Module 1 - Bài 2. Giao tiếp hiệu quả với AI

Main ideas & lessons

• Effective AI use is a skill, not “magic.” This lesson focuses on practical methods for communicating with AI chat tools.
• Context beats keywording. Unlike search engines where short keywords work well, AI prompts require descriptive sentences and strong context to get valuable results.
• Mindset shift from search to collaboration. Instead of acting like a Google user (short queries), you act like a student/teacher-in-charge who provides requirements and constraints so the AI can help appropriately.
• Use a structured prompting framework to improve outcomes. The lesson teaches a reusable prompt structure to reliably produce correct, usable outputs.
• Don’t rely on a single prompt/round. The best results come from iterative refinement (multiple rounds of editing and questioning).
• Use AI as a tutor/assistant that supports learning. The goal is that the less the AI “does everything,” the more you learn and gain real understanding.
• Avoid harmful prompting habits, especially around data safety and academic integrity.

The “six accomplishments” the lesson claims you will achieve

1. Master contextual objectives and output tasks when chatting with AI.
2. Understand why context matters more than keywords.
3. Use sample examples to improve prompt conversations (the subtitle references “sample ROM upgrade examples” and “room two conversations,” but the intent is to practice prompting with examples).
4. Identify five common mistakes so you don’t repeat them later.
5. Learn how to change prompting style away from short keyword habits.
6. Learn strategies (iteration and review moves) to get better results over multiple rounds.

Instructional methodology: prompting framework (MBND)

Four-part framework for building prompts (memorize/check for missing pieces)

• M = Objective (目的)

  • State what you need the AI to do for learning (e.g., explain the essence, summarize for an exam, rephrase to guide understanding).

• B = Context (背景/背景信息)

  • Provide the AI with details that help it tailor results:
    • Who you are / skill level / audience level
    • Domain/business context
    • Input data (your problem statement, code snippets, or your written work for analysis)
  • Safety constraint: do not share sensitive info (e.g., login names, passwords, personal information, unpublished documents/research).

• N = Task (任务/指令)

  • Use an appropriate intent verb (examples given: explain, compare, refute, ask questions).
  • Prefer not doing it for you:
    • Ask leading questions so you can figure it out yourself.
    • Add constraints so the AI acts like a tutor/companion, not an automatic doer.
  • If unsure, say you are not sure rather than forcing guessing.

• D = Desired Final Output (产出/格式/成品要求)

  • Specify the format and structure:
    • length, structure, bullet points, tone, and what the final “product” should look like
    • optionally request sources, but sources are not proof—they’re references you must verify.

Prompting advice about the framework

• You don’t always need all four parts:
  • For trivial questions, task only (N) may be enough.
  • For moderate needs, use 3 components.
• In practice, you don’t need to type the letters; the point is to ensure those elements are present.

Techniques for better interaction with AI

1) “One-prompting” / One-Prompting (show an example/template)

• Provide an example of the desired output (e.g., a sample flashcard), because:
  • Examples are more powerful than descriptions.
• Warning:
  • The AI may mimic errors you didn’t intend (like formatting specifics), so your example must be accurate.

2) Multi-round workflow (don’t treat AI like a vending machine)

• Subtitles criticize the pattern:
  • “Put in a question → get the answer → leave.”
• Instead:
  • Round 1 is usually tentative.
  • Your job is to edit, refine, and keep prompting until the result is high quality.
• Competitive advantage claim:
  • Since many students already use AI (the subtitle states over 95%), the advantage is in deeper rounds (2nd/3rd/4th).

3) Six conversational moves (strategy)

The subtitle describes a set of “six conversational moves,” with explicit mention of:

• Ask AI to narrow down an idea
• Ask it to change perspective
• Ask it to challenge its own answer
• Move #6 (explicit): Ask for a review:
  • “Please type a review of my method below and point out any mistakes.”
  • Then ask three questions to check understanding.

Case studies / examples used to illustrate the method

Case study A: Writing a psychology essay

• Wrong/naive approach:
  • Students ask for “write an essay” → output gives little learning value and can trigger integrity problems.
• Improved prompting approach:
  • Request not the full essay, but:
    • three main points
    • the types of evidence needed
  • Then ask for a map/outline to navigate and learn independently.
• Further refinement:
  • Ask the AI to point out three weaknesses in your draft/prompt plan.
  • Emphasis: better prompts lead to the AI doing less, while you learn more.

Case study B: Programming/binary code (engineering/information technology)

• Wrong/naive approach:
  • Ask the AI for code directly → it may run, but you don’t understand when the teacher asks follow-up questions.
• Improved workflow:
  • Paste your code.
  • If it fails, ask:
    • don’t rewrite automatically
    • ask three questions to help you find the error.
• Core rule reiterated:
  • When you prompt well, the AI’s responses can drive you to think more and therefore learn more.

Five harmful habits to avoid (explicit list + additional emphasis)

1. Prompt that’s too long
   • Overstuffing methods/techniques can cause confusion and break requirements.
2. Copying a prompt from the internet without adapting context
   • Results won’t match your situation/audience.
3. Asking and then copying/pasting
   • Leads to lack of understanding, poor learning, and potential school disciplinary issues.
4. Feeding sensitive personal/account/financial data directly to AI
   • Examples of dangerous data mentioned:
     • login credentials, passwords
     • personal information
     • credit card details
     • bank account information
5. Accuracy/integrity negligence (implied through “common problem” + verification reminders)
   • Don’t treat outputs/sources as automatically correct or “proof.”
   • The subtitle stresses: you must ensure the information is accurate and verify sources.

Overall summary (what to take away)

• No magic spells: use a framework built on purpose/objective + context + task/output requirements—especially strong context.
• Don’t stop after the first answer: iterate through multiple rounds; use AI as an assistant while you remain the learner.
• Better prompts lead to more learning: the best learning happens when the AI does less “for you” and more “with/around you.”
• Safety & integrity matter: avoid harmful habits, especially sensitive data sharing and academic misconduct.

Speakers / sources featured

• No specific named speakers or external sources are identified in the subtitles.
• The “AI chat tools” referenced include ChatGPT-like tools and Google Gemini (mentioned as examples).