Chapter 21 Exercises: Personalization, Filter Bubbles, and the Algorithmic Self

Instructions

These exercises engage with filter bubble and personalization concepts through personal observation, empirical research, critical analysis, creative work, and structured discussion. Exercises are labeled by type: [Reflection], [Research], [Analysis], [Creative], and [Group Discussion].

[Reflection] Your Personalization Audit Spend 30 minutes scrolling your primary social media platform's feed without clicking anything. Categorize every piece of content you see: What topics appear? What is the political valence (if any political content appears)? What types of accounts are represented? How many items feel surprising or outside your expectations vs. highly predictable? Write a one-page reflection on what your feed's composition reveals about how the algorithm has modeled you.
[Research] Pariser's Original Argument Read Eli Pariser's The Filter Bubble: What the Internet Is Hiding from You (2011), or at minimum the first three chapters, and watch his TED Talk of the same name. Write a 500-word evaluation: What is Pariser's strongest argument? What has held up empirically since 2011? What aspects of his argument seem less persuasive now, given subsequent research?
[Analysis] Filter Bubble vs. Echo Chamber Distinction The chapter distinguishes between filter bubbles (algorithmic) and echo chambers (social). Choose a specific community you belong to or are familiar with — a neighborhood, a religious community, a professional network, a college campus. Analyze the community through both lenses: What filter bubble effects (algorithmic curation) affect what members know? What echo chamber effects (social selection) affect what members know? Which mechanism do you think is more powerful in this community?
[Research] Bail et al. (2018) Primary Source Find and read Bail et al.'s 2018 PNAS paper "Exposure to opposing views on social media can increase political polarization." Write a 600-word analysis: What was the study design? What were the main findings? What are the study's limitations? How do you reconcile its counterintuitive findings with the filter bubble concern?
[Creative] The Pariser Test Create two social media accounts (on the same platform, if the platform allows it) or two Google accounts. Over two weeks, curate very different engagement profiles on each: one that consistently engages with progressive political content, one that consistently engages with conservative political content. After two weeks, search for the same three politically contested topics on each account. Document what is different about the results, recommendations, or news feed content. Write a 600-word case study of what you found.
[Reflection] Your Algorithmic Self Portrait Based on your engagement history on your most-used social media platform, write a description of "the user the algorithm thinks you are." What interests, political views, consumer preferences, and demographic characteristics would the algorithm infer from your behavior? How accurately does this portrait match who you actually are? Where does it diverge? What does the divergence reveal about the limits of behavioral profiling?
[Group Discussion] Is the Filter Bubble Overstated? Prepare for a structured debate: One half of the group argues that filter bubbles are a major epistemic crisis requiring urgent intervention; the other half argues that filter bubbles are largely an elite concern overstated by researchers and journalists who don't account for the full information environments people occupy. After 20 minutes of debate, discuss as a group: What is the most empirically honest position on filter bubble magnitude?
[Research] Collaborative Filtering Explained Research how collaborative filtering algorithms work at a technical level. Explain in 400 words, for a non-technical audience, what collaborative filtering is, how it creates "people like you" recommendations, and what its specific limitations and failure modes are. Include a specific example of how collaborative filtering could create a filter bubble effect without any explicit political targeting.
[Analysis] The Feedback Loop in Practice Choose a topic that you started engaging with on social media within the last year that you didn't engage with before. Trace the feedback loop: What initial exposure led to engagement? How did the algorithm respond to your engagement? What was your feed like around this topic after one week, one month, and three months of engagement? Has the topic crowded out other content? Write a 500-word analysis of the feedback loop's operation in your case.
[Research] Identity Lock-In and Algorithmic Bias Research at least two documented cases of algorithmic personalization systems producing biased or stereotyped recommendations based on demographic characteristics. Cases might include: Amazon's hiring algorithm, Spotify's recommendation biases, YouTube recommendations and racial categorization, or Facebook ad targeting and discrimination. Write a 500-word analysis connecting these cases to the identity lock-in concept described in the chapter.
[Creative] Design an Anti-Filter-Bubble Platform Design a social media platform or news aggregation service specifically intended to counteract filter bubble effects. Your design should address: How does the recommendation system work? How does it balance personalized relevance with epistemic breadth? How does it handle the Bail et al. finding that cross-cutting exposure can increase polarization? Write a design brief of 600-800 words describing your platform.
[Group Discussion] Editorial vs. Algorithmic Curation The chapter argues that algorithmic news curation represents a governance shift — from accountable editors to unaccountable algorithms. In groups, discuss: Is editorial news curation actually more accountable to public interest than algorithmic curation? What mechanisms of editorial accountability exist? What mechanisms of algorithmic accountability exist? Which system do you find more trustworthy for civic information, and why?
[Reflection] The Local News Gap The chapter describes Maya's experience of being politically aware about national issues but largely uninformed about her local community's civic affairs. Honestly assess your own knowledge: What do you know about your city council? Your state legislature? Local housing, education, and transportation policy? Local court decisions? Where did you get this information — from local news sources or from social media? What gaps do you notice?
[Research] Cross-Platform Data Sharing Research the extent of Meta's cross-platform data collection across Facebook, Instagram, WhatsApp, and Messenger. Find documentation of how behavioral data from one Meta platform is used to personalize experiences on others. Also research "off-Facebook activity" — the data Meta collects from third-party websites and apps. Write a 400-word summary explaining what the typical Meta user's data profile consists of.
[Analysis] The Bakshy et al. (2015) Controversy Research the controversy surrounding Bakshy, Messing, and Adamic's 2015 Science paper on Facebook's News Feed and ideological diversity. Find: the original paper, at least two critical responses to the methodology or findings, and Facebook's public commentary on the research. Write a 500-word analysis: What did the study find? What are the most valid criticisms? What would you need to trust the findings more?
[Creative] The Algorithm's Version of You Write a short story (800-1,000 words) told from the perspective of an algorithm's model of a user. The story should be written in first person as the algorithm — describing what it "knows" about the user, how it curates the user's experience, what it gets right, and what it gets wrong. The story should make the personalization dynamic vivid and emotionally resonant.
[Group Discussion] Serendipity and Discovery Discuss as a group: Think about the most important ideas, cultural works, or people you have discovered in your life that significantly shaped who you are. How did you discover them — through personalized recommendation, through serendipitous encounter, through someone else's recommendation, through random browsing? What does your collective experience suggest about the role of serendipity in epistemic growth? What does it suggest about the value and limits of personalized recommendation?
[Research] The Reuters Institute Digital News Report Find the most recent Reuters Institute for the Study of Journalism Digital News Report (available at reutersinstitute.politics.ox.ac.uk). Look specifically at the data on how different age groups and countries access news, how much news comes through social media vs. traditional sources, and what levels of concern exist about information quality. Write a 400-word summary of the most relevant findings for understanding how personalization affects news consumption at a population level.
[Analysis] Location Data and Personalization Research how social media platforms use location data to personalize content and advertising. Find the relevant sections of Facebook's and Google's privacy policies describing location data use. Also find any academic or journalistic research on specific cases where location data was used for personalization in ways that surprised users. Write a 400-word analysis of what "ambient personalization" through location data means for the filter bubble model.
[Reflection] Trying to Escape For one week, make a deliberate effort to see content outside your typical social media filter. This might involve: following accounts with different political perspectives, searching for topics you don't usually engage with, reading news sources with different editorial orientations than you typically use, or using your platform's "explore" or "discover" functions. At the end of the week, write a reflection: How difficult was it? Did you encounter anything genuinely surprising or valuable? Did any of the cross-cutting content activate defensiveness (consistent with Bail et al.)?
[Research] TikTok's Algorithm and Personalization Research TikTok's recommendation algorithm, specifically the For You Page (FYP) system, which is widely regarded as the most accurate personalization system in social media. Find: any available documentation of how the FYP works, academic or journalistic research on its effects, and any evidence of filter bubble effects (or their absence) on TikTok compared to other platforms. Write a 500-word analysis comparing TikTok's personalization model to the general filter bubble model described in the chapter.
[Group Discussion] The Democratic Stakes The chapter suggests that dramatically different information environments among liberal and conservative social media users may have consequences for democratic deliberation and shared governance. Discuss: How much information fragmentation can a democracy tolerate? What historical cases of severe information fragmentation (within countries, not between them) can you identify? How were those situations resolved or not resolved? What does history suggest about the viability of democracy with extreme information asymmetry?
[Analysis] The 2020 Election Information Environment Research the academic and journalistic documentation of information environment differences between liberal and conservative social media users during the 2020 US election. Find at least two studies documenting specific content asymmetries. Write a 500-word analysis: What were the most significant documented differences? What were their likely sources (filter bubbles, echo chambers, or both)? What were their potential democratic consequences?
[Creative] Redesign the News Feed for Epistemic Health You have been asked to redesign a social media platform's news feed specifically to promote "epistemic health" — the conditions for informed democratic citizenship. This means: exposure to diverse political perspectives, reliable local civic information, high-quality rather than merely engaging political content, and awareness of one's own personalized information environment. Write a detailed design specification describing your redesigned feed, the tradeoffs you've made, and how you would measure success.
[Research] Algocracy and Content Governance Research the concept of "algocracy" — governance by algorithm — as it applies to information environments. Find academic or policy analysis of what it would mean to subject recommendation algorithms to the same transparency and accountability requirements as public broadcasting or news editorial standards. Write a 500-word analysis: What would algorithmic accountability look like in practice? What models (from other sectors) could inform its design?
[Reflection] Your Information Ecosystem Map Draw or describe a map of your complete information ecosystem — every source through which you regularly receive news, information, and ideas. Include social media platforms, news websites, podcasts, newsletters, conversations, television, and any other sources. Categorize each source by: how personalized it is, what political/ideological orientation it has, whether it is primarily local or national/global, and how you came to use it. Analyze your map: How diverse is your information ecosystem? Where are the filter bubble risks greatest?
[Group Discussion] Platform Transparency and User Control Some researchers and advocates argue that users should have more transparency about and control over their personalization — being able to see why they are shown particular content and actively adjust their personalization profiles. Others argue that full user control over recommendation creates burdensome cognitive labor and may not actually improve epistemic outcomes. Discuss: What level of transparency and control over personalization should platforms be required to provide? What should be the default? Who should make these decisions?
[Research] EU Digital Services Act and Filter Bubbles Research the EU Digital Services Act (DSA, 2022) provisions specifically related to recommendation algorithms and personalization. What does the DSA require platforms to do regarding: transparency about recommendation systems, user control over personalization, and risk assessment for systemic risks from recommendations? Write a 500-word policy analysis: Are the DSA's provisions likely to meaningfully address filter bubble effects? What are their limitations?
[Analysis] Comparing Personalization Strength Across Platforms Compare the personalization approaches of five platforms — Facebook, Twitter/X, TikTok, LinkedIn, and Reddit — based on available documentation and academic research. For each: How does the recommendation algorithm primarily function? How much user control is available? What signals does it primarily use? Based on your analysis, rank the platforms from most to least likely to create significant filter bubble effects for a politically engaged user.
[Creative] The Serendipity Proposal Write a proposal for a "serendipity tax" — a regulatory or platform requirement that a certain percentage of every user's recommended content must be outside their established engagement profile. Your proposal should: define what qualifies as "outside profile," specify what percentage would be required, address the Bail et al. concern about cross-cutting exposure increasing polarization, explain implementation challenges, and address free speech and platform autonomy concerns. Present your proposal as a policy brief.
[Reflection] After Reading This Chapter Before reading this chapter, how did you think about your social media information environment? Were you aware of personalization? Did you think of your feed as representative of social reality or as a curated selection? How has reading this chapter changed (or not changed) how you think about your personalized feed? Write a one-page reflection on what you've learned and what it means for how you will use social media going forward.
[Research] Personalization and Health Information Research documented cases where social media personalization affected health information exposure. Look for cases involving: vaccine hesitancy content, mental health information, dietary advice, or medical misinformation. Write a 500-word analysis of how filter bubble effects in health information specifically can cause harm, using specific documented cases.
[Group Discussion] Who Should Govern Personalization? Currently, personalization algorithms are governed primarily by platforms — private companies — with limited regulatory oversight. Discuss: Who should have authority over how personalization algorithms work? Options include: platforms themselves (with transparency requirements), government regulators, independent oversight bodies, civil society organizations, or users themselves through democratic processes. What are the strengths and weaknesses of each governance model? Which does your group find most appropriate?
[Analysis] The Algorithmic Self and Personal Growth The chapter describes identity lock-in as a potential limitation of personalization — the algorithm models who you were, not who you are becoming. Think about your own history of significant personal change: political evolution, changes in values, interests that you've moved away from or toward. For each change: Would an algorithm trained on your behavioral history before the change have helped or hindered the change? Write a 500-word analysis of personalization's relationship to personal growth and change.
[Creative] A Letter from 2035 Write a letter from yourself in 2035 to a younger person entering adulthood. The letter addresses: how personalization algorithms have developed since 2026, what the long-term consequences of the filter bubble era have been (imagine both optimistic and pessimistic scenarios), what you wish you had understood about your personalized information environment when you were young, and what practices for maintaining epistemic diversity you have found valuable over time. The letter should be warm, specific, and honest.