Memory Printer: Exploring Everyday Reminiscing by Combining Slow Design with Generative AI-based Image Creation

Paper Title

Memory Printer: Exploring Everyday Reminiscing by Combining Slow Design with Generative AI-based Image Creation

Publication Info

• Topic area: Human–AI interaction in memory reconstruction and reminiscing.
• Keywords: Generative AI, slow design, tangible interaction, memory reconstruction, human–AI interaction, reminiscing, agency, screen printing, algorithmic bias, data privacy.

Background and Problem

• Problem / challenge: Existing web-based generative AI tools for memory reconstruction undermine user agency, are emotionally disengaging, and lack predictability and control.
• Significance: Memory reconstruction is critical for personal and therapeutic contexts, such as reminiscing and alleviating cognitive diseases like Alzheimer’s. Effective tools can deepen emotional connections and preserve unrecorded memories.
• Motivation and related work: Prior studies explored AI-assisted memory tools, focusing on large language models (LLMs) and visual stimuli. However, these tools often prioritize efficiency over meaningful interaction, failing to address the reconstructive and reflective nature of memory. Tangible and slow design principles remain underexplored in this domain.

Solution

• Proposed approach: Memory Printer—a tangible interface inspired by screen printing that combines slow design principles with generative AI to support memory reconstruction.
• Novelty:
  1. Introduces deliberate slowness to create reflective space during memory reconstruction.
  2. Embodies interaction through physical tools (e.g., wooden scraper) to restore agency.
  3. Decomposes image generation into layered control for iterative negotiation with AI.
• Procedure and key techniques:
  • Users narrate memory impressions, generating a base image via AI.
  • A physical scraper controls image revelation, enabling gradual observation and comparison.
  • Layered redrawing allows users to modify specific regions incrementally.
  • A built-in Zink printer produces tangible photos, transforming digital interaction into physical artifacts.

Results

• Concrete findings:
  • Memory Printer enhanced users’ sense of control, emotional resonance, and creative exploration compared to web-based tools.
  • Tangible interaction and slow design principles facilitated deeper memory evocation and reflection.
  • Participants viewed printed images as meaningful souvenirs, compensating for unrecorded memories.
• Advantage over baselines:
  • Compared to web-based tools, Memory Printer provided richer agency cues, reduced cognitive load, and encouraged systematic memory reconstruction.
  • Participants reported stronger emotional engagement and immersion with the tangible interface.
• Experiments / evaluation:
  • 24 participants (ages 22–57) compared Memory Printer with a web-based GAI tool in a laboratory study.
  • Data collection included audio/video recordings, generated images, and thematic analysis of interviews.
  • Findings highlighted Memory Printer’s ability to support reflective and structured memory reconstruction.
• Limitations and future work:
  • Limited age and cultural diversity among participants; future studies should include older adults and broader demographics.
  • Ethical concerns remain regarding false memories, algorithmic bias, and data privacy.
  • Future research will explore local AI deployment, cultural diversity in training datasets, and proactive notifications for false memory risks.

Summary

Memory Printer combines slow design principles, tangible interaction, and generative AI to support memory reconstruction in emotionally sensitive contexts. Empirical evidence from a study with 24 participants demonstrates its ability to enhance user agency, emotional engagement, and reflective practices compared to web-based tools. The physical scraper and layered redrawing features enable gradual, systematic memory reconstruction, while printed photos serve as meaningful souvenirs. Despite its promise, challenges such as false memory formation, algorithmic bias, and privacy concerns highlight the need for responsible design in GAI applications. This work contributes grounded insights into human–AI interaction and positions Memory Printer as a design exemplar for future research in emotionally significant AI applications.