Student learningfor Grades 9–12
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.
All activities Student learning
The Feed Game
Students play a card-based recommendation engine by hand and watch their feed shrink to a few topics, then design ways to take it back.
Overview
Recommendation systems don't know what is true, kind, or good for you. They predict what will keep you watching, based on patterns in what you did before. In this game, one student plays the Algorithm and follows a printed rule card to choose posts for a Viewer, scoring every reaction. In five rounds, most feeds narrow to one or two topics, and posts designed to spark strong feelings rise to the top. Students then change one rule and discover what they can and can't control.
Objectives
- Students will simulate a recommendation algorithm by following written rules and explain, in their own words, how it predicts what to show next.
- Students will use their round-by-round data to describe how a feed narrows and why strong-feeling posts get boosted.
- Students will propose and test one change (to the rules or to their own behavior) that widens a feed, and explain the trade-off.
Materials
On paper
- Handout A: Feed Cards (1 set of 16 per team, cut apart: 14 post cards and 2 rule cards)
- Handout B: Feed Tracker (1 per team)
- Handout C: Take Back Your Feed (1 per student)
- Tokens or paper clips for scoring (about 30 per team), or tally marks
On screen
- Optional: a teacher-projected slide showing the rule card and a timer
- Optional: a shared spreadsheet where Recorders enter topic counts to make a class chart
Before you start
- Print and cut Handout A. Keep the two rule cards separate so each team's Algorithm gets them first.
- Draw a class tally on the board with columns for the five topics (Gaming, Animals, Sports, Science, Cooking) and rows for Round 1 and Round 5.
- Decide how teams form. Mix students so Viewers aren't all friends who share interests; varied feeds make a better debrief.
Step by step
-
10–5 min
Hook
Why did that show up?
Ask: "Have you ever watched one video and then your whole feed changed?" Take quick stories. Then say: "Today one of you gets to be the algorithm. You don't get to use your opinion. You follow the rules, exactly like a computer does."
Facilitator noteKeep it light. Students don't need to name apps or admit screen time.
-
25–10 min
Model
How to play
Read Rule Card 1 aloud and play one round with a volunteer Viewer in front of the class. The Algorithm deals 4 random post cards. The Viewer reacts honestly to each: Skip (0), Watch (1), Like (2), or Comment/Share (3). Posts marked Spark earn +1 because people linger on things that make them feel strongly. The Recorder writes the topics and points on Handout B. Then show Rule Card 2: next round, the Algorithm deals 3 cards from the top-scoring topic and 1 from the second-highest.
Facilitator noteStress that Viewers should react as themselves. The simulation only works with honest reactions.
-
310–22 min
Practice
Five rounds
Teams play five rounds with roles fixed, so one Viewer builds one feed. After each round the Recorder totals points by topic and writes which topics are still showing up. When a topic runs out, the Algorithm reshuffles its already-seen cards back in ("feeds repeat"). After Round 5, Recorders report on the class tally: how many topics appeared in Round 1, and how many in Round 5?
Facilitator noteCirculate and ask Algorithms: "Did you check whether that post was true?" The answer is no. The rules never ask.
-
422–30 min
Debrief
What the Algorithm knew
Look at the class tally together. Ask: "What did the Algorithm know about the Viewer?" (Only their reactions.) "What did it not know?" (Whether a post was true, kind, or good for them; their mood; what they wanted to learn.) "Why did Spark posts keep winning?" Connect to AI: "Real recommendation systems are far more complex, but the core idea is the same: they learn patterns from past behavior and predict what will keep you engaged. That's prediction, not understanding."
Facilitator noteWatch for the misconception that the algorithm is "trying to trick you." It's optimizing a goal someone chose. The fair question is: whose goal?
-
530–40 min
Apply
Change one rule
Teams choose one change and play two more rounds with it. Option A (you change): the Viewer deliberately watches and likes a topic they'd never pick. Option B (the designers change): rewrite Rule Card 2 (for example, "1 card from each topic" or "Spark posts earn no bonus"). Teams record whether the feed widened and one trade-off ("more variety, but more stuff I skipped").
Facilitator noteOption B is computational thinking in action: students modify an algorithm and test the output. Ask them to predict before they play.
-
640–45 min
Reflect
Take back your feed
Students complete Handout C individually. Close by asking two or three students to share the one habit they'll try this week.
Facilitator noteCollect Handout C. Look for answers that link a specific behavior (liking, rewatching, sharing) to what the feed shows next.
Paper or screen
Unplugged
This is designed as an unplugged activity: cards, tokens, and a tally on the board simulate the whole system. The physical rule cards make the algorithm's logic visible in a way an app never does.
Digital
Recorders enter Round 1 and Round 5 topic counts into a shared spreadsheet so the class sees a live bar chart of narrowing feeds. For a follow-up, students (with family permission, at home) look at why an app says it recommended a post, if it offers that option, and compare the app's explanation to their rule cards. No student accounts are needed in class.
Does it need a screen? The core learning is better unplugged: students can see and hold every rule, which a real app hides. A shared class chart adds a clear picture of the narrowing pattern across many feeds at once.
Evidence of learning
What you should be able to see or collect if it worked.
- Handout B shows fewer topics in Round 5 than in Round 1, and students can explain why using the rule card.
- Students distinguish what the algorithm "knows" (reactions) from what it doesn't (truth, kindness, the viewer's goals).
- Teams predict the effect of their rule change before testing it and compare the prediction to their results.
Adaptations
- Grades 3–5
- Use three topics (Animals, Gaming, Cooking), play three rounds, and skip Spark scoring. Focus on one question: "Why did you keep seeing the same kind of post?"
- Grades 9–12
- Add a third rule card that a "company" writes: sponsored posts are inserted every other round. Debate who the algorithm serves: the viewer, the creators, or the advertisers.
- Students with IEPs or 504s
- Give the Algorithm role a printed step-by-step checklist version of the rule cards, and let the Viewer point to reaction cards (Skip, Watch, Like, Share) instead of saying them.
Standards connections
Students act as a rule-following recommendation algorithm, track topic counts across rounds, and test a rule change against their prediction.
- TEKS
- computational thinkingdata literacy, management, and representationTEKS sections: Technology Applications §126.17–§126.19, high school Technology Applications courses (19 TAC Chapter 126)
Reflect
- Do I stay safe and make good choices online? How do my likes and shares shape what I, and other people, see next?
- Why might a feed that shows only one kind of post make it harder to find trustworthy information?
- If I could rewrite one rule for the apps I use, what would it be, and who would object?
What comes next
For homework, students notice one thing their feed (or a family member's, with permission) keeps showing and write one sentence guessing which past action caused it. That observation opens the next lesson on persuasion and sponsored content.
Pairs well with
- Gen AI Literacy Breakouts (K–8)Follow-up breakouts on how AI systems learn from patterns.
- Critical Thinking Breakouts: Signal Check, The Viral ClaimPairs well: why viral posts spread, and how to check them.
