I appreciate how Chiang reframes the fear of AI “taking over” by comparing it to management-consulting logic rather than superintelligence. His argument that powerful institutions often use technology as a justification for harmful decisions — rather than technology making those decisions itself — really stuck with me. It made me think about how often companies claim, “The algorithm says we have to do this,” the same way executives once said, “McKinsey says we have to cut costs.”

I think this kind of moral courage is undervalued in discussions of “tech ethics.” Too often we focus on the abstract risks of algorithmic bias, misinformation, privacy — but forget that behind every system is a person (or a team) making design choices. The chapter’s call for potential tech workers to reflect on these issues — and to be ready to push back — feels deeply important.

Doctorow’s idea of enshittification—that platforms inevitably degrade as they shift value first toward users, then toward business partners, and finally toward extracting maximum profit for themselves—really stuck with me. It feels like an unusually clear description of how many social-media sites evolve over time.

How might capitalism look different (in theory or practice) if “free market” were more tightly regulated — for example, with labor protections, antitrust enforcement, or community ownership models? Does that still count as capitalism?

I find its framework interesting — they don’t just dismiss shaming out of hand, but carefully analyze when and how it might be morally justified. Their conditions (like proportionality, necessity, respect for privacy, non-abusiveness, and reintegration) seem really well suited to thinking about social media shaming.

I appreciate how it doesn’t shy away from the limits of reconciliation. By comparing the Nuremberg Trials with the South African Truth and Reconciliation Commission, it highlights how not all wrongdoing can be “fixed” in the same way — and sometimes punishment, not forgiveness, is what justice requires.

Jacobson’s work impressed me because it shows how harassment can become a kind of background radiation in someone’s life—persistent, invisible to outsiders, and extremely draining. What stood out most is how the cyberstalker’s behavior followed her across platforms and into adulthood, showing that online harassment isn’t always an isolated flare-up; sometimes it becomes a long-term pattern that shapes someone’s sense of safety.

I find the framing morally compelling — it challenges the common notion that only “illegal” speech is worth regulating or worrying about. Harassment that skirts legality can still deeply damage people. It underscores the responsibility of platforms and communities to think beyond just what is legally prohibited, and consider what is socially or ethically harmful.

Foldit is an interesting case because it shows that crowdsourcing isn’t just about distributing simple microtasks — sometimes the crowd can outperform algorithms on highly complex scientific problems. I love how Foldit turns protein-folding into a game, and how ordinary players (not trained biochemists!) have actually contributed to real scientific discoveries.

This defines crowdsourcing pretty broadly — but are there important ethical differences between volunteer crowdsourcing (like Wikipedia) versus paid microtask platforms (like paid data labeling)? How should we think about those differences, especially in a course about ethics and automation?

I think the labour issues Roberts raises are under‐recognized. As a user I rarely think about the human moderators behind the platform I use. It made me more aware of the ethics of consumption: the “clean” feed I enjoy is enabled by invisible people working in difficult conditions. That leads me to a question: How much responsibility should platforms bear toward these moderators’ mental health, and how transparent should they be to end users about who does this work?

I wonder if the dichotomy “trained vs untrained” might be easier said than done: even trained moderators face huge grey areas and emotional toll. So “training” helps, but doesn’t eliminate risk.

This article by Rick Paulas provides a visceral, first-hand feel for what it means to go “viral” in social media contexts—describing it as something like “doing cartwheels on the water-spout of a giant whale.” That kind of metaphor really brings home how thrilling yet unstable virality is: fun, exhilarating, but also out of control and potentially dangerous.

As someone who uses social platforms and watches how memes / posts spread, I’ve observed that sometimes the version that goes viral isn’t the original but a mutated one (someone adds a caption, remix, or cross-posts to another network). The chapter’s point that inheritance matters jumped out: once a variation exists and spreads with the change, future copies carry that change. That resonates with seeing e.g. a tweet being quote-retweeted, then everyone repeats the quote-tweet version, not the original tweet.

It struck me how relevant this paper is to the chapter’s point that recommendation algorithms don’t just serve content but shape what we see and how we interpret it. The study shows how digital media can drive affective polarization via “partisan sorting” — which nicely connects to the chapter’s warning that algorithms can deepen divisions by reinforcing “you vs them” dynamics.

In my own experience, I’ve seen both “good” and “bad” recommendations: one time a platform surfaced a deeply niche topic I had read once and I found it fascinating; another time it repeatedly suggested things totally irrelevant and frankly annoying.

It's interesting to see it explores how accessibility features aren’t just “extra” for a few users, but foundational for designing inclusive social-media environments. It challenged me to rethink how often accessibility is treated as an afterthought rather than baked in from the start.

I really appreciated how the chapter highlights different approaches to accessibility — making the environment work for all, adapting tools for users, and the burden often being placed on the user instead of the design. One question I have is: what’s a practical checklist or metric developers could use early in the UI design to shift from the “modifying the person” model to the “making the tool adapt” model?

I found the inclusion of Bowman’s article about the 530 million-user breach striking — it grounds the discussion of privacy in real, large-scale harm rather than abstract theory. From my experience, seeing such breaches makes the “privacy isn’t just about secrets, it’s about control and trust” line hit home.

I found how users often feel they’ve lost control over their data interesting — it reminds me of the moments when I accept a “Cookie/Privacy” pop-up without really reading it, then later wonder how much the platform knows about my interests.

I found Catherine Stinson’s “The Dark Past of Algorithms That Associate Appearance and Criminality” especially compelling — it highlights how seemingly neutral data-mining efforts (for example facial recognition or risk scoring) embed deep historical biases and reinforce harmful associations. It makes me reflect: when we apply mining methods in social-media contexts, it’s not just about data quality but also which associations we’re willing to carry forward.

I found the discussion of unintended versus intentional data poisoning especially striking — it reminded me of how a viral trend on a platform can distort a research survey in ways the authors likely never anticipated. One thing I’m wondering though: given that many social-media datasets are collected passively and opportunistically, how can researchers realistically detect when the data has already been poisoned by normal platform usage (rather than a malicious actor)?

I really appreciated the inclusion of Whitney Phillips’s “Internet Troll Sub-Culture’s Savage Spoofing of Mainstream Media” (Scientific American, 2015) — it gives a grounded, cultural perspective on trolling that complements the ethical discussions nicely. I wonder whether there are more recent studies that explore how trolling has shifted into coordinated “political influence” efforts that would update or extend Phillips’s observations.

I think the section on Trolling and Nihilism raises an important point that some trolling communities don’t just push boundaries playfully but actually seem to treat ethics as irrelevant. What struck me is how this can lead to a kind of moral vacuum where the harm to people or groups is dismissed as part of the “game” of disruption.

The varianceexplained.org analysis of Trump’s tweets is a great example of how data science can uncover patterns behind online personas — it really connects to this chapter’s discussion of authenticity. I found it fascinating how statistical language analysis can expose who’s “really” speaking behind a public account, showing that authenticity online can sometimes be measured rather than just perceived.

I was struck by the tension in the chapter between authenticity as honesty and authenticity as the matching of presentation and interaction. The idea that social media personas can “feel” authentic even when they’re partly performed makes me wonder: in platforms driven by algorithms and engagement metrics, do we discourage real authenticity because performative authenticity “wins” more often?

It’s cool to see Tom Standage’s Writing on the Wall: Social Media — The First 2,000 Years cited here. His historical framing helps us see that social media isn’t entirely new, only evolved. I’d also suggest adding Shoshana Zuboff’s The Age of Surveillance Capitalism as a counterpoint source: it connects design history with power, data extraction, and economic incentives, deepening the discussion of how “design” choices embed commercial values.

I appreciated the tension you raised around friction vs. frictionless design — it really made me think, on one hand, low friction feels like “good usability,” but as you note, design can deliberately add friction to push users toward more thoughtful behavior. I wonder: is there a risk that too much deliberate friction becomes paternalistic or manipulative (assuming users can’t be trusted)?

I’m glad to see Design Justice: Community-Led Practices to Build the Worlds We Need by Sasha Costanza-Chock in the bibliography. This work deeply connects design choices with power, oppression, and equity. It would be interesting to more explicitly surface how “data simplification” (in 4.2) is a design choice that can perpetuate injustice.

I really appreciated how it highlights that any dataset is already a selection — a simplification— and that those choices shape the story the data can tell. And I wonder: in practical terms, how might we audit or compare different simplification choices across platforms to surface which versions produce more (or less) ethical distortions?

It’s clever because it both acknowledges that bias is baked into datasets and algorithms, yet still tries to offer practical design strategies to deal with harms. I’d love to see people push further by testing or critiquing Jeong’s prescriptions in real bot designs: do they scale? Do they fail under adversarial inputs?

I liked the donkey metaphor — it really drives home how bots (or other tools) can act without “understanding,” which loosens the connection between intention and consequence. It makes me wonder: when a bot does something harmful (even unintentionally), should we treat that more like an accident or a responsibility failure — and how much should we hold the creator, operator, or even the platform accountable?

I really liked the metaphor of ethical frameworks as “tools” for seeing different aspects of a situation. It emphasizes that no single moral theory will show everything. One question I had: how do you choose which pair of frameworks to apply in a given scenario, and what criteria should guide that choice?

One additional ethics framework that is worth including is Discourse Ethics. This framework emphasizes the ethics of communication: acting only under norms that all affected can accept in rational discourse. It helps especially in social media settings, because these platforms involve many stakeholders (users, platform designers, advertisers...) engaging via mediated communication. Discourse Ethics adds tools for evaluating whether moderation policies, recommendation algorithms, or community norms are acceptable to all affected rather than simply optimizing outcomes or following abstract rules.