In this paper we map out several distinct ethical challenges that long-term care poses relative to the other forms of care (inpatient, domestic companionship, informal mental health consul- tations online, and others) and explore particular difficulties of memory, mental modeling, decline vs. sustained ability, and the sustenance of caregiving support. All of these, we argue, suggest that current deployment of LLMs, even with improving practices of benchmarking and attempts at preventing potentially harmful confabulations, face a daunting chasm of accountability if the dignity of care recipients is to be preserved.
@inproceedings{arnold26icres,
title={Dignity Over Alignment: Long-term Care and Human-AI Interaction},
author={Thomas Arnold and Matthias Scheutz},
year={2026},
booktitle={Proceedings of ICRES},
url={https://hrilab.tufts.edu/publications/arnold26icres.pdf}
}