How Do Apps Decide What We See Next?

Full script

Read the whole talk.

All slides in one place. Read it through once, then practice slide by slide.

1. 01How Do Apps Decide What We See Next?

   Good evening, everyone. Last month, I watched one video about bread baking. The next day, my whole feed was bread. How did the app know? Today, I'll explain, in simple terms, how recommendation systems decide what we see.

   大家晚安。上個月，我看了一個關於烤麵包的影片。第二天，我的整個首頁全都是麵包。這個應用是怎麼知道的呢？今天，我會用簡單的話來解釋推薦系統是怎樣決定我們看到什麼內容的。

2. 02What Is a Recommendation System?

   A recommendation system is a computer program that tries to predict what you'll want to watch, read, or buy next. Video apps, music apps, and online shops all use them.

   推薦系統是一種計算機程式，它會試著預測你接下來想看什麼、讀什麼或買什麼。影片應用、音樂應用和網路上商店都在使用它。

3. 03Signal 1: What You Do

   The first type of information is your own behavior: what you click, how long you watch, what you like or share, and even what you skip. Watching a video to the end is often a strong signal.

   第一類資訊是你自己的行為：你點了什麼、看了多久、按讚或分享了什麼，甚至你跳過了什麼。把一個影片看到最後，往往是一個很強的訊號。

4. 04Signal 2: People Like You

   The second idea is comparing you with similar people. If many people who watched the same videos as you also enjoyed something else, the system may guess you'll enjoy it too.

   第二個思路是把你和相似的人進行比較。如果很多和你看過同樣影片的人也喜歡另一個內容，系統就可能會猜你也會喜歡它。

5. 05The Good Side

   This can be very helpful. I've discovered great music and useful cooking videos that I would never have found by myself.

   這可能非常有用。我發現了很多很棒的音樂和實用的烹飪影片，要是靠我自己，可能永遠也找不到它們。

6. 06The Risks

   But there are risks. If the system only shows us more of what we already like, our view of the world can become narrower. And because many apps want us to keep watching, it's easy to lose track of time.

   但這也有風險。如果系統只給我們看更多我們已經喜歡的東西，我們看世界的視野就可能會變窄。而且，因為很多應用都希望我們一直看下去，所以我們很容易忘記時間。

7. 07Taking Back Control

   The good news is that we can influence what we see. Search for new topics on purpose, use the "not interested" button, and follow people with different opinions. Our actions teach the system.

   好訊息是，我們可以影響自己看到的內容。有意識地去搜尋新的話題，使用「不感興趣」按鈕，並關注那些有不同觀點的人。我們的行為會教會這個系統。

8. 08Conclusion

   To conclude, recommendation systems learn from what we do and from people like us. They can open doors or close them. Remember: while the app is learning about you, you're also training it. Thank you.

   最後，推薦系統會從我們的行為和與我們相似的人身上學習。它們可以為我們開啟一扇門，也可以把門關上。請記住：在應用瞭解你的同時，你也在訓練它。謝謝大家。