Innovation and User Engagement: A Case Study of TikTok (6/3/2023)

Abstract

This paper explores the impact of information communication technology (ICT) innovations, including human-computer interaction (HCI) theories, mobile-first design, and responsive web design on TikTok's interface. Analyzing the platform's design principles, focusing on the unique aspects of its machine learning algorithm and user psychology, establishes what sets TikTok apart from competing social media platforms. The paper also discusses TikTok's infrastructure architecture, emphasizing using content delivery networks (CDNs) and edge computing to optimize content distribution and enhance the user experience. The platform's future development direction is then hypothesized based on previous research. Similarly, this paper highlights the innovative Monolith system, which employs data science techniques to deliver personalized recommendations at scale. However, privacy and data protection concerns significantly threaten TikTok's future. A comprehensive analysis of TikTok's web design is necessary to fully understand the impact of data science, evolving infrastructure architecture, security concerns, and constant innovation on user experiences. TikTok serves as a valuable case study for examining the intricacies of social computing.

I. Introduction

TikTok has been a leader in using web design to revolutionize social computing. Integrating advancements in ICT and a mobile-first approach, TikTok emphasizes organization, simplicity, and readability. Through the mobile-first design method, TikTok has designed an interface— rooted in HCI theory— built for maximizing user engagement that has distinguished it from preceding forms of social computing. Moreover, TikTok capitalizes on advancements in ICT and human psychological factors that influence user engagement. Beyond integrating established technologies with psychology, TikTok's development of its Monolith system has furthered its innovative approach to user engagement on its platform. Furthermore, harnessing the power of data science by using machine learning in a novel way has brought TikTok to the forefront of its industry. However, despite these revolutionary applications of ICT and machine learning, like all social computing companies, TikTok faces concerns over security, including privacy and data protection, which urgently need to be addressed to secure the future of TikTok.

II. Exploring ICT Innovations and TikTok

This section provides a concise overview of crucial ICT innovations and their influence on web design for social computing, focusing on Human-Computer Interaction (HCI) theories, mobile-first approaches, responsive web design, and how TikTok exemplifies these.

Since the early 1990s, the rapid evolution of web design has developed in lockstep with technological advancements and changing user behaviors. Platforms like Facebook, WordPress, and YouTube have shaped web design by pushing for more dynamic and interactive experiences. Furthermore, introducing the iPhone in 2006 revolutionized social media and web design, leading to the adoption of responsive web design concepts in 2010 (Bonsor 2020). Responsive web design, with its fluid grids, flexible images, and media queries, ensures accessibility and optimization of web content across devices (Marcotte 2010). In addition, it enables a seamless user experience, catering to the increasing trend of mobile web usage and the need for adaptability (Wroblewski 2011).

Recently, the mobile-first approach, driven by the dominance of mobile devices in social media usage, has become a priority in web design (Johnson, 2019). For example, over 98% of Facebook users use a mobile device to access the site's content (Dixon 2023). This approach advocates for designing for mobile devices first and progressively enhancing them for larger screens to ensure optimal user experiences and accommodate device constraints (Johnson 2019). TikTok provides an example of this mobile-first design process in action.

Within the realm of mobile-first design, Activity Theory serves as a valuable framework for understanding the intricate connections between human activity, consciousness, and the design principles that prioritize mobile interactions, as it examines the interplay between individuals, their goals, the tools employed, and the social contexts in which these interactions occur. When evaluating social computing platforms like TikTok, Activity Theory becomes particularly helpful in analyzing the complexities of human activity and interaction within these digital environments, shedding light on user behavior, engagement patterns, and the impact of social factors on the overall user experience.

Interaction design, a practice within HCI, focuses on developing interactive digital goods, settings, systems, and services (Cooper et al. 2014). This practice strives to establish meaningful interactions between people and products, emphasizing human-centered user requirements, as TikTok exemplifies. The way users see and use interfaces is evaluative through usability or user experience (Murray 2011). HCI theories, such as Activity Theory, contribute to understanding human-computer interaction in the context of social computing platforms like TikTok, providing valuable insights into user behavior and engagement. Design teams utilize these theories to enhance user experiences and drive technology adoption, particularly in mobile short-form video apps like TikTok (Wang 2020). A good interface should not call attention to itself but should let us direct our attention to the task because the visual system is limited, with one portion of our visual input taken at a time (Murray 2011; Kim and Cave 1995). With its full-screen video playback and easy navigation, TikTok's interaction design achieves transparency, allowing users to focus on the app's content.

TikTok's interaction design, characterized by its full-screen video playback and easy navigation, exemplifies the principle of least astonishment by directing users' attention to the app's content and minimizing cognitive load. The principle of least astonishment argues that a component should act how most users anticipate it to behave (Seebach 2001). With that goal in mind, as Harvard Business Review highlights, prior experience(s) with competing products also influence expectations (Schwager and Meyer 2007). In addition, visual design, encompassing aesthetics and usability, plays a vital role in evaluation and engagement, highlighting the importance of well-crafted interfaces (Jongmans et al. 2022). For instance, with Tiktok, icons and pop-ups are limited to the most essential for easy usability while maintaining continuous video play. Correspondingly, the post, like, comment, and share icons used by Tiktok will all be familiar to users that have experience with other social computing platforms; the most notable similarities are Facebook, Instagram, and Youtube, making navigation easier for the user. Likewise, TikTok allows users to scroll up/down to switch between videos, with all this familiarity directing users' attention toward the app's content.

Furthermore, previous research has established that the essential design elements are navigation, graphical representation, organization, simplicity, and readability, regarding their impact on user engagement (Garett et al. 2016). For example, TikTok's design emphasizes organization, simplicity, and readability by providing a continuous watching environment by displaying only one comment pop-up while the full-screen video continues to play. After all, HCI is the region of intersection between psychology and the social sciences, on the one hand, and computer science and technology, on the other (Carroll 1997). Therefore, designers of these interfaces need to consider psychological factors influencing user engagement laboriously. For illustration, short-form content reduces cognitive load, and TikTok's machine-curated feed offers a variety of videos that stimulate the reward system and capitalize on the mere exposure effect (Sweller 1988; Biederman and Vessel 2006; Zajonc 1968). These design choices align with the psychological factors influencing user engagement.

TikTok's user engagement is in no small part thanks to the platform's differing approach compared to other social computing platforms because the "For You" page of TikTok uses machine learning to personalize each user's experience by deciding what video to show. The algorithmic approach is not new — it is just unique by removing the need for a user to follow other specific users or pages that align with their interests (Chow 2022). Instead, the app will quickly figure out what videos the user will like. The mechanism for this is a complex analysis of their reactions and a consideration of what content users who behave similarly have liked to decide what additional content they would enjoy (Smith 2021). This socio-interest graph, where the platform decides what to consume, differs from the social graph used by Facebook, Instagram, and Twitter (Chow 2022). Social media sites like Facebook and Instagram have historically employed social graphs to link users to a network of people they know and familiar friends (Newport 2022). In addition, balancing a user's autonomy in their interests and the desire for connection is arguably more aligned with one of the most influential motivation theories of human behavior, Self Determination Theory, than the other platforms in its design. TikTok's competitors have acknowledged the efficacy of this approach - Meta and Snap are increasingly incorporating socio-interest graphs into their algorithms (Chow 2022). Finally, efficient gesture interaction simplifies design and reduces the likelihood of errors during interaction (Lidwell, Holden, and Butler 2010). Minimizing user error is essential for the app's approach to be able to discern interest. These aspects of Tiktok's design are the foundation that provides such consistently high levels of user engagement as an app that is simple to learn, easy to navigate, well organized, and continuously stimulating to the brain's reward system.

ICT has profoundly impacted web design in social computing, as seen with a focus on Human-Computer Interaction (HCI) theories, mobile-first approaches, and responsive web design in the case study of TikTok. TikTok is a prime example of a mobile-first design process, where the interface delivers seamless user experiences on smaller screens while accommodating device constraints. In addition, TikTok's emphasis on design elements that are proven to impact user engagement allows the app to capitalize on conventions established by other social media platforms. Furthermore, this familiarity helps shepherd users' attention toward what differentiates TikTok from other social media platforms - its unique algorithm that applies a profound psychological understanding and machine learning to captivate users continuously.

III. Infrastructure Architecture: Enabling Scalability and Transformation

An essential aspect of TikTok is utilizing a content delivery network (CDN), a group of geographically distributed servers that speed up web content delivery by bringing it closer to where users are. The CDN platform balances, manages, and processes TikTok application traffic across all TikTok's edge clusters. At its simplest, edge computing clusters capture, process, and analyze data near the user to improve performance. Also, the CDN system contains varied network services to orchestrate the delivery of bits from their servers to users' phones. In addition, they operate an extensive network of POPs worldwide to accelerate site traffic and cache CDN content for TikTok edge computing. A point-of-presence (POP) is a point or physical location where two or more networks or communication devices build a connection from one place to the rest of the internet. TikTok is actively using new-generation technologies to drive innovation and build cloud-native content delivery networks and edge computing, ultimately serving TikTok's user base by minimizing load times for the user.

Incorporating CDNs and edge computing in TikTok's design is an example of using infrastructure advancements to optimize content distribution to benefit users. Researchers (Henfridsson and Bygstad 2013) explored how infrastructure architectures have evolved. They identified three key factors: adoption, innovation, and scaling. By studying these factors, they showed how changes in infrastructure technologies and systems had influenced the design process. In their study, (Helmond, Nieborg, and van der Vlist 2019) investigate how Facebook has transformed from a simple social networking site into a more complex "platform as infrastructure." Over a decade, they uncover a gradual process of growth and development, focusing on the interaction between becoming a platform and becoming an essential part of digital infrastructure. By doing so, they challenge the dominant perception of social media platforms and shed light on their evolving nature. While Henfridsson and Bygstad (2013) studied digital infrastructure more broadly, Helmond, Nieborg, and Van Der Vlist (2019) study focuses on how this process manifests itself in social computing. Although TikTok and Facebook are distinct platforms, the study by Helmond, Nieborg, and van der Vlist (2019) on the evolution of Facebook as a "platform as infrastructure" can offer some implications for TikTok. TikTok's rapid rise and dominance in the social media landscape suggest that it, too, has the potential to undergo a similar transformation. By understanding the interplay between platform and infrastructure, TikTok can strategically shape its development to become a more robust and influential platform, expanding its services and offerings beyond a conventional social networking site.

TikTok's infrastructure is critical for delivering a consistent user experience and high availability. It employs load balancing, redundancy, fault tolerance, and a content delivery network to achieve high availability and efficient content delivery for the user's benefit.

IV. Harnessing Monolith

Having established the importance of TikTok's algorithm to the user experience and the infrastructure that ensures its timely delivery, this section will dive deeper into what is unique about TikTok's algorithm. The influence of data science on web design in social computing is significant and plays a crucial role in improving user experiences. For example, TikTok's design inherently utilizes user behavior data analysis, personalization, A/B testing and optimization, and user feedback for continuous improvement.

TikTok's parent company ByteDance's Monolith system, is an example of the application of data science in web design to optimize user experiences and improve the effectiveness of social computing platforms. More specifically, this innovative development by ByteDance on behalf of TikTok is a unique methodology for achieving personalized content recommendations for users at scale with TikTok via machine learning. Real-time machine learning is the technique of continually improving a machine learning model by training it using live data. Accordingly, this system is tailored for online training and offers high fault tolerance and real-time learning. Online training for machine learning, as seen with TikTok, is to dynamically adapt to new patterns in user data. The Monolith's high fault tolerance is noteworthy for TikTok because it can help alleviate reliability, security, and other concerns associated with machine learning models conducting online training. With TikTok, what this means is that the model the company uses to decide which video to present on the "for you" page of any given user is constantly adapting to new user trends while simultaneously continuously improving itself with a high degree of accuracy in its recommendations.

This large-scale recommendation system can capture as many features as possible and adapt to concept drift (Liu et al., 2022). In machine learning, concept drift is when the real-world context changes, but the model does not recognize it. Adapting to concept drift is especially important for delivering a quality user experience in social computing, where context is dynamic and ever-changing. Monolith addresses challenges arising from sparse, categorical, and dynamic data from real-world user behavior (Liu et al., 2022). These elements are well-known challenges for data scientists applying machine learning to real-world applications. Due to these limitations, which occur with data, Monolith provides full expressive power for sparse features and a real-time feedback loop for training (Liu et al., 2022). This method helps cater the model to sparse data and provides feedback loops, critical elements of real-world machine learning systems. Additionally, feedback loops can improve or degrade the machine learning model's performance. In the case of TikTok, these feedback loops improve performance by adjusting parameters.

In order to organize data so that this model can be used to its fullest potential, it uses a data structure called a collisionless hash table that organizes and stores data in a way that minimizes the occurrence of two different pieces of data coexisting in a storage location, ensuring efficiency and reliability. By utilizing deep learning and innovative techniques like a collisionless hashtable, Monolith achieves improved model quality and reliability for online serving (Liu et al., 2022). In addition, TikTok utilizes these methods for predictive analytics to anticipate user preferences and behaviors. By analyzing historical data, TikTok's algorithms can predict users' interests, suggest relevant videos, and customize the user experience in real time. This personalization based on predictive analytics helps TikTok deliver a highly engaging and tailored user experience. Deep learning is a subset of machine learning that uses three or more layers to simulate the behavior of the human brain, allowing it to learn from large amounts of data (IBM 2023). In addition, hidden layers can help optimize and refine for accuracy (IBM 2023). Furthermore, the collisionless hash table achieves its namesake by using a specialized hashing function that guarantees unique hash codes for each item - this simplifies implementation and improves performance compared to traditional hashtables. Ultimately, these methods for improved model quality and reliability are central to delivering an excellent user experience for TikTok because they allow for astonishingly personalized content recommendations.

As demonstrated in TikTok's Monolith system, which combines revolutionary techniques such as collisionless hash tables and deep learning to give unique content suggestions for each user, data science continues to exert increasing influence on web design in social computing.

V. Unveiling Critical Security Concerns in TikTok's Web Design for Social Computing

Despite TikTok's massive success and innovative techniques for web design for social computing, there are critical concerns regarding privacy and data protection; some are novel concerns, while others have plagued its American social computing rivals over the past decade.

Before delving into the security concerns for TikTok, it is vital to note the distinction between TikTok and American-based social media platforms, underscored by the United States Department of Defense (DoD). (Vergun 2023). Chiefly among the concerns of the DOD is that China has historically used its cyber capabilities to steal sensitive information and engage in espionage (Vergun, 2023). According to the Department of Defense, TikTok is a "potential threat vector" to the United States due to its ownership by the Chinese company ByteDance (Vergun, 2023). This concern goes far beyond the perspectives on privacy, such as moral and communitarian views, shaping our understanding of privacy as a fundamental right or a balance between individual interests and the common good. Given its large user base and scale, the concern is that China may be able to direct misinformation and collect data through TikTok (Vergun, 2023). Especially given that algorithmic content recommendation poses privacy risks, as seen with the Congressional hearings for Facebook and Twitter but particularly in the case of TikTok, where user profiles and location data influence personalized video content. Finally, a history of noncompliance with data protection regulations, exemplified by the UK ICO's accusations against TikTok, emphasizes the need for transparent data handling and user consent. Addressing these challenges requires prioritizing privacy, data integrity, and ethical practices to ensure a safer and more secure social computing environment.

VI. Conclusion

In conclusion, TikTok's web design is a model of social computing that embodies the transformative power of ICT innovations, employing a mobile-first design philosophy and brilliant interaction design rooted in a robust application of HCI theories. TikTok leverages this knowledge by prioritizing design elements backed by empirical evidence to allow intuitive navigation for a captivating and stimulating user experience. Likewise, its infrastructure architecture enables scalability and paves the way for TikTok's future growth and continued transformation. By leveraging data science and the Monolith system, TikTok harnesses the power of personalized content recommendations with real-time learning, further optimizing user experiences and driving continuous improvement. Despite all the success of TikTok, it will be crucial to address security and privacy concerns in order for TikTok to sustain its success and continue forward as a leader in social computing.