Application of Machine Learning: Personalization in UX Design; Definition and Frameworks of Personalized Systems

Lee, Minjae, School of Engineering and Applied Science, University of Virginia
Wayland, Kent, EN-Engineering and Society, University of Virginia
Vrugtman, Rosanne, EN-Comp Science Dept, University of Virginia
Graham, Daniel, EN-Comp Science Dept, University of Virginia
Laugelli, Benjamin, EN-Engineering and Society, University of Virginia

The modern User Experience (UX) industry is increasingly moving towards the use of personalized user interfaces and feeds. The major drivers of the tech industry all utilize an extensive form of personalization through the use of machine learning. These personalization systems utilize personal and behavioral data to provide a personalized experience without direct input from the user, presenting itself in forms like individualized products, services, recommendations, and pricing. This reliance on machine learning and data collection introduces issues related to data privacy and vulnerability, and the elevated convenience produces social side effects related to the reduction of user autonomy. However, because personalization is still a relatively new concept, there is a lack of public understanding on the implementation process and risk assessment of the technology. In order to address this problem, the technical and STS research works to better educate the public and future engineers on forms of assessment and implementation of personalization.

The technical research redesigns the current Human Computer Interaction (HCI) course at the University of Virginia to better prepare students and future engineers for the changing standards of UX design by compensating for machine learning as a part of the design process. The core concepts of the UX lifecycle from the original course structure is mostly retained. However, because the UX lifecycle is primarily designed to function around fixed attributes and data, it cannot fully accommodate for machine learning and its self-evolving nature. In order to accommodate for the volatility and malleability introduced by machine learning into the design process, an additional Material Lifecycle Thinking (MLT) design method is introduced along with a Machine Learning Lifecycle Canvas. The MLT method brings into perspective the various stakeholders attributes in context to an evolving machine learning system. The Canvas is divided into sections which correspond to the Machine Learning Lifecycle and integrates MLT within the UX design process during the contextual inquiry and analysis stages. This allows students to visually map the volatile attributes of the machine learning system and gain insight on how the final UX design prototype with the application of personalization should function. By applying the theoretical method and lecture content to the semester-long project which works to address a specific UX problem, students can learn the process of designing a personalized UX solution, the new standard for UX design.

The presence of machine learning in the front end is more significant than most individuals realize, and the tradeoffs between convenience and privacy risks can impact the lives of future generations. In order to combat these potential consequences, a standard definition and framework for personalization is established in the STS research to better educate the public on the potential risks of relying on the technology. Due to the multiple origins and context of personalization, the term has lacked a commonly accepted definition. Various definitions were compared and reviewed in accordance with the origins and current use of personalization as a term, resulting in the definition of personalization as an individualized machine learning solution to providing a product or service individualized to users based on collected and available user data. To apply this definition and establish a framework for assessing personalization, Vesanen’s Framework of Personalization and the Enhanced Value Net (EVN) were compared. The EVN was most in line with the proposed definition and allowed personalization systems to be analyzed vertically between the customers, company, and suppliers as well as horizontally between the company, competitors, and complementors. It was most in line to the modern application of personalization with machine learning, allowing for insights on potential side effects of user autonomy, price discrimination, data privacy, and data vulnerability.

The technical research provides a means for individuals with less experience in machine learning a method to practically incorporate and design a UX solution with personalization. The STS research provides a definition and framework for the analysis of personalization systems. In conjunction, the research offers both a means of design and analysis of personalization in its modern usage to increase understanding and awareness of the technology. The technical research supplements the STS research by providing a means of understanding how the system was designed and the considerations put into the process. The STS research supplements the technical by establishing a means to evaluate the finalized design and its potential impact on consumers and the entity utilizing the system amongst other stakeholders. Although the means are proposed in the research, the redesigned course structure and application of the MLT and Canvas should be tested further in order to assess its effectiveness as a design tool. Furthermore, the utilization of the EVN should also be assessed through a case study of an existing personalization system to evaluate its effectiveness in providing effective and useful insight.

BS (Bachelor of Science)
Machine Learning, Personalization, UX Design

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Rosanne Vrugtman
Technical Advisor: Daniel G. Graham

STS Advisor: Kent Wayland
STS Advisor: Benjamin Laugelli

All rights reserved (no additional license for public reuse)
Issued Date: