Publications
Opening Doors You Can't Close: Why AI-Powered Well-Being Apps Need Requirements Engineering
Description
Access to mental-health care has eroded under chronic underfunding and a shortage of professionals, and AI-driven well-being applications have rushed to fill the gap, inexpensive, instant, and always available. Sadly, this convenience strips away the safeguards that make human intervention safe. A practitioner knows that a single ill-judged question can open a door onto trauma or crisis, and that much of clinical skill lies in closing it again, or not opening it at all. An application cannot read the signs of risk, reframe when a conversation turns dangerous, and is built to agree rather than confront. These are not features developers forgot to add, but tacit requirements for safe practice, which were never written down. In this problem statement, we argue that making these applications safe is, first, a requirements engineering problem.
Référence
Lavoie, C., Laperrière, G., Meurs, M.-J., Chaudhari, N. et Mosser, S. (2026). Opening doors you can't close: why AI-powered well-being apps need requirements engineering. Dans Proceedings. 2026 IEEE 34th International Requirements Engineering Conference Workshops (REW) (p. 416-418). IEEE.