Did you know that there are personalization algorithms in the social media you use?

Image credit: AI，KELLEPICS

1.Introduction

In recent years, the rapid development of artificial intelligence has brought unexpected and earth-shaking changes to countless people’s lives in all aspects. I believe that many readers also feel that the use of artificial intelligence has had a profound impact on many fields, among which the actual progress of artificial intelligence is of great significance to the entire digital media field (Karnouskos, 2020). Among them, the personalized algorithms in major social media platforms are increasingly affecting the development of users, media content creators, merchants and the platforms themselves. The extensive use of this algorithm allows users to quickly and accurately match their needs with products and services, greatly reduces the cost of information dissemination and acquisition, and improves users’ consumption experience (People’s Daily, 2023). Rules, norms and regulatory frameworks for digital platforms are increasingly important, with huge political and social implications. Whether it’s watching news, watching videos, shopping online, or dining out, personalized recommendations on Internet platforms are increasingly affecting users’ daily lives. In order to achieve accurate and efficient personalized recommendations, the platform needs to use algorithms to collect and process users’ personal information to track and identify users. At the same time, there are also some platforms that abuse algorithms, causing problems such as “information cocoons”, “big data killing familiarity” and “inducing addiction”.

In this blog, I will use YouTube’s video recommendation algorithm, a mature leading platform, as an example to discuss the application of digitization, automation, artificial intelligence and algorithms on digital platforms more clearly, explore their benefits and problems, and discuss its development puts forward the prospect and thinks about the impact on users and society.

2. Digitization: the basis of personalization algorithms

In the process of personalized video recommendation, data is the foundation of everything. Data sources generally include content data and user activity data. The former is the metadata of the original video stream and video, and the latter is explicit and implicit. (Davidson et al. 2010). In other words, the titles, expressions, and descriptions in the videos we browse are content data, while the explicit data are users’ likes, favorites, comments, and ratings on the videos. Invisible activities are the data generated by the interaction between users and videos, such as the length of time and the frequency of viewing rates. Only with these databases can we proceed to the next step.

This kind of digitization occupies a major position in social media platforms. The existence of digitization allows videos and media content to appear on the user interface more accurately, and the user’s feeling and experience will be more comfortable so that users have more confidence in the platform. The viscosity increases. Just as Flew believes that Google’s search engine can more accurately display some relationships between search and purchase decisions than traditional advertisements. Because of these obvious relationships, advertisers can be correctly matched with demanders and potential users. Thus establishing a stronger and far-reaching relationship between users and the platform.

For digital platform companies, various data collected from digital platform users and further analysis mechanisms can bring them new forms of revenue (Flew, 2021). The more effective user data is, the more efficiently and accurately the platform and advertisers can deliver products.

3. Automation: Improve recommendation efficiency and accuracy

By automating the collected data, the processing and learning of the machine make the usefulness of the data greatly returned. In this process of accumulating data for search and matching, users’ behavioral experiences on social media platforms allow machines to automatically process them, allowing them to learn and understand, influence and shape user behavior on a large scale (Zuboff, 2019). For example, when a user “likes” a media content, after the data is collected, the machine automatically analyzes the user’s personality characteristics, personality type and content preferences.

Automation plays a key role in efficiency and precision in social media platforms. First, compared with the speed of manual identification, extraction, and classification, automated personalized algorithms can not only expand the scope of data but also automatically identify and analyze user behavior data. For example, for a huge platform like YouTube, many videos are uploaded every second, and the automated algorithm system needs to respond quickly enough to better balance the relationship between content and users (Covington, Adams & Sargin, 2016).

Secondly, manual intervention in traditional algorithms has a certain degree of subjectivity, while automated personalized algorithms can largely avoid human bias. In the research on the recommendation system of deep learning, Zhang et al. (2019) stated that the automatic learning system reduces manual interference and improves the quality while saving manpower. However, prejudice cannot be completely avoided, because the makers of the algorithm and related regulations and rules still have a certain influence. Automated algorithms can understand users from different dimensions and levels through a large amount of data learning and data feedback, so as to achieve more accurate push of media content. Artificial intelligence automates learning and makes technology fast and deep. Its development is very rapid, and users are influenced bit by bit in the process of using social media. According to Flew’s (2021) analysis of “behavior types” in digital platforms, new behavioral technologies of big data can influence and regulate users’ emotions and emotions on digital platforms. Many people have had this experience. Just chatting with friends about what they want to buy, they saw this product on the digital platform in a blink of an eye. Or maybe we watched food videos twice more, and in the next few days, the platform started to push food videos, and even started to push videos on how to cook food. Over time, various kitchen utensils and delicacies began to be pushed on the advertising page to lure us to buy.

Image credit: Application of AI in Social Media

4. Potential hazards of personalized algorithm recommendations

In digital platforms, personalization algorithms are important but also bring about developments beyond expectations. In addition to allowing users to communicate and communicate on the Internet in a variety of ways (Schultze and Whitt, 2016), it promotes a closer connection between users, platforms and merchants, allowing rapid matching of needs with products and services to achieve better benefits. However, while the personalized algorithm pushes on social media platforms expands human capabilities, it also brings many new problems and risks. Although personalized algorithms have improved the level of personal information services, they may also invisibly manipulate people, such as information cocoons, algorithm bias, and user privacy issues. Among them are filter bubbles that algorithms may generate, a danger that steadily increases (Krafft, Gamer & Zweig, 2019). In easy-to-understand words, the gradually formed information blind spots in our commonly used social media may cause individuals or groups to be informed of different facts. Because of the risks of these potential harms in algorithm recommendation, many people have begun to propose regulation of recommendation algorithms.

5. The Importance of Algorithmic Governance in Personalization Algorithms

In addition to the potential challenges and problems of personalized algorithms, there is also a very important issue that needs to be paid attention to is algorithm governance, because this history is related to the rights and interests of users and other stakeholders, and is also related to maintaining social justice. For example, behaviors like abusing algorithms violate consumers’ right to know, right to choose, and personal information rights, and also have a negative impact on the order of communication in cyberspace. Algorithmic content is highly specialized and technology updates are fast, which determines that algorithm governance is a long-term work that requires pooling of energy (Renming Daily, 2023). Since algorithms are developing so fast on digital platforms, the governance and supervision related to them are also facing huge challenges and tasks. Therefore, relying solely on the internal autonomy of the digital platform and the self-regulation of the enterprise is not enough to make stakeholders involved them and every participant feel completely safe in the environment, because of its unique scale and form, the formulation of external rules and regulations is also an issue that requires great attention (Flew, 2021). For example, the government needs to formulate effective policies and regulations, enterprises must have independent supervision measures, and groups at the social level participate in algorithmic governance and supervision. Algorithm regulation cannot only rely on the self-consciousness of the platform, but also requires the establishment of an all-around supervision system, and the concept and method of data governance are constantly innovated and improved.

Image credit: MS. TECH; IMAGES: SCREENGRABS FROM YOUTUBE

6. Case Study of YouTube’s Video Recommendation Algorithm

Nowadays, social media platforms are ubiquitous in the daily life of readers. YouTube is currently the largest video search and sharing platform in the world. All readers should have used this digital platform. Among several major companies, Google quickly targeted YouTube, acquired it and successfully allowed YouTube to occupy the position of the largest mobile phone in the field of social video (Flew, 2021). Based on YouTube on Google, a powerful search engine, it shows a different pattern. Combining previous data and relevant information to advance current transactions makes interactions more customized and personalized (Varian, 2010).

The implementation of recommendation algorithm for YouTube is mainly divided into data collection, recommendation generation and recommendation service (Davidson et al., 2010). Here artificial intelligence exerts its high efficiency. The data formed by users and media content combined with the automation of the algorithm system allows YouTube to provide personalized recommendations to users.

In detail, YouTube’s video recommendation algorithm is based on the user’s viewing, likes and favorites, and generates a video set through a video image with “co-access” (Davidson et al., 2010). The purpose of such an assembly line process is to allow users to find videos that belong to their preferences and to have a more sense of participation and entertainment. As Flew (2021) stated that this new surveillance capital model produces a “behavior type” that further refines personality assessment, and this type of behavior is very helpful for commercials and businesses. This also means that there is a need for a wide range of media content on YouTube. But the recommendation algorithm on YouTube will generate moderate filter bubbles for most of the set personas, and personalization has a greater impact on the YouTube homepage than others (Ledwich, Zaitsev & Laukemper, 2022). Therefore, although the operation of its algorithm meets the needs of users, it may also bring certain challenges to the diversity of information and the formation of public opinion. This kind of personalized recommendation algorithm not only appears in YouTube, but many social media platforms also make full use of personalized algorithms to carry out a new capitalist model.

7. Prospects for Algorithmic Governance Problem Solution Proposals

In the process of governance, for the challenges brought by artificial intelligence, dataization, automation and algorithms, such as automated decision-making, the invisibility of algorithmic systems and the opacity of operations, issues related to ethics and values, for all parties involved in this situation， there are responsibilities and measures (Gritsenko & Wood, 2022).

• Digital platforms strengthen self-governance and supervision, continuously optimize algorithm mechanisms, and expand information coverage to prevent users from falling into filter bubbles.

• Regulatory authorities have strengthened supervisory measures to crack down on violations of laws and regulations.

• Enterprises should also implement their own responsibilities to make data more secure.

Through the establishment of rules and regulations, self-supervision, and strengthening supervision in all aspects, this series of behaviors is expected to usher in a healthier and more sustainable development of algorithm applications and related industries.

Reference list

1. Covington, P., Adams, J., & Sargin, E. (2016). Deep Neural Networks for YouTube Recommendations. Proceedings of the 10th ACM Conference on Recommender Systems, 191–198. https://doi.org/10.1145/2959100.2959190
2. , J., Liebald, B., Liu, J., Nandy, P., Van Vleet, T., Gargi, U., Gupta, S., He, Y., Lambert, M., Livingston, B., & Sampath, D. (2010). The YouTube video recommendation system. Proceedings of the Fourth ACM Conference on Recommender Systems, 293–296. https://doi.org/10.1145/1864708.1864770
3. Flew, Terry (2021) Regulating Platforms. Cambridge: Polity, pp. 79-86.
4. Gritsenko, D., & Wood, M. (2022). Algorithmic governance: A modes of governance approach. Regulation & Governance, 16(1), 45–62. https://doi.org/10.1111/rego.12367
5. Huffman, S. (2008, September 15) Search evaluation at Google.[Blog post].Google official blog. https://googleblog.blogspot.com/2008/09/search-evaluation-at-google.html
6. Karnouskos, S. (2020). Artificial Intelligence in Digital Media: The Era of Deepfakes. IEEE Transactions on Technology and Society, 1(3), 138–147. https://doi.org/10.1109/TTS.2020.3001312
7. Kalpokas, I. (2019). Data: The Premise of New Governance. In Algorithmic Governance (pp. 11–25). Palgrave Macmillan Limited. https://doi.org/10.1007/978-3-030-31922-9_2
8. Krafft, T. D., Gamer, M., & Zweig, K. A. (2019).What did you see? A study to measure personalization in Google’s search engine. EPJ Data Science, 8(1), 1–23. https://doi.org/10.1140/epjds/s13688-019-0217-5
9. Ledwich, M. ., Zaitsev, A., & Laukemper, A. (2022). Radical bubbles on YouTube? Revisiting algorithmic extremism with personalised recommendations. First Monday, 27(12). https://doi.org/10.5210/fm.v27i12.12552
10. Varian, H. (2010). Computer-enabled transactions. American Economic Review, 100(1), 1–10.
11. Swart, J. (2021). Experiencing algorithms: How Young People Understand, Feel About, and Engage With Algorithmic News Selection on Social Media. Social Media + Society, 7(2), 205630512110088–. https://doi.org/10.1177/20563051211008828
12. Schultze, S. J., and Whitt, R. S. (2016). The interent as a complex layered system. In J. M. Bauer and M. Latzer (eds), Handbook on the Economics of the Internet (pp. 55–71). Cheltenham: Edward Elgar.
13. Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. New York: Public Affairs.
14. Zhang, S., Yao, L., Sun, A., & Tay, Y. (2019). Deep Learning Based Recommender System: A Survey and New Perspectives. ACM Computing Surveys, 52(1), 1–38. https://doi.org/10.1145/3285029