Dust Off Kindle Highlights With Quologue: Surfacing Personal Data With Generative AI for Reflective Experiences

Paper Title

Dust Off Kindle Highlights With Quologue: Surfacing Personal Data With Generative AI for Reflective Experiences

Publication Info

• Topic area: Human-Computer Interaction (HCI) and generative AI for personal data reflection.
• Keywords: e-books, annotations, generative AI, reflective experiences, personal data, slow technology, large language models, metadata, digital reading, self-expression.

Background and Problem

• Problem / challenge: E-book annotations, stored as digital metadata, often remain invisible and underutilized, limiting opportunities for self-reflection and engagement with personal reading histories. Current e-reading platforms restrict user interaction with this data, offering minimal support for creative or reflective practices.
• Significance: Revisiting personal annotations can foster self-reflection, identity construction, and deeper engagement with one’s past experiences. Addressing the invisibility and inaccessibility of e-book metadata could unlock its potential as a resource for meaning-making.
• Motivation and related work: Prior research highlights the reflective potential of personal metadata and the limitations of current e-reading systems in enabling meaningful interactions with annotations. HCI studies have explored slow technology and generative AI for fostering reflection, but applications for e-book metadata remain underexplored. This paper seeks to bridge this gap.

Solution

• Proposed approach: Quologue, a web application powered by a large language model (LLM), enables users to engage with their e-book highlights through stepwise interactions, using keywords and remixing to foster reflection and self-expression.
• Novelty:
  1. Introduces a dialogical model for interacting with e-book metadata using generative AI.
  2. Applies slow technology principles to foster long-term, reflective engagement with personal data.
  3. Demonstrates how minimal, ambiguous prompts (keywords) can catalyze creative and reflective experiences.
  4. Explores the influence of generative AI on users’ digital reading and annotation practices.
• Procedure and key techniques:
  • Users upload e-book highlights to Quologue.
  • Each week, the system selects a random highlight and generates three keywords (two from the highlight, one semantically related but unrelated to the text).
  • Users write a response based on the keywords, which is then synthesized with the original highlight to create a “remix.”
  • Users can review and compare remixes, track their interaction history, and optionally influence future keywords through an “interest of the week” feature.

Results

• Concrete findings:
  • Participants reported diverse reflective experiences, including recalling past memories, connecting with recent life events, and discovering new insights about themselves.
  • Several participants became more intentional and strategic in their e-book highlighting practices.
  • Generated remixes often resonated with participants, blending past and present contexts meaningfully, though some desired more control over the tone.
• Advantage over baselines:
  • Quologue transformed static e-book metadata into a dynamic medium for reflection and creative expression, unlike conventional e-reading platforms that treat annotations as static reference points.
  • The system’s stepwise, keyword-based interaction model encouraged deeper engagement with personal data over time.
• Experiments / evaluation:
  • An eight-week field study with 10 participants (aged 20–64) from North America.
  • Participants interacted with Quologue weekly, and researchers conducted three rounds of interviews to gather qualitative insights.
  • Data analysis revealed themes such as creative appropriation of keywords, reflective experiences, and changes in highlighting behaviors.
• Limitations and future work:
  • Limited participant pool (North America, self-identified avid e-book readers).
  • Some participants found the generated keywords or remixes misaligned with their expectations.
  • Future work could explore diverse cultural contexts, extend to other forms of personal textual data, and refine user control over remix tone and granularity.

Summary

This paper introduces Quologue, a generative AI-powered system that transforms e-book highlights into a medium for reflection and self-expression. By leveraging slow technology principles and a keyword-based interaction model, Quologue fosters long-term engagement with personal reading data. An eight-week field study revealed that participants experienced meaningful reflections, discovered creative uses for their metadata, and adapted their digital highlighting practices. The findings suggest opportunities for reimagining e-reading infrastructure to support lifelong, dialogical interactions with personal data. Future research could explore broader applications of this approach across diverse contexts and personal data types.