Enhancing Dashboard Onboarding via LLMs

Information

• Publication Type: Bachelor Thesis
• Workgroup(s)/Project(s): not specified
• Date: January 2025
• Date (Start): February 2024
• Date (End): January 2025
• Matrikelnummer: 12028208
• First Supervisor: Vaishali Dhanoa

Abstract

Data visualization offers powerful insights that drive better decision-making, but its inherent complexity can often be challenging. To understand these visualizations, an effective onboarding process is essential, providing clear guidance on the purpose and functionality of visualizations for users of all experience levels. Traditional onboarding approaches, such as static tutorials, frequently fall short in addressing the unique needs of diverse users, leading to confusion and inefficiency. This thesis presents an innovative onboarding solution powered by Large Language Models (LLMs), designed to provide personalized, context-aware assistance that adapts dynamically to each user. Our approach harnesses the capabilities of prompt engineering, adaptive sequencing, and conversational interactions to create a dynamic, engaging onboarding experience. Key features include custom prompts to clarify complex visual elements, adaptive sequencing that responds to user behavior, and tailored narratives that adjust to users’ expertise levels. We implemented these features into a prototype system, powered by Llama 3.1 and ChatGPT 4o, to provide real-time, responsive assistance. By transforming dashboards into more approachable tools, this work makes data visualization accessible to a wider audience, ultimately enhancing the way users interact with and extract insights from complex data.

Additional Files and Images

Weblinks

No further information available.

BibTeX

@bachelorsthesis{holzer-edo,
  title =      "Enhancing Dashboard Onboarding via LLMs",
  author =     "Philipp Holzer",
  year =       "2025",
  abstract =   "Data visualization offers powerful insights that drive
               better decision-making, but its inherent complexity can
               often be challenging. To understand these visualizations, an
               effective onboarding process is essential, providing clear
               guidance on the purpose and functionality of visualizations
               for users of all experience levels. Traditional onboarding
               approaches, such as static tutorials, frequently fall short
               in addressing the unique needs of diverse users, leading to
               confusion and inefficiency. This thesis presents an
               innovative onboarding solution powered by Large Language
               Models (LLMs), designed to provide personalized,
               context-aware assistance that adapts dynamically to each
               user. Our approach harnesses the capabilities of prompt
               engineering, adaptive sequencing, and conversational
               interactions to create a dynamic, engaging onboarding
               experience. Key features include custom prompts to clarify
               complex visual elements, adaptive sequencing that responds
               to user behavior, and tailored narratives that adjust to
               users’ expertise levels. We implemented these features
               into a prototype system, powered by Llama 3.1 and ChatGPT
               4o, to provide real-time, responsive assistance. By
               transforming dashboards into more approachable tools, this
               work makes data visualization accessible to a wider
               audience, ultimately enhancing the way users interact with
               and extract insights from complex data. ",
  month =      jan,
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Research Unit of Computer Graphics, Institute of Visual
               Computing and Human-Centered Technology, Faculty of
               Informatics, TU Wien ",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2025/holzer-edo/",
}