An algorithm is a series of mathematical instructions or a set of rules you follow to solve a problem or complete a task. Almost all our daily activities can be considered an algorithm, from your morning routine to the assembly instructions for your IKEA furniture. Nowadays, it is also used in a narrower definition, as a mechanism that selects and recommends data in advance to increase relevancy and usefulness for each user. As most are aware, algorithms are used in search engines such as Google, and social media platforms such as YouTube, Instagram, Twitter and Facebook. They are even used in your GPS navigation systems, web banner advertisements, and auto-correction systems. Furthermore, it is now starting to be used in institutions as a tool to determine employment status, personnel evaluations, credit ratings, and even the length of a criminal sentence.

Since algorithms are protected by the Patent Act and are business secrets under the Unfair Competition Prevention Act, the operation of each company's algorithms are not known specifically. However, most of them use 'content-based filtering' and 'collaborative filtering' to screen information on those bases. ‘Content-based filtering’ recommends information to users according to the characteristics of their previously preferred viewed content. ‘Collaborative filtering’ makes recommendations based on the assumption that people with the same tendencies or characteristics make similar choices or repeat their previous choices. The recipe for the algorithm differs based on the characteristics and needs of the company. For example, Facebook prioritizes customer engagement, leading them to arrange posts in popularity, content type, relationship, and recency. Pinterest on the other hand, serves the purpose of an archive, sharing ideas and photos with unrelated individuals. Therefore, it is based on relevance, quality of the Pin, domain quality, and the quality of the Pinner.

Although algorithms are necessary for filtering out irrelevant and low-quality content, there are severe problems with it that need to be pointed out. One of the side effects algorithms can cause on society is the risk of expanding discrimination. A joint research team at the University of Chicago and MIT found that company response rates to resumes were 50% higher when typical white names were used. This happened because human values and prejudices were reflected in the process of designing what data to collect and what criteria to be evaluated. Algorithms also make decisions based on years of accumulated data that contains previous discriminations. This results in the continued repetition of discrimination rather than resolving the problem. Another problem is the exacerbation in the competitive imbalance to gather information. The next era of technology has a high possibility of being monopolized by a few large corporations, since the company with the most customers can collect the most data, learn the best model with the data, and attract more customers again. This same problem exists among nations too. People, especially in countries in the Asia-Pacific region have large differences in their perspectives on the utilization of personal data. According to Asian security news outlet Security Brief, when customers were asked if they would like to provide personal information for a better consumer experience, the Japanese recorded the lowest approval rate at 43%, but the Chinese recorded the highest rate at 82%. This could increase a difference in technological developments between countries, companies and individuals. Another problem is with the operability and non-disclosure of the algorithm. As mentioned above, algorithms are subject to patent law protection, and most of the time they are not disclosed to the public. Thus, there is the possibility of manipulating them to derive desired results. At the same time, when an algorithm fails unintentionally, it is difficult to recognize the problem and is unclear who will be responsible for the failure.

Algorithms can also cause personal problems too, with one of them being addiction. You probably have experienced scrolling through social media for several hours. Even the CPO of YouTube Neal Mohan, admitted in a New York Times interview that 70% of YouTube viewing time is the result of recommended algorithms, and that total video viewing time has increased more than 20 times since the algorithm was introduced. Isolation from opposing views is also one of the major issues with algorithms. This phenomenon, called the filter bubble, narrows people’s view by making them believe the contents that they are exposed to are the opinions of the majority of people or even all of them. They don’t consider other points of view because they are not hearing them. Filter bubbles limit the type of content a person sees to previous interests, hindering the chance for people to explore something new. Some even complain that the repetitive content bores them quickly. Last but not least, algorithms have problems related to personal information infringement in the use of data. Data is usually collected through the opt-out method, which assumes that they had your consent to use the information unless you specifically refuse to give them permission individually. Most people skim through the user agreements, allowing companies to trace all their digital footprints and utilize personal information without them knowing how the company will be using their data.

It cannot be denied that algorithms will be needed to manage and circulate vast amounts of data in the future. However, it is crucial for companies to specify the detailed use of people’s data in advance when using algorithms and collecting data from them. Users should also acknowledge that an algorithm on their preferred website is only a suggestion tool, and that the results may not always be the optimum for what they want or need. Therefore, we users should aim to think critically and independently when viewing information instead of relying on the technology entirely to provide that information.

Choi Ga Yun  choia67@hotmail.com

<저작권자 © 홍익대영자신문사, 무단 전재 및 재배포 금지>