Research map / Multi-agent systems and games / Human-AI and goal-conditioned
SPASCA: Social Presence and Support with Conversational Agent for Persons Living with Dementia
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents SPASCA, a conversational AI system with a customized LLaMA2-based dialogue model and a controllable talking-head avatar that gives social presence and support to persons living with dementia, with three prototypes compared qualitatively.
Problem
Persons living with dementia need constant interaction and emotional support, which is taxing for human caregivers; existing computer-assisted interventions (e.g. pre-recorded simulated presence therapy) do not support true dialogue or personalization.
Approach
Combines a dialogue model customized by prompting LLaMA2 on reminiscence-therapy conversations with a talking-head avatar that renders a familiar caregiver's identity with lip-sync and controllable emotional expression, and compares scripted/recorded versus generative configurations in prototypes.
Key results
- No quantitative results reported; the paper states evaluation with target users is in progress.
- Three prototypes built: text/audio chatbot (Prototype-1), scripted DM with recorded video (Prototype-2), and full DM with generated talking face (Prototype-3).
- Qualitative expert review is said to report pros and cons of configurations, but detailed findings are not given in the text.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace generated talking-face videos with retrieved pre-recorded clips (ablation) | SPASCA with DM-scripted · Prototype 2 | generation speed | Helped |
| replace generated talking-face videos with retrieved pre-recorded clips (ablation) | SPASCA with DM-scripted · Prototype 2 | visual fidelity | Helped |
Limitations
- No user evaluation with PLWD reported yet; evaluation is planned as future work.
- Ablated scripted dialogue restricts interaction to pre-defined options.
- Data from only ten sessions with one physiotherapist and a small set of face images.
Experimental setup
Datasets: DementiaBank MEAD in-house recorded therapy sessions
Baselines: DM-scripted TF-recorded Deep-Live-Cam LivePortrait
Builds on: simulated presence therapy reminiscence therapy LivePortrait
Closest papers by meaning
- SocialSim: Towards Socialized Simulation of Emotional Support Conversation (AAAI 2025 · similarity 0.82)
- MemoryBank: Enhancing Large Language Models with Long-Term Memory (AAAI 2024 · similarity 0.82)
- LifeEval: A Multimodal Benchmark for Assistive AI in Egocentric Daily Life Tasks (CVPR 2026 · similarity 0.81)
- MetaMind: Modeling Human Social Thoughts with Metacognitive Multi-Agent Systems (NeurIPS 2025 · similarity 0.81)
- Mem-PAL: Towards Memory-based Personalized Dialogue Assistants for Long-term User-Agent Interaction (AAAI 2026 · similarity 0.81)
- MimeQA: Towards Socially-Intelligent Nonverbal Foundation Models (NeurIPS 2025 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.