JMIR Publications

ABSTRACT

Background:

Young adults in the United States experience high rates of suicidal thoughts and behaviors (STBs) while having low engagement with traditional care, highlighting the need for accessible and scalable digital tools. Digital safety planning interventions (SPIs) can expand access to support, but self-guided digital SPIs introduce critical decision points about what to deliver, when to intervene, and how users retain authority over system actions. Personalization, defined as system-initiated adaptation, and customization, defined as user-initiated control have been proposed as complementary mechanisms for making these decisions, yet how they should be coordinated to preserve autonomy and reduce burden in crisis contexts remains unclear.

Objective:

To develop a framework of decision points that specify what, when, how, and by whom (system vs. user) key safety-planning functions should be controlled in self-guided digital SPIs, and to clarify how personalization and customization jointly structure autonomy, burden, and safety in moments of acute distress.

Methods:

Young adults (ages 18–24) with lived experience of STBs were recruited from individuals who engaged with Mental Health America (MHA), then they participated in semi-structured individual interviews. Transcripts were analyzed using thematic analysis by three coders, and resulting themes were interpreted through the 5W1H framework as an organizing lens for identifying and structuring decision points.

Results:

Participants identified three classes of decision points that organize effective self-guided digital SPIs: (1) system-level decisions about what content and actions are automatically generated, including crisis resources, coping suggestions, outreach scripts, and affirmations; (2) user-level decisions about when and at what pace safety planning unfolds, including message timing, and response latency; and (3) shared interaction of user- and system-level decisions governing whether and how the system may initiate safety-relevant actions, such as contacting supports or escalating care. Across these domains, participants emphasized that automation should reduce cognitive burden while remaining explicitly subordinate to user consent and revisable control.

Conclusions:

Personalization and customization function as mechanisms for allocating decision authority within digital SPIs. Organizing these mechanisms around the 5W1H framework clarifies how adaptive automation can be coordinated with autonomy preservation across the key decision-point domains of what is delivered, when the system acts, and how users participate. This decision-point framework offers an actionable foundation for designing, evaluating, and optimizing self-guided suicide prevention technologies.