Background and motivation. Connection platforms are among the most consequential social infrastructure built in the past two decades. Their dominant design pattern, optimising for engagement, session frequency, and retention is structurally misaligned with the conditions under which good relationships form. A 2025 pre-registered meta-analysis found modest but consistent associations between heavy dating-app use and poorer wellbeing, body image, and self-esteem. The World Health Organization's 2025 Commission on Social Connection found that one in six people worldwide experiences loneliness, that the condition is most prevalent among young adults aged 18–29 (the heaviest users of connection platforms), and that social disconnection is associated with approximately 871,000 deaths per year. This paper argues that these public-health stakes warrant a fundamentally different design paradigm. Framework overview. The readiness-first framework proposes that connection technology should first establish whether a person is in a constructive state to pursue a relationship, and if not, what would help. It is organised around four constructs: Safety (vulnerability to coercive dynamics; capacity to hold boundaries), Self-Awareness (ability to observe relational patterns with self-compassion and without defensive distortion), Relational Capacity (communication, emotional regulation, and conflict-repair skills), and Readiness (an integrative probabilistic judgement derived from the other three). Constructs are operationalised through a structured self-report instrument and a Bayesian inference engine that represents each construct as a probability distribution rather than a point estimate, preserving and surfacing uncertainty explicitly. Ethical design specification. The paper specifies ethical commitments as falsifiable, auditable properties rather than values statements: no variable-ratio reward mechanics, no countdown timers or artificial scarcity, no engagement metrics as success measures, exit always accessible and non-manipulative, consent to sensitive processing opt-in by default, the triggering threshold for any protective hold never shown to the user. This operationalisation enables external audit and distinguishes the work from assertion-based "ethical AI" positioning. Research agenda. The paper provides a ten-phase, ten-year research roadmap sequenced so that downstream efficacy and population-scale claims are withheld until upstream validation exists. It presents a 100-item research question bank across ten themes (construct validity, predictive validity, safety and gating, ethics and autonomy, cross-cultural validity, Bayesian methods, module intervention, population health, AI governance, and longitudinal effects), each item written to be answerable as a master's thesis, doctoral study, or multi-site collaboration. A pilot framework specifies five settings: university wellbeing services, young-adult community cohorts, digital-wellbeing initiatives, research institutions, and public-health/NGO partners. Theoretical foundations. The framework synthesises Bowlby (1969/1982) and Ainsworth et al. (1978) on attachment; Porges (2011) on polyvagal theory and the physiological gating of social engagement; van der Kolk (2014) and Freyd (1996) on trauma and betrayal-trauma dynamics; Gottman and Levenson (1992) and Gottman (1994) on relationship quality and repair; and Neff (2003) on self-compassion and reflective self-awareness. It draws on the dark-patterns research programme (Mathur et al., 2019, 2021) and regulatory guidance (European Data Protection Board, 2022; Federal Trade Commission, 2022). Limitations. Stated at length and without hedging: no empirical validation; readiness is an unvalidated construct; instrument reliability is unknown; self-report is susceptible to social desirability and deliberate misrepresentation; gating may compound isolation in false-positive cases; cultural validity is unestablished; full reproducibility is constrained by proprietary parameters. These limitations are not rhetorical; several are currently unresolved and are the primary motivation for the proposed research programme. Intended audience. Researchers in digital wellbeing, human-computer interaction, social psychology, relationship science, public health, and AI ethics. University wellbeing practitioners and institutional research directors interested in pilot collaborations. Grant reviewers evaluating proposals for Horizon Europe Clusters 1 and 2, EU4Health, or Digital Europe. Policy actors responding to the 2025 WHO Commission on Social Connection mandate.
Vicky Verma (Tue,) studied this question.