Teaching design comparison

Read date: 2026-09-08. This is a comparison of seven publicly readable teaching resources, plus the supplied Shenzhen Polytechnic University training material and the project's earlier public-course outline. It records observed course features, not claims about comparative learning effectiveness. Pages were read directly; inaccessible pages are listed separately and do not count toward the seven.

## Sources actually read

| Resource and intended audience | Structure and practice visible on the source | Adopt for this course | Do not carry over |
|---|---|---|---|
| [Google AI Essentials](https://grow.google/ai-essentials/), adults with no prior experience | Five modules: introduction, productivity, prompting, responsible use, staying current. Public overview describes workplace activities for brainstorming an outline, researching/organizing information against criteria, and drafting email responses. The page currently advertises under five hours; this is a changing catalog fact. | One useful artifact per session; brief explanation immediately followed by action; work-related language. | Productive-looking output is not sufficient evidence. Our tasks require source excerpts and retained observations. No subscription or provider-specific account is required. |
| [Elements of AI](https://www.elementsofai.com/), non-experts; University of Helsinki and MinnaLearn | Introduction combines theory and practical exercises, is self-paced, and requires no complicated mathematics or programming. Building AI is a separate next step, with Python recommended. Public landing page read; logged-in exercises not claimed to have been completed. | Separate accessible core from optional technical extension; explain concepts before terms; keep algorithms outside the required practical path. | No compressed survey of every AI algorithm. This course trains a specific evidence workflow. |
| [Microsoft Learn: Introduction to generative AI and agents](https://learn.microsoft.com/en-us/training/modules/fundamentals-generative-ai/), beginner students and technical roles | Seven units: introduction, LLMs, prompts, agents, a generative-AI exercise, assessment, summary. Objectives use observable verbs. The page lists basic AI/ML familiarity as prerequisites. | Give explicit objectives, demonstrate one example, use a short independent check at the end. | Do not assume that a beginner label means no prerequisites. Introduce our small vocabulary and avoid Azure/Python requirements. |
| [Data Carpentry: Data Organization in Spreadsheets](https://datacarpentry.github.io/spreadsheet-ecology-lesson/), learners without prior tool knowledge | Six episodes cover entry, formatting, dates, quality control and export. A deliberately messy dataset is supplied. The formatting-problems episode warns against multiple facts in a cell, color-only labels, mixed units, and replacing missing values with zeros; it is intended as discussion support while students diagnose data. | Supply messy questions and a broken page; students repair a copy. Use one observation per row, explicit status columns and blanks for missing outcomes. | The ecological subject matter and programming analysis are unnecessary; use a small, familiar service case. |
| [SIFT: The Four Moves](https://hapgood.us/2019/06/19/sift-the-four-moves/), students learning online verification | Stop; investigate the source; find better coverage; trace claims, quotations and media to their original context. The guide explains why leaving an unfamiliar page can be more useful than inspecting its visual polish. | Make source tracing a practiced action: open the publisher and original text before judging a claim. Include a polished promotional page that is not independent evidence. | A checklist score is not proof of truth. No claim that all third-party pages are reliable. |
| [Cornell University Library: Evaluating Web Pages](https://guides.library.cornell.edu/evaluating_Web_pages), undergraduate information-literacy learners | Accuracy, authority, objectivity and currency pages give questions about authorship, publisher, purpose, affiliations and correctly cited references. The accuracy and authority pages were read in full. | Turn questions into concrete worksheet fields: publisher, date, purpose, independence, supporting passage and scope. | Domain names and credentials alone do not validate a specific sentence. Evaluate the relationship between the claim and the cited text. |
| [University of Edinburgh: Research Data MANTRA](https://mantra.ed.ac.uk/), people managing research data | Separate learning units cover context, planning, organizing, archiving, safety, sensitive data and FAIR access. Organizing explicitly includes names, versions and collaboration; tutorials provide open datasets and exercises in PDF/HTML for analysis packages. | Teach file naming, immutable originals, versions and a handover index before large-scale collection. Make the final project reopenable by another learner. | Data governance is taught through small practical decisions; this introductory course does not reproduce a research-compliance program. |

## Additional materials

- Supplied Shenzhen Polytechnic University ZIP: five 90-minute integrated exercises were read from DOCX XML; topics are acquisition/annotation, SVM classification, MNIST classification, image enhancement and dataset preparation. The useful pattern is stated inputs, numbered operations, screenshots and itemized scoring. No code or media from the attachment is redistributed. Its 10 DOC guides and three Python/NumPy/Pandas slide files were inventoried; their full contents were not claimed to have been reviewed.
- Earlier project outline `GEO课程目录重构版.md`: the 10-module, nine-assignment sequence links project setup, website facts, query maps, evidence pages, third-party sources and repeated testing. It is marked historical, so it informs scope rather than superseding current course materials.
- `course-labs/README.md`: eight reproducible research labs inform retained originals, stable identifiers, claim-evidence links, missingness and reproducible handover. We replace required Python execution with local pages and a spreadsheet.

## Resulting organization

Retain **ten 90-minute workshops**, organized into **five two-session cycles**: (1) define and collect questions; (2) observe answers and investigate sources; (3) verify and revise content; (4) inspect pages and measure repeated observations; (5) review and hand over. Ten is an instructional design choice based on the number of distinct artifacts and the need for a guided attempt plus independent practice. It is not copied from the ZIP and is not an empirically proven optimum.

Each workshop follows a 10/15/40/15/10-minute rhythm: retrieve prior knowledge, watch one worked example, complete a guided task, check a partner's evidence, write an exit note. The same fictional campus bicycle service reduces topic-switching burden. The final task asks learners to transfer the workflow to a new small local service brief; transfer is assessed through evidence handling rather than brand visibility gains.

Every unit includes: a concrete brief; files to open; an observable result; a worked passage or screenshot; exact steps; a submission list; four scoring dimensions; common errors and recovery; a short independent check. Early tasks use filled examples; later tasks withhold the solution until review. An optional live extension has a separate record type and does not require paid automated requests.

## Access limitations

OpenLearn and FutureLearn returned anti-bot challenge pages; IBM legacy links redirected to a new catalog/educator overview without readable course details; the Stanford COR curriculum index exposed navigation but not lesson descriptions in the fetched HTML; Check, Please redirected without readable lessons. They were explored but excluded from the seven-source comparison. No unavailable lesson content was inferred.

## 中文设计摘要

实际读取七项公开课程或教学指南，并另与深职大附件和项目旧版社会课程比较。最终保留十次、每次九十分钟的工作坊，组织为五组连续实践：定义与提问、观察与查源、核验与改写、页面与测量、复核与交付。每次围绕一个具体成果展开，依次安排示范、动手、互查和离堂小结。这个数量来自任务依赖和课堂时间分配，不是照搬附件，也不声称已有教学实验证明最优。

核心路径只需浏览器和表格；同一虚构校园服务案例贯穿全程。示例答案逐步减少，末次加入新服务场景的迁移练习。英文讲义作为主稿，中文按相同任务、数据、图号和验收口径翻译，术语统一为 observation/观察记录、claim/主张、evidence/证据、source/信源、mention rate/提及率、independent corroboration/独立印证。