[Book Review] David Sumpter, Outnumbered - Part 2

https://www.amazon.com/Outnumbered-Facebook-Filter-bubbles-Algorithms-Control/dp/147294741X

Part 2. Algorithms that influence us

Part 1 Algorithm that influence us was about showing algorithms; no matter how they analyzed a massive amount of data, their performance is as good as the data quality. Therefore, it is not magic. From Part 1, we could throw out blind trust in AI. However, this doesn't mean that algorithms are bullshit. From the recommendation algorithms of YouTube to target advertisement algorithms of Instagram, a castle built from numerous data heavily influence our daily life. Like👍 reactions in social media and filter bubble issues represent the penetration of algorithms into our social life.

 

1. Popularity Competition

 

Today, popularity and success not only comes from one's extraordinary abilities or skills. Instead, these largely depend on the first few exposures and reactions from other people. The initial response draws the following attention; finally, its popularity grows exponentially. This phenomenon can be found widely, for example, in the number of YouTube subscribers, best-seller books and citations in Google Scholar papers. Algorithms that work behind the scene can be called "like👍 addition models."

 

The author simulated this model's results by making an algorithm based on the book-purchasing scenario. In this model, an initial buyer chooses two books randomly. However, from the next turn, customers reference early purchases. The writer modelled this tendency by increasing the statistical possibility of buying formerly taken books. In contrast to the first twenty purchases showing fairly distributed results, results after five-hundreds of the trial were concentrated on specific authors. The total sales of the top 5 authors' books are similar to that of the twenty authors. However, if the initial purchase varied, the outcome was reproduced differently. These astonishing results remind us how our success is accidental and why we can not help ourselves to be humble in front of our prosperity.

 

On the contrary, another form of algorithm doesn't filter our access to information based on collective inclination. For example, FaceBook and Twitter will sort our friend's posts based on like reactions they received. Yet, the dating application Tinder algorithm is just the opposite of that. This is because Tinder was developed to match individual needs. Therefore, the user only manages simple behaviours, swiping to the left if they like someone or to the right if they dislike someone. Eventually, the simple algorithm will match the individual's preferences and alert them if they have co-likes. The Tinder algorithm is straightforward but noteworthy as it does not filter our exposure to news, posts, events, etc. Under the algorithm, we can control a random winner-takes-all system and information bias.

 

2. Filter Bubble

 

Filter bubble means filtered space created by algorithms based on ones like👍, web browsing and search records. Most social network services utilize this algorithm to provide personalized information. So, what is the underlying concept of this system? Two factors determine the visibility of specific posts and news in our feed: first, one's attention to the topic and second, intimacy between one and the person who shared the posts. To sum up, if the topic is something you have been interested in and the person who shared the news is close to you, the algorithm will post it very top of your feed. In other words, it has high visibility.

 

The problem is that once the algorithm captures your slight preference, it will keep showing posts related to your initial activities. Then, there are more possibilities for the user to click presented news because it has high visibility. Finally, the algorithm will fix your taste regardless of whether you are being centrist in real life. Once caught in such bubbles, you will be more difficult to expose to the opponent's thoughts, losing the ability to critically assess the information's truthfulness. 

 

However, different from exaggerated beliefs about filter bubbles and their effect on our rational decision-making, it turned out that conspiracy theories and fake news have little influence because most users discriminate fake from the truth. They didn't believe the content blindly. Furthermore, another research verified we are not emotionally vulnerable to social media content. They observed changes in social network users' attitudes when all of the positive content were intentionally erased from their feeds. The result was that they were barely affected by the manipulation. The usage of positive words in the experimental group was just 0.1 % point less than that of the control group. Therefore, we could conclude concerns that filter bubbles withdraw democracy are inflated.

 

Then why do we show such immunity against filter environment to some degree? The author guesses we are considerably open to different views due to our hobbies. Hobbies in this context mean various environments we experience other than our profession or political stance. To reference Bob Hockfeldt's comments, "We manage our lives in the various contexts defined by many social dimensions." To sum up, due to the complexity of our social relationship, most of the time, social media space cannot become a strong resonance room. However, this draws me to a further question: "What about platforms that lack social interaction?" For example, YouTube and Netflix only run based on users' tastes. Therefore, there are no rooms for a reference check. For instance, videos in my YouTube feed verifies I'm a solid left-winger. On the other hand, my grandmother's feed is filled with conservative viewpoints. However, there is little chance for me to be exposed to Grandma's like👍 videos because there is no social networking in the YouTube ecosystem. Therfore, the author's claim remains half ture.