The Block: CASE 9
Introductory algorithmic literacy
1: STORY
Two students search the same topic and receive different recommended videos. That difference raises a useful question about what the platform might know from earlier activity. It does not, by itself, reveal the exact reason one video appeared above another. A recommendation diary can make possible inputs visible without pretending to read the platform's formula.
Malik's practice log records five basketball-shoe clips watched to completion over two days, two likes, and a search for one brand. Those actions are observable signals. Watching to the end could reflect interest, curiosity, distraction, or something else. The log cannot directly measure Malik's private reasons or prove what he wants to see next.
The platform explainer lists categories that recommendations may use: viewing, likes, follows, searches, location and language settings, and information about similar users. It does not say how much each category counts. The comparison observation shows mostly general trending clips in a fresh logged-out browser, but several conditions changed together, including account state and available history.
For a 24-hour diary, record recommendation topics, time, account state, and relevant actions without collecting private message contents or classmates' histories. A paper chart can record observations from teacher-provided practice screenshots. Compare patterns cautiously, and name alternative explanations. A useful ending might identify viewing completion as a possible signal while explaining why this diary cannot establish that it caused a particular recommendation.
2: EVIDENCE
These are NPW practice documents. Real-world sources are credited separately below.
SOURCE A — Recommendation log
NPW practice document. Over two days Malik watches five basketball shoe clips to completion, likes two, and searches one brand.
What it establishes: Shows interaction signals the platform could use. It does not prove the exact ranking formula.
Question while reading: Which actions might be signals rather than preferences you consciously chose?
SOURCE B — Platform explainer excerpt
NPW practice adaptation of common platform disclosures. “Recommendations may use viewing, likes, follows, searches, location/language settings, and information about similar users.”
What it establishes: Shows categories of signals. It does not reveal exact weights.
Question while reading: What remains unknown about why one post ranked above another?
SOURCE C — Compare account
NPW practice observation. A fresh logged-out browser shows mostly general trending clips instead of shoe content.
What it establishes: Supports the idea that personalization matters. It does not isolate which signal caused the difference.
Question while reading: What would a stronger test hold constant?
Source A provides concrete interactions that fit signal categories in Source B: completed viewing, likes, and search. Their match supports a hypothesis that these activities could influence recommendations. The explainer's “may” leaves the actual inputs and their weights unresolved. Neither record exposes the ranking calculation for a specific clip.
Source C strengthens the possibility of personalization because its logged-out feed differs from the shoe-focused context. It is not a controlled test of one signal: account state, history, and potentially other conditions vary together. A diary should preserve those differences rather than label one of them the proven cause. The platform controls ranking; users can document patterns and consult its stated controls.
3: PRACTICE
C shows a different feed under different conditions. B lists several possible signals, and neither source supplies exact causal weights.
Responses stay in this browser. Use “Download my notes” to keep a copy.
4: TASK
Keep a 24-hour recommendation diary and identify possible signals. Include one clear claim or question, at least two source references, one statement of uncertainty/scope, the person or office with relevant authority, and one realistic next step.
My 24-hour diary will have columns for time, recommended topic, account state, and preceding actions such as search, like, or completed viewing. Source A suggests these actions are worth tracking; Source B lists them as possible signals. Source C's logged-out feed offers a comparison, but it changes several conditions at once. I will mark possible connections, not claim an exact formula or invent results before collecting them. The platform determines ranking and can clarify its available recommendation controls. My next step is to record a small set of observations using the same categories throughout the day, or equivalent teacher-provided snapshots on paper.
If you wanted to test whether language settings affect recommendations, which conditions would you try to hold steady and what would remain uncertain?
A host-led explanation of recommendation feeds using TikTok and a computer-science interview. The older publication date makes it useful for comparing enduring mechanisms with changing platform features.
Which part of the explanation is supported by an expert interview, and what current platform documentation would help test it?
Create a paper feed from teacher-provided post cards. Change one engagement signal and explain why the next recommendations might change. Do not require students to reveal their personal feeds.
Make: A six-card feed map with two evidence-backed explanations.
Public PBS episode page. PBS lists CC and a transcript; playback and school-network access still need checking.
45–55 min core, 80–100 min full
| Criterion | Beginning | Developing | Strong | Advanced |
|---|---|---|---|---|
| Accuracy & scope | Repeats claim. | Mostly accurate; scope incomplete. | Accurate and properly scoped. | Accurate, scoped, and explicit about uncertainty. |
| Evidence | Little or unrelated. | Relevant source, sometimes overstated. | Explains what sources establish. | Compares sources and limits. |
| Authority | Names actor without role. | Plausible authority. | Explains who decides and why. | Maps proposal, approval, implementation, review. |
| Communication | Vague or audience-mismatched. | Main point understandable. | Clear, specific, audience-aware. | Clear, accessible, precise, well-qualified. |
