Student learning · Grades 9–12 · 75 minutes
Algorithm Autopsy
Students dissect the design features of a fictional social app, trace how each one shapes attention and belief, and redesign the feed for people instead of engagement.
Recommendation systems are a form of AI most students use every day without choosing to. They predict what will keep a person watching and rank content accordingly, and the app around them is designed with features like autoplay, streaks, and infinite scroll that keep people on the app longer. In this autopsy, teams examine the features of a fictional app called Loop, sort each one by what it is optimized to do, and discuss in a Socratic seminar whether a feed can shape what people believe. They finish by designing a healthier feed with specific rules, trade-offs, and a way to measure whether it works.
In this packet
- Handout A: Loop Feature Autopsy (card sort)
- Handout B: The Loop Memo (Seminar Text) (reading)
- Handout C: Healthier Feed Design Spec (worksheet)
- Facilitator key (last page)
TCEA ELE indicators
- AI1.1 I know how AI systems work and their capabilities in educational contexts.
- S5.2 I can use creativity and logic to find innovative solutions.
- S5.1 I know how to approach complex problems systematically.
- AI1.3 I am aware of the current limitations and potential biases in AI systems.
- S2.2 I can identify trustworthy sources of information on the internet.
Full facilitator guide: mglearn.github.io/eles/activities/algorithm-autopsy.html
Handout A for Algorithm Autopsy
Loop Feature Autopsy
Loop is a fictional video app. For each feature, decide what it mostly does, and say aloud what it is optimized for and who benefits.
Keeps you on the app
Shapes what you see and believe
Gives you control
Handout A for Algorithm Autopsy
Loop Feature Autopsy: cards to sort
Cut apart the strips. Sort each one onto the mat and be ready to explain why.
Sheet 1 of 2
1. Infinite scroll: the feed never ends; there's no bottom of the page.
2. Autoplay: the next video starts automatically after three seconds.
3. "For You" ranking: the main feed is ranked by predicted watch time, not by who you follow or when it was posted.
4. Streaks: a counter shows how many days in a row you've posted or messaged a friend.
5. Trending topics: a sidebar lists the topics getting the most reactions right now.
6. "Not interested" button: tap it and the feed shows less of that kind of content.
7. Push notifications: "3 friends posted while you were away!"
8. Following-only view: a setting to see only accounts you follow, in time order.
9. Engagement boost: posts with many comments and shares get shown to more people, whether the reactions are positive or angry.
10. Like counts: every post shows its number of likes.
11. Daily time limit: users can set a reminder after a chosen number of minutes.
12. "Why am I seeing this?" panel: explains that a post appeared because you watched similar videos.
Handout A for Algorithm Autopsy
Loop Feature Autopsy: cards to sort
Cut apart the strips. Sort each one onto the mat and be ready to explain why.
Sheet 2 of 2
13. Similar-account suggestions: after you follow an account, Loop suggests five more like it.
14. Variable rewards: new likes and comments are delivered in bursts when you open the app, not as they arrive.
Handout B for Algorithm Autopsy
The Loop Memo (Seminar Text)
This memo is fictional. Annotate: underline claims, circle words that sound neutral but aren't, and write one question in the margin for the seminar.
1LOOP INTERNAL MEMO · Product Team · Subject: Q3 Feed Update
2Team, great news. Our new ranking model, which predicts how long each user will watch each video, has increased average daily time on Loop. Users are clearly finding more of what they love.
3The model learns from every signal: watch time, rewatches, pauses, comments, shares, and how fast someone scrolls past. We've found that videos generating strong emotional reactions, whether excitement, surprise, or anger, keep people watching longest, so the model naturally shows more of them.
4Some team members raised a concern that users who watch one conspiracy-style video are now seeing many more. We want to be clear: Loop doesn't choose what people believe. We simply show people what they want. If someone watches it, they must want it.
5We also tested a "Following-only" default. Time on the app went down, so we've kept it as an optional setting, available three menus deep.
6Next quarter, we'll add streaks to direct messages and move the daily time limit reminder to after 90 minutes instead of 30, since users told us it was annoying.
7Our mission remains the same: connect people with the content that matters to them.
Handout C for Algorithm Autopsy
Healthier Feed Design Spec
Design a feed for real people. Be specific enough that an engineer could build it.
Who is this feed for, and what is it optimized for instead of pure engagement? (for example: learning something new, connection with friends, accurate news)
Three ranking rules: how the feed decides what comes first.
Two user controls, and whether each is on or off by default (and why):
One Loop feature we would remove, and the evidence from the autopsy or seminar that convinced us:
The trade-off: what would the company or users lose with our design?
How we'd measure whether it works (something you could actually count or ask users):
Sketch the feed screen, showing where the controls live:
Facilitator only for Algorithm Autopsy
Answer key and notes
Handout A: Loop Feature Autopsy
| Item | Sort | Why |
|---|---|---|
| 1. Infinite scroll: the feed never ends; there's no bottom of the page. | Keeps you on the app | Removes natural stopping points, so leaving requires a deliberate decision. |
| 2. Autoplay: the next video starts automatically after three seconds. | Keeps you on the app | Makes continuing the default; also decides what you see next without your choice. |
| 3. "For You" ranking: the main feed is ranked by predicted watch time, not by who you follow or when it was posted. | Shapes what you see and believe | The system, not you, decides what appears, optimized for keeping your attention. |
| 4. Streaks: a counter shows how many days in a row you've posted or messaged a friend. | Keeps you on the app | Turns daily use into an obligation and makes missing a day feel like a loss. |
| 5. Trending topics: a sidebar lists the topics getting the most reactions right now. | Shapes what you see and believe | Signals what's important based on reaction volume, which can reward outrage or surprise. |
| 6. "Not interested" button: tap it and the feed shows less of that kind of content. | Gives you control | Lets users push back on the system's predictions, though users must know it exists. |
| 7. Push notifications: "3 friends posted while you were away!" | Keeps you on the app | Pulls you back in at moments the app chooses. |
| 8. Following-only view: a setting to see only accounts you follow, in time order. | Gives you control | Replaces predicted ranking with your own choices; often not the default. |
| 9. Engagement boost: posts with many comments and shares get shown to more people, whether the reactions are positive or angry. | Shapes what you see and believe | Reactions, including anger, become a signal of value, so provocative content spreads further. |
| 10. Like counts: every post shows its number of likes. | Shapes what you see and believe | Popularity can be mistaken for accuracy or quality. |
| 11. Daily time limit: users can set a reminder after a chosen number of minutes. | Gives you control | Supports balance, but only if users turn it on and don't dismiss it. |
| 12. "Why am I seeing this?" panel: explains that a post appeared because you watched similar videos. | Gives you control | Adds transparency so users can understand, and question, the prediction. |
| 13. Similar-account suggestions: after you follow an account, Loop suggests five more like it. | Shapes what you see and believe | Can narrow your feed toward one viewpoint over time. |
| 14. Variable rewards: new likes and comments are delivered in bursts when you open the app, not as they arrive. | Keeps you on the app | Unpredictable rewards encourage checking again and again. |