Publications

Peer-reviewed work, newest first. My name appears in bold.

CHI ’25 2025 · Yokohama, Japan

Promoting Cognitive Health in Elder Care with Large Language Model-Powered Socially Assistive Robots

Maria R. Lima, Amy O'Connell, Feiyang Zhou, Alethea Nagahara, Avni Hulyalkar, Anura Deshpande, Jesse Thomason, Ravi Vaidyanathan, Maja J. Matarić

Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems

My partDesigned and built the robot's full LLM dialogue system, supervised every field session in the retirement homes, and developed new methods for assessing task outcomes and analysing the data.

Read the paper DOI 10.1145/3706598.3713582

Abstract

Can a low-cost social robot promote cognitive health where older adults actually live? We built an LLM-powered system on Blossom, an affordable, home-assemblable robot with a knitted skin, and ran a five-week within-subjects field study with 22 adults aged 65 and over in retirement homes, in English and Spanish. The robot administered two clinically validated verbal tasks (picture description and semantic fluency); a human administered the baseline sessions. After the robot-run weeks, task performance improved, participants made more spontaneous social comments with the robot than with the human, and acceptance stayed high after the novelty wore off.

7 figures

Fig. 1 Blossom, the low-cost robot platform, with and without its knitted skin.

Fig. 2 A session in a retirement home.

Fig. 3 System: speech-to-text, an LLM generator with population-specific prompts, an independent LLM moderator, then text-to-speech synchronized with robot motion.

Fig. 4 Idle motions designed to keep the robot feeling present.

Fig. 5 Task performance improved after the robot-administered weeks.

Fig. 6 More spontaneous social comments with the robot than with the human administrator.

Fig. 7 Acceptance: fun, useful, low anxiety, trust rising over time, SUS 85.6 after novelty.

Collaboration: Imperial College London · UK Dementia Research Institute · University of Southern California.
- socially assistive robots
- LLM
- older adults
- field study
Show BibTeX
```
@inproceedings{lima2025promoting,
  title     = {Promoting Cognitive Health in Elder Care with Large Language Model-Powered Socially Assistive Robots},
  author    = {Maria R. Lima and Amy O'Connell and Feiyang Zhou and Alethea Nagahara and Avni Hulyalkar and Anura Deshpande and Jesse Thomason and Ravi Vaidyanathan and Maja J. Matarić},
  booktitle = {Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems},
  year      = {2025},
  address   = {Yokohama, Japan},
  doi       = {10.1145/3706598.3713582},
  url       = {https://doi.org/10.1145/3706598.3713582},
}
```

Publications

Promoting Cognitive Health in Elder Care with Large Language Model-Powered Socially Assistive Robots

Do not press

Stop the cat?