Junior AI Coder

Lesson 3

Build Byte's Recommendation Skill
建立 Byte 的智能推薦技能

Lesson 2 taught Byte to detect “Get a Recommendation.” Now students build the skill that compares safe preferences, scores fixed choices, and explains one useful suggestion.

第二課中，Byte 學會辨認「推薦」意圖。第三課會建立真正的推薦技能：比較安全的喜好、計算分數，並解釋建議原因。

Open Recommendation Sandbox
開啟推薦沙盒 View Lesson Tasks
查看課堂任務 Review Lesson 2
重溫第二課

Core Question

Why did the assistant choose this suggestion, and does it really fit the user?

Course Journey

From Intent to a Useful Skill
由理解意圖到提供實用技能

Lesson 3 gives the personal AI assistant its first decision-making skill.

  1. 1 Conversation Safe replies
  2. 2 Intent Choose the job
  3. 3 Skills Recommendation focus
  4. 4 Memory Coming next
  5. 5 Knowledge Coming later
  6. 6 Personal Assistant Final project

Lesson Story

Byte Knows the Job. What Should It Suggest?

When a user asks for a recommendation, a useful assistant should compare a small number of relevant choices. It should not guess from private information or pretend that one answer is perfect for everyone.

“I want a quiet activity I can do alone for 20 minutes.”

Byte can compare that safe profile with fixed activities and explain why one choice scored highest.

A recommendation system follows a simple pattern: collect suitable inputs, compare them with features, calculate scores, choose a result, and explain the match.

推薦系統會收集合適的輸入、比較選項特徵、計算分數、選出結果，再解釋為甚麼這個建議可能適合。

What Students Learn
學生會學到甚麼

  • A preference is a choice, not a private fact.喜好是一個選擇，不是私人資料。
  • Features describe useful differences between choices.特徵用來描述不同選項的差異。
  • Scores turn matching rules into a result.分數把配對規則變成推薦結果。
  • A reason helps the user judge the suggestion.解釋原因可以讓使用者判斷建議是否合理。
  • Ties and weak matches need careful handling.分數相同或配對較弱時，需要小心處理。
  • Recommendations can be useful without being perfect.推薦可以有用，但不代表一定正確。

Key Words
關鍵詞

Recommendation
推薦

A useful choice suggested by a system.

Preference
喜好

What the user says they would like.

Feature
特徵

A quality used to compare choices.

Score
分數

Points added when a choice matches.

Ranking
排名

Ordering choices from highest to lowest score.

Bias
偏差

A rule or data choice that may unfairly favour one result.

How It Works

One Small Recommendation Pipeline

Students can see every step, change the code, and compare the local result with a real AI model.

Step 1 Choose safe preferences

Energy, solo or together, and available time.

Step 2 Add scores

Each matching rule gives points to one or more fixed activities.

Step 3 Explain and check

Show the top choice, the reason, and a reminder that the user should judge the fit.

Interactive Lab

Compare Code Rules with Real AI
比較程式規則和真實 AI

The sandbox uses the same profile for both results. The real AI result is accepted only when it uses one of five fixed activity names.

Launch Lesson 3 Sandbox

Quick Multiple-Choice Challenge

Task Checklist

Testing Profiles

Test AQuiet + Solo + 20 minutes

Does the reason connect to quiet and solo preferences?

Test BActive + Together + 10 minutes

Does a short active choice rise to the top?

Test CEither + Either + 20 minutes

Several choices tie. How should a recommender explain that uncertainty?

Safety and Fairness Check

The demo uses only classroom preferences. It does not collect names, locations, health information or family details. A high score is evidence from the current rules, not proof that the suggestion is perfect.