DESIGN RESEARCH - AI
Unlock your memories with AI
Treasurefinder is a device powered by a large language model (LLM) that generates open-ended questions based on stories stored in NFC-tagged physical objects or cards. The device is designed to explore a novel approach to see whether and how integrating AI, especially LLMs, into physically meaningful objects can support people’s reminiscing practices.
My Role
Research, Analysis, Product Design, Interaction Design
System Design, User-Testing
Tools
OpenAI API, Arduino, NFC, ESP32, Solidworks, PS, Illustrator
Featured in
Date
Feb 2023 – July 2023
Overview
Opportunity
Technology-mediated reminiscing has been explored for decades. However, generating relevant cues to trigger reminiscing remains challenging. LLMs have shown great potential in generating relevant content across various domains, but how LLMs can be used to facilitate reminiscing has not been explored in depth.
Goal
This work aims to provide explorative insights into the use of AI, specifically LLM (large language models), in supporting remembering practices on personal, significant objects at home.
Research
Challenges
Many experts in the field have dedicated their efforts to developing technology-enhanced artifacts to reminisce. Despite their efforts, people often have difficulty forming emotional bonds with these artifacts. This is because people value the physical traces and marks left on their objects, as these tangible cues serve as reminders of the past. As a result, integrating technology into meaningful objects remains challenging. Moreover, asking open-ended questions can improve memory recall. However, crafting relevant, open-ended questions is challenging for people as they rely on their interests or existing knowledge. AI, particularly LLMs, can help by generating contextually relevant questions as cues to facilitate memory recall.
Prototype
Ideation
Various ideas were explored to discover how to design an engaging user experience that could support users in recollecting memories associated with meaningful objects. These ideas included sketching out possible designs, creating low-fidelity prototypes to test the user experience, and crafting foam prototypes to assess the tactile experience of different designs.
Based on research, it was determined that the device had to meet several essential design requirements. The first one was to enable users to actively share stories related to the objects they have at home, stimulating interactions. Additionally, the device needed to provide a tangible and portable experience since most LLMs (like chatbots) are digital. Finally, the device had to be able to share stories about objects of any size, preserving the object’s identity.
Results
Treasurefinder Design
Treasurefinder is a compact device featuring an LLM to generate engaging open-ended questions. Activated by scanning an NFC-tagged object or inserting an NFC-tagged card, it plays voice-recorded stories previously captured by the object’s owners, facilitating user interaction with both the objects and their associated narratives.
An exploratory study was conducted with 12 participants in pairs to understand reminiscing behaviors with Treasurefinder. The findings showed that the AI-generated questions 1) supported individuals to recall the past, 2) fostered new insights about the other person, and 3) encouraged reflection. More interestingly, the device facilitated the active retrieval of memories associated with frequently overlooked objects they cherish.