Deco Framework Extends Physical Objects into AI Companions via Dual Embodiment
A team of researchers has introduced Deco, a mobile platform that converts tangible items such as plush toys into AI companions via a dual-embodiment approach. A preliminary study involving nine participants (N=9) revealed four key design principles: Faithful Identity, Calibrated Agency, Ambient Presence, and Reciprocal Memory. Deco employs multimodal Large Language Models (LLMs) alongside Augmented Reality to produce synchronized digital representations of users' physical companions. In a within-subjects experiment with 25 participants (N=25), Deco demonstrated a marked improvement over a baseline digital companion powered by a personalized LLM in terms of perceived companionship and emotional connection. This system enhances personal physical objects into interactive AI companions, effectively linking static items with dynamic digital agents.
Key facts
- Deco is a mobile system integrating multimodal LLMs and AR.
- Formative study (N=9) derived four design principles.
- Design principles: Faithful Identity, Calibrated Agency, Ambient Presence, Reciprocal Memory.
- Within-subjects study (N=25) showed Deco outperformed baseline.
- Baseline was a personalized LLM-empowered digital companion.
- Deco creates synchronized digital embodiments of physical companions.
- Aims to bridge emotional bond between users and physical objects.
- Published on arXiv with ID 2605.03882.
Entities
—