Interests

These are the interests that drive relevance scoring. Editing interests.md in the repo changes what the digest surfaces. No code changes required.

tocify research interests

This file drives relevance scoring. digest.py reads the ## keywords, ## narrative, and ## sections headings below. Edit freely. No code changes needed.

narrative

I am a psychological scientist studying suicidality, with a primary focus on pediatric and adolescent populations. My work centers on intensive longitudinal data (ILD), meaning ecological momentary assessment (EMA), digital phenotyping, passive sensing, wearables and biosensors. It also centers on the statistical and computational frameworks needed to model short-timescale risk processes from those data.

Methodologically I care most about machine learning and dynamical systems modeling. That covers two things. First, substantive applications to suicide risk assessment, prediction, and intervention. Second, technical developments in those method families in their own right, even when the application domain is not suicide. Examples of the second: time series modeling, idiographic and person-specific modeling, early warning signal detection, state-space and nonlinear dynamics methods, just-in-time adaptive interventions (JITAI) and micro-randomized trial design, and the calibration, validation, and generalizability of clinical prediction models.

I care about methodological rigor. External validation, replication, calibration, handling of class imbalance and rare events, measurement reliability, and honest reporting of predictive performance. I would rather see a well-validated null than an overfit headline.

keywords

suicide suicidal ideation suicidal behavior suicide attempt self-harm non-suicidal self-injury NSSI self-injurious crisis safety planning suicide prevention adolescent adolescence youth pediatric child teen ecological momentary assessment EMA experience sampling intensive longitudinal digital phenotyping passive sensing wearable biosensor actigraphy smartphone sensing real-time monitoring just-in-time adaptive intervention JITAI micro-randomized machine learning deep learning predictive model risk prediction prediction model classifier natural language processing large language model dynamical systems nonlinear dynamics early warning signal critical slowing state space time series idiographic person-specific network analysis multilevel model DSEM latent state calibration external validation generalizability class imbalance measurement burst mental health psychiatry depression mood

sections

Each item is assigned to exactly one section. IDs must match digest.py.

• suicide, Suicide & self-harm. Suicidal ideation, behavior, attempts, self-injury, crisis services, means safety, prevention and intervention trials, epidemiology and risk factors. Boost heavily for pediatric and adolescent samples, and for ILD or ML components.

• sensing, Intensive longitudinal & sensing. EMA and ESM, digital phenotyping, passive smartphone or wearable sensing, biosensors, actigraphy, measurement-burst designs, JITAI and micro-randomized trials, real-time risk monitoring. Any clinical population, mental health preferred.

• methods, ML & dynamical systems methods. Technical advances in machine learning, prediction modeling, dynamical systems, nonlinear time series, idiographic and person-specific modeling, early warning signals, network psychometrics, causal inference for longitudinal data, model calibration, validation and generalization. Include strong methods papers even when the application sits outside psychiatry, if the method plausibly transfers to ILD or risk prediction.

• adjacent, Adjacent mental health, genetics & neurobiology. Two kinds of thing land here. First, good psychiatry, clinical psychology, or child and adolescent mental health work that does not fit the sections above. Second, psychiatric genetics and neurobiology: GWAS and polygenic scores for psychiatric or suicide-related phenotypes, gene-environment interaction, epigenetics, neuroimaging of adolescent development, biomarkers, and mechanistic work on stress, reward, or impulsivity systems relevant to self-harm risk. Prefer human studies and large or consortium samples. Keep this section selective.