Video summary

CAEP LAB 03 Video Recording

Main summary

Key takeaways

Educational

Main ideas, concepts, and lessons

1) Course framing (Lab 3, end of program)

  • The session is CAEP Lab Three (final day).
  • The focus shifts from building artifacts/apps to:
    • AI ethics
    • Turning scattered LLM capabilities into repeatable educator workflows
    • Comparing three major LLMs: ChatGPT, Claude, Gemini
  • Participants are reminded of deadlines:
    • Upload and complete tasks by next Sunday
    • Receive certificates after completion (Inshallah)

2) Core principle: “Human judgment stays central”

  • AI can generate drafts, analysis, and outputs, but:
    • AI can hallucinate / make errors
    • Human facilitation and checkpoints are required
  • Mindset: use AI to speed work, while humans remain responsible for verification and final decisions.
  • “Human checkpoints” are treated as a required stage in ethical AI educator practice.

Methodology / workflow taught (LLM prompt → prep → workflow → agents)

A) Prompting basics (“copy-paste prompt”)

  • Participants practice using the same prompt across three LLMs to compare outputs.
  • Goal: understand how different models handle the same instructions.

B) Five-step workflow for creating lesson/activity outputs (Prep → workflow)

The instructor models an end-to-end workflow using prompts that gradually become more “locked down”:

  1. Prep / Analysis step (no activity creation yet)

    • Task: analyze a topic and identify:
      • key concepts
      • tensions and debates
      • misconceptions
      • ideas requiring deeper student thinking
    • Constraint: do not create an activity yet.

  2. Learning design step (produce learning intention, not an activity yet)

    • Task: choose one strong learning opportunity and define:
      • learning intention
      • evidence of student thinking/learning
    • Constraint: still do not create an activity yet.

  3. Activity design step (create an inquiry-based activity)

    • Task: design an inquiry-based activity requiring students to:
      • use evidence
      • compare perspectives
      • justify interpretations
    • Constraint: do not provide conclusions in advance.

  4. Audit step (educator checks alignment and weaknesses)

    • Task: audit the created activity for:
      • alignment with the learning intention and cognitive challenge level
      • what thinking must be done by students vs. by AI
      • unsupported assumptions
      • gaps requiring educator verification
    • Constraint: identify genuine weaknesses and recommend only necessary revisions.

  5. Final document assembly

    • Produce a consolidated “final” output after revisions.

Key takeaway: this structure turns LLM output into a repeatable professional educator workflow, not a one-shot response.

C) AI agents (advanced orchestration)

  • Next concept: AI Agents (goal-defined systems where the AI decides steps/tools/actions within boundaries).
  • Difference from prompting/workflow:
    • In prompting, you request output.
    • In workflow, you define stages; AI performs each stage with your structure and human checkpoints.
    • In agents, you define the goal, and the AI selects/executes steps toward that goal while still respecting constraints that keep human checkpoints intact.

How participants were taught to choose an approach

  • Use simple prompting for:
    • straightforward Q&A or quick generation (low risk, small edits)
  • Use workflow for:
    • lesson planning or structured designs needing fewer mistakes
  • Use agents for:
    • more complex programs with many steps (e.g., an entire course or certificate program), where breakdown/coordination is needed

“Deep Research” practice across LLMs

What “Deep Research” was positioned to do

  • The instructor demonstrates a research-augmented workflow feature (in supported tools/plugins), described as:
    • gather multiple credible perspectives
    • compare and synthesize
    • verify/transform outputs
    • produce evidence-based results with citations/links

Demonstration activity (academic authorship + generative AI)

  • Topic used repeatedly: generative AI, academic authorship, and research implications.
  • Participants run multiple prompts in ChatGPT, Claude, and Gemini with “deep research” enabled where possible.
  • Observed practical issues:
    • availability varies by free vs paid plans
    • traffic/capacity limits occurred in Gemini
    • different timing and quality patterns across Claude/Gemini/ChatGPT

Reported comparative takeaway (from instructor)

  • Deep research generally produced more credible, citation-based outputs.
  • The instructor also emphasized personal judgments about model reliability (e.g., concern about Gemini being weaker or sometimes repeating/mistaking), reinforcing that results still require human review.

AI Ethics segment (explicit scenarios + rules)

What “AI ethics” was defined as in practice

  • Not banning AI—rather:
    • use it wisely
    • apply privacy protection
    • verify outputs (AI can be wrong)
    • keep professional responsibility with educators
    • ensure transparency with students

“Five question” ethics scenarios (student privacy → grading → cheating → teaching materials → adaptive bias)

  1. Student privacy

    • Scenario: teacher uploads student counseling notes, medical info, behavior records, and full name to AI for advice.
    • Response: No (privacy/data protection breach).

  2. AI grading without human review

    • Scenario: AI grades 50 student essays against a rubric and records grades officially without teacher review.
    • Response: No (teacher accountability and verification required).

  3. Student using AI for brainstorming but final assignment is theirs

    • Scenario: student uses AI to brainstorm and receive feedback, but final submission is student-written.
    • Response: Generally No cheating automatically if used ethically (focus is on learning + proper instruction).

  4. AI-generated historical facts/quotes/citations used without checking

    • Scenario: teacher uses AI-generated facts/quotes/citations without verification.
    • Response: Not ethical (citations/facts must be verified).

  5. AI personalization that lowers expectations

    • Scenario: AI analyzes prior performance and assigns weaker students easier tasks.
    • Response: Acceptable only with human monitoring to avoid reinforcing low expectations and bias.

Instructor’s concluding ethics principles (practical rules)

  • Do not outsource high-stakes professional responsibility (grades, final assessment decisions).
  • Protect student privacy at every cost.
  • Verify AI outputs (never blindly trust).
  • Transparency: don’t present AI work as entirely flawless or fully “perfect”; teach students as life-long learners.
  • AI may support productivity, but human judgment is irreplaceable.

Content creation / infographic and media demo

  • The session transitions to creating visual content from a YouTube video (handnotes/infographics) using an app demonstrated live.
  • A suggested tool/app was demonstrated:
    • participants generated a graphical/handnote style infographic from a provided YouTube webinar link.
  • Participants also shared examples of educational content they created (e.g., posters/infographic-style visuals, educational videos, narrated content).
  • The instructor emphasized that generated visuals can support classroom teaching, but should still be used responsibly (consistent with the ethics framing).

Final wrap-up / logistics

Participants must:

  • complete tasks from earlier labs
  • ensure their artifact/app/website is live (or share links)
  • use Google Classroom for submissions and receive tutorials

Instructor availability:

  • questions via the WhatsApp group
  • additional tutorial/video resources planned for upload

Speakers / sources featured

Speakers / instructors / participants (identified in subtitles)

  • Main instructor/facilitator (repeated throughout; leads Lab 3, demos workflow/agents/deep research, delivers ethics rules)
  • Sana Bil / Sanabil (Sana) (participant; responds to ethics scenarios and prompts; shares content/insights)
  • Rafat (participant; responds to ethics discussion; shares examples and feedback)

Systems/tools mentioned (used as “sources” for output)

  • ChatGPT
  • Claude (referred to as “Claude / clot” in subtitles)
  • Gemini
  • “Deep Research” feature/plugin (where available in the tools)
  • Google Classroom (submission platform)
  • WhatsApp group (support and coordination)

Original video