ARTEMIS (Adaptive Real-Time Emotional and Metacognitive Intelligence System): A Novel Framework for Passive Cognitive-Affective Profiling Through Naturalistic Smartphone Interaction Patterns

Mental health conditions affect over 1.1 billion people globally, with Sub-Saharan Africa experiencing a treatment gap exceeding 90% due to limited psychiatric infrastructure, cultural stigma, and shortage of trained clinicians. Current screening approaches require active participation, clinical expertise, and controlled environments; presenting barriers that are nearly insurmountable in resource-constrained settings. We present ARTEMIS (Adaptive Real-Time Emotional and Metacognitive Intelligence System), a groundbreaking passive monitoring framework that detects early-stage anxiety, depression, ADHD, and cognitive decline through naturalistic smartphone usage patterns without requiring specialized hardware, active user engagement, or clinical supervision. ARTEMIS introduces three novel contributions: (1) Temporal Cognitive Load Signatures (TCLS); a new mathematical formalism capturing attention fragmentation through app-switching entropy and task completion dynamics; (2) Affective Micropattern Recognition (AMR); detection of emotional state transitions through typing cadence variability, pressure dynamics, and pause-pattern irregularities invisible to conscious awareness; (3) Metacognitive Coherence Indices (MCI); quantification of executive function integrity through decision consistency, error-correction latency, and goal-directed behavior maintenance across extended temporal windows. In a prospective observational study with 847 participants across Kenya, Tanzania, and Uganda (ages 15-65), ARTEMIS achieved 91.3% sensitivity and 89.7% specificity for major depressive disorder, 88.4% sensitivity for generalized anxiety disorder, and 86.9% accuracy for ADHD screening; comparable to gold-standard clinical interviews while requiring zero active assessment time. The system operates entirely on-device, preserves privacy, functions offline, and runs on entry-level smartphones ($50-100 range), making it deployable across Africa's 495 million smartphone users. ARTEMIS represents a paradigm shift from active assessment to passive digital phenotyping, enabling continuous, stigma-free mental health monitoring at population scale. This work establishes smartphone interaction dynamics as a valid, reliable, and culturally-agnostic biomarker for cognitive-affective states with transformative implications for global mental health equity.

Keywords: Digital phenotyping, passive mental health screening, smartphone behavioral biometrics, cognitive load dynamics, African mental health, low-resource screening, computational psychiatry