1. A video platform notices that users who watch one politically charged video are 73% more likely to be recommended a more extreme version next. This is best explained by:

• Deliberate censorship of moderate content by platform moderators.
• An engagement algorithm that optimizes for watch time using past behavior as signal.
• Users actively searching for increasingly extreme content through keyword queries.
• A legal requirement that related content be grouped together.

2. Two students at different schools search the same news event on their phones and receive substantially different top results. The most likely cause is:

• Different ISPs routing traffic through different DNS servers.
• Personalization algorithms using each student’s browsing history to rank results.
• The news event is classified differently under COPPA regulations for each user.
• Geographic bandwidth limitations that affect which servers respond first.

3. Consider the following statements about filter bubbles:
I. Filter bubbles result from algorithms optimizing for user engagement.
II. Filter bubbles can reduce exposure to viewpoints that differ from a user’s existing preferences.
III. Filter bubbles are only created by search engines, not social media feeds.

Which are correct?

• I only
• III only
• I and II only
• I, II, and III

4. A social platform adds a feature: “You might also like content from people with different views.” This change most directly attempts to address:

• Algorithmic bias caused by unrepresentative training data.
• The narrowing of information exposure created by engagement-based recommendation.
• Copyright issues with cross-platform content sharing.
• Metadata privacy concerns in content delivery.

5. Which scenario best illustrates an unintended consequence of recommendation algorithms?

• A streaming service recommends movies the user has not seen but rated similar to past favorites.
• A news aggregator shows only articles that confirm a user’s pre-existing beliefs, reducing their exposure to contradicting evidence.
• An e-commerce site displays “frequently bought together” item bundles.
• A music app creates a playlist based on the user’s most-played tracks.

6. A researcher finds that users who rely primarily on a single personalized news feed score lower on tests of awareness of opposing viewpoints than users who actively seek multiple sources. This finding most directly supports the claim that:

• Search engine ranking algorithms deliberately suppress minority viewpoints.
• Personalization that maximizes engagement can reduce the breadth of information users encounter.
• Users who score lower are less intelligent and therefore seek less diverse information.
• Recommendation algorithms are more accurate than human editorial selection.

7. A social media platform’s algorithm shows users more content similar to what they previously engaged with. Over time, a user who reads conservative news sees only conservative content. This is:

• A technical error in the recommendation system.
• A filter bubble — the algorithm optimizes for engagement and progressively narrows the user’s information exposure.
• An intentional political manipulation by the platform.
• Beneficial because users see content they find relevant.

8. Which action would most effectively reduce a user’s filter bubble on social media?

• Using a faster internet connection.
• Logging out while browsing to avoid personalization.
• Actively seeking out and engaging with diverse viewpoints and sources outside usual recommendations.
• Sharing more posts to increase algorithmic exposure.

9. Consider: I. Filter bubbles only affect political content. II. Recommendation algorithms optimize for engagement, which tends to reinforce existing user preferences. III. Users may not be aware that their information feed is algorithmically curated.

• II only
• II and III only
• I, II, and III
• I only

10. A news aggregator shows headlines algorithmically ranked by predicted user interest. Compared to a random sample of the day’s news, a user of this aggregator most likely:

• Receives a more balanced view of events because the algorithm finds the most important stories.
• Sees a narrower slice of news reflecting their existing interests and prior reading patterns.
• Discovers more diverse viewpoints because algorithms sample broadly.
• Gets identical news to all other users because the algorithm uses the same ranking formula.