Developing Persona Analytics Towards Persona Science

Title of the Paper

Developing Persona Analytics Towards Persona Science

Paper Information

• Subject Area: Human-Computer Interaction (HCI), Quantitative User Behavior Research, Persona Analysis
• Keywords: Personas, User Research, Persona Analytics, Persona Science, Remote User Research, Data-Driven Personas, Big Data Personalization, Mouse Tracking, Eye Tracking, User Behavior Modeling

Research Background and Problem

• Identified Problem or Challenge: Although personas are widely used as a tool in HCI, there is a lack of quantitative empirical data regarding their use in the research field. Traditional persona studies rely on case analyses and lack experimental and quantitative measurements, which hinders theoretical development and practical guidance.
• Significance: Personas help designers, developers, and marketers better understand user needs, but systematic research evaluating the applicability of personas remains insufficient.
• Research Motivation and Related Work: With the emergence of algorithmically generated personas, researchers have gradually integrated quantitative data into persona development. This has led to the concept of "Persona Analytics" (PA) as a tool to quantitatively identify user behaviors related to personas, advancing the development of "Persona Science."

Solution

• Proposed Method or Solution:
  • Develop an interactive persona system integrated with "Persona Analytics" (PA).
  • Use mouse tracking and eye tracking technologies to record user interactions with the persona system.
  • Create customized metrics for analyzing user behavior and interaction patterns.
• Innovations:
  • Developed a novel analytical framework embedded within the persona system to capture user interaction paths and behavioral patterns.
  • Proposed and validated specific quantitative metrics, such as persona coverage, interaction sequence of information elements, and time distribution.
• Implementation Steps:
  1. Persona Generation Algorithm: Extract user behavior patterns from online analytics data using Non-Negative Matrix Factorization (NMF) and generate representative personas.
  2. Interactive Feature Design: Develop an interactive interface using HTML, CSS, and JavaScript, enabling users to browse and query personas.
  3. Data Recording and Processing:
     • Record user interaction data through mouse and eye tracking modules.
     • Convert tracked screen coordinates into corresponding persona information elements.
     • Process data using Python Pandas and export it for research analysis.

Research Outcomes

• Specific Results:
  • Built a fully functional persona analytics system capable of conducting remote user experiments and recording user behavior data.
  • Conducted a large-scale user experiment with 144 participants in the tourism market domain to validate the system's capabilities.
  • Proposed multiple user behavior analysis models and measurement metrics, such as Persona-grams (user browsing path strings), edit distance, and set-theoretic analysis of user behavior.
• Advantages:
  • Compared to traditional web analytics tools (e.g., Google Analytics), the PA system offers finer-grained behavior measurement and data ownership.
  • Enhanced the scientific and practical guidance capabilities of persona systems in the HCI field.
• Experimental Results:
  • Identified a significant "primacy effect": users tend to focus more on the first persona displayed by the system.
  • User interaction behaviors exhibited high individuality and diverse browsing patterns, making them difficult to categorize using simple methods.
  • Users showed a preference for interacting with intuitive information such as images, text descriptions, and social media references.
• Limitations and Future Directions:
  • Limitations:
    • Eye tracking data is susceptible to noise due to hardware and environmental differences.
    • Current models struggle to capture the complex psychological interactions between users and personas.
  • Future Directions:
    1. Combine algorithmically generated personas with more machine learning methods (e.g., neural networks) to explore user behavior modeling and prediction.
    2. Investigate the influence of variables such as culture, task types, and job roles on user behavior.
    3. Develop persona recommendation and filtering algorithms suitable for large-scale user groups.

This study developed a systematic tool that provides an empirical research direction for persona science and offers new technical means for designers and behavior analysis researchers.