The Coldewey TechCrunch article on Reddit subreddit bans provides a lot of insight regarding the issues raised in this chapter on moderation tradeoffs. In a study conducted by Georgia Tech, the results showed that banning subreddits on the basis of hate speech actually helped in reducing the occurrence of such activities on Reddit. People who remained after the ban greatly decreased their usage of hate speech while people migrating to other subreddits did not show an increase in such activities. However, what I find to be understated in this chapter is the effectiveness of the policy itself. Banning the hate subreddits may have 'solved' the issue, but the banned users merely migrated to sites such as Voat and Gab, which openly encourage such activities.

The fact that 'quality' is dependent on the target audience of the platform is a perspective I need to take into consideration when thinking about moderation in the future. Before reading the article, I assumed that content moderation was an absolute necessity for all platforms since certain content could be classified as 'bad'. However, the 4chan example showed that content moderation may also involve a very different form of moderation based on a new definition of quality. This brings up the question of whether content moderation is truly ethical or simply another aspect of doing business with one's own target market demographic.

The Meyer Atlantic article about Facebook’s secret mood experiment seems truly disturbing from the perspective of this chapter. In 2012, the company modified the news feed of almost 700,000 users, exposing some to more positive stories while others to more negative stories. This study revealed that the exposure to either kind of news feed affected the extent to which individuals shared positive or negative sentiments. I think what makes this experiment even more disturbing is the paradox – the chapter explains how social media can contribute to the improvement of mental well-being; yet, in this case, the social networking site chose to make users feel bad without any user awareness and consent. When asked about the significance of the experiment's findings, the lead author mentioned the minimal effect in real-life contexts. However, as stated by one school of social work's dean, users who were emotionally vulnerable could become depressed or anxious due to the manipulation in news feed, and no one would ever know.

It certainly hits close to home for me – I believe the difference between seeking out community versus simply being discovered by an online community is often overlooked. The case provided by Fiesler is particularly relevant since it shows that she was not simply scrolling; instead, she went looking for her own community. This element of agency seems to be crucial for whether or not being in an online community positively impacts mental health. Is it possible that being forced into communities through popular platforms such as TikTok or Instagram Reels may be even less effective for this purpose than old-school forums or niche spaces?

The article in Wired by Lauren Goode exemplifies how recommendation algorithms are oblivious to the complexity of real life. The author had broken up from her partner, called off her wedding and got divorced, yet she kept receiving wedding offers and anniversaries on the platform. To me, the most striking part is the fact that we are looking at an extreme case but one that exposes a problem that cannot easily be overlooked. Algorithms are trained to see patterns, not individuals. That makes me think about whether it would make sense for social networks to make it possible for people to exclude themselves from specific life events or topics.

This explanation of recommendation algorithms really shows how much decision-making takes place behind the scenes every time we use an application. What strikes me as strange is the fact that the word 'recommendation' somehow sounds a little too neutral – as if we are receiving a favor from the system in the form of recommendations. In fact, recommendation algorithms may be focused on maximizing some parameters, which do not necessarily correspond to the needs and interests of the users. It would be interesting to know whether there is any kind of responsibility regarding disclosure of criteria behind the algorithms.

Social Model of Disability, referenced in this chapter, emphasizes the difference between individuals' impairment and barriers to accessibility imposed on people by society. Instead of treating disability as a flaw or a deficiency, Social Model of Disability states that individuals become disabled through social and physical barriers rather than impairment. Thus, people are disabled by society, not their disability; that is to say, disability becomes a barrier imposed by society. To me, it seemed to be an interesting approach, especially taking into account social media which either promotes or eliminates barriers.

Masking is another term which clearly emphasizes on the harsh reality that we live in an environment which is very much unwilling to accommodate its inhabitants. If someone feels like he or she needs to disguise their own self just to make their way in the environment, then such an act reflects the society rather than the individual who suffers from a disability. In my opinion, the phrase "mental or physical toll" stands out because society admires individuals who "push through" but does not question why they have to push through.

The Mary Sue story about Facebook workers exploiting their privileges to spy on women is horrifying. It relates very much to the chapter's message that our 'private' information is not really private. The most disturbing aspect of this story is that this was not some kind of breach or hack, but rather an internal affair where insiders used their privilege to spy on women for personal gain. This leads one to wonder whether the greatest privacy risk we face is not big companies and governments themselves, but rather the individuals working inside them.

The example of sometimes wishing companies to access our personal communications, such as during situations involving threats of death, emphasizes the issue of privacy. Privacy is generally perceived as a right that only needs protection. However, this particular example opened my eyes on the matter and made me realize that sometimes there may come a time when it would be to our benefit to sacrifice a little bit of privacy. Where do we draw the line in terms of whether or not accessing personal communications is allowable?

The Vox article written by Kurt Wagner highlights that Facebook creates 'shadow profiles' for those individuals who have never signed up for any Facebook account. They track the movements of such individuals using different tactics like placing tracking pixels on third-party sites and uploading contact information submitted by other users to Facebook.

This is one of the most disturbing aspects discussed in the chapter. The revival of discredited pseudoscience like physiognomy under the guise of 'AI' illustrates how easily a new technology can be used as a vehicle for laundering dubious concepts. Wrapping phrenology in algorithms will not make it a scientific concept; it only makes it easier for the ordinary citizen to object, since there is an algorithm backing it now. Rather, the use of AI in physiognomy is far more damaging than its original form, because it enables scaling and implementation in applications such as employment assessment or security screening in ways that 19th-century phrenologists would have never thought possible.

In his article published in the journal Scientific American, Whitney Phillips claims that trolling did not appear on its own. It is, in fact, a copy of the actions carried out by mainstream media organizations. By looking at what is being done by the media industry, trolls have found out how to troll people on the internet.

In this regard, I agree with Film Crit Hulk completely. "Don't feed the trolls" has always sounded like the equivalent of advising a victim to take the beating rather than the advice to stop the attacker from committing an attack in the first place. This is an inversion of logic, since platforms have both the power and means to ban offensive users but choose not to do so out of the economic benefit of engagement and the financial burden of moderating.

This article is about a fake COVID death story, which is an example of how fast misinformation can spread on social media when the story cannot be verified. What I find concerning is that the false story gained a lot of attraction before anybody did their due diligence on the story, showing how even when the truth comes out, damage may already be done.

To me, it is very interesting to see the phenomena of catfishing develop as a result of the digital age. This phenomenon definitely seemed much less popular in the past, with the invention and surge in popularity of social media, we are seeing how harmful this is starting to become, especially with the development of generative AI. Personally, with this growth, I believe platforms do not do enough to prevent catfishing, where innocent people get hurt thinking the person they are interacting with is real. Furthermore, this leads to safety issues that affect users on the platform, so social media can benefit from adding more catfishing prevention features.

Including the segment from Drunk History regarding John Adams and Thomas Jefferson as part of a lesson about social media history is quite fascinating. I am reminded that political antagonism and the use of media as a tool of the trade did not begin with modern day social media sites, but Adams and Jefferson used papers and pamphlets for their own political gain as well.

This shows how email was one of the most significant moments in the evolution of communication. The ability to transport information digitally increases the convenience factor. This reminded me of how e-mail wasn't always a public tool; in the 1960s-1970s, it was a private tool used by researchers in the U.S. military.

I was reading the Julia Evans article, and what stuck out to me was specifically the odometer example. This was supposed to be a program that tracked 10,000 km, but it only actually tracked 262 km because floating-point errors kept accumulating over time. I originally thought of floating point as a super technical concept, but this example makes it seem much clearer and less technical to me, and makes me think twice about how much we should trust computers when it comes to accuracy.

Learning about metadata made me realize how much information I have that is going out to the public that I do not even know about. Every photo I take shares a location and a timestamp, and most apps track when and how the app is being used, so it sort of bothers me how so much of my information and schedule is going out to these companies, which can potentially sell it to other groups or organizations.

The Huck Magazine provides an article that goes over click farms and added an observation that I did not expect earlier. It isn't just bots on automated scripts running servers, rather it is actual humans manually liking and following posts at a large scale. This blurs the distinction between "bot" and "human" which I believe this chapter doesn't fully address since it is the artificial manipulation of metrics, regardless of who does it.

This stat that states how over half of the negative tweets about The Last Jedi were from bots or accounts that were politically motivated make me rethink how I interpret online backlash. Whenever I see this online, I assume it is a group psychology, but this statistic has showcased to me that a lot of this really just comes down to bots. This makes me wonder how often public opinion is shaped without anyone noticing.

The idea of deontology seems difficult to apply in real life because situations vary, and you aren't always able to apply a strict set of rules. For example, in social media, people may feel obligated always to tell the truth. In doing so, this could spread negativity and be seen as harmful, suggesting that consequences may need to be seen alongside different moral rules.

This section of consequentialism is interesting to me since it talks about how actions are judged based on the total of their consequences, sort of acting as a cumulative judgment. Personally, I do not agree with the idea; however, consequentialism can be seen as problematic since it can justify harming a small group of people as a justification for supporting a larger group. For example, using social media to maximize engagement could lead to promoting harmful content for maximum exposure.