Brooke A. Ammerman

Suicidal risk may be encoded in everyday communication patterns but diluted in routine digital interactions. We introduce a method for surfacing this latent signal: training per-person language model agents on individuals’ authored text (the on-screen text each participant typed, captured whenever a keyboard was visible in screenshots) and placing those agents in simulated social interactions—a “communicative stress test.” Using data from 79 adults with recent suicidal ideation, we fine-tuned individual LoRA adapters on Qwen3-8B using each participant’s authored text, then placed agents in standardized conversations with probe personas. Agent-generated risk language was associated with EMA-measured suicidal ideation ( r = .576, p < .001), with a single neutral small-talk probe performing nearly as well ( r = .551). A shuffle control confirmed the signal is person-specific ( r = .071 when adapters were mismatched), and automated descriptions of participants’ general smartphone activity produced no signal, confirming specificity to interpersonal communication. A prompt ablation demonstrated partial robustness to removal of disclosure-encouraging language ( r = .430). This proof-of-concept demonstrates that simulated social interaction can amplify latent vulnerability signals, bridging digital phenotyping, generative AI, and suicide theory.

Background: Passive smartphone sensing shows promise for suicide prevention, but behavioral metadata (GPS, screen time, and accelerometry) often lacks the contextual information needed to detect acute psychological distress. Analyzing what people actually see, read, and type on their phones—rather than just usage patterns—may provide more proximal signals of risk. Objective: This study aimed to test whether vision-language models (VLMs) applied to passively captured smartphone screenshots can predict momentary suicidal ideation (SI). Methods: Seventy-nine adults with past month suicidal thoughts or behaviors completed ecological momentary assessments (EMA) over 28 days while screenshots were captured every 5 seconds during active phone use. We fine-tuned open-source VLMs (Qwen2.5-VL [Alibaba Cloud], LFM2-VL [Liquid AI]), and text-only models (Qwen3 [Alibaba Cloud]) to predict SI from screenshots captured in the 2 hours preceding each EMA. We evaluated performance with temporal and subject holdouts. Results: The analytic sample comprised 2.5 million screenshots from 70 participants. Temporal holdout models achieved strong discrimination at the EMA level (AUC=0.83; AUPRC=0.77), with image-based models outperforming text-only models (AUC=0.83 vs 0.79; 95% CI 0.003-0.07). Subject holdout generalization was near chance (AUC≈0.50), though a simple lexical screening method retained modest discrimination (AUC=0.62). Smaller models performed comparably to larger models, supporting feasible on-device deployment. Conclusions: Screen content predicts short-term SI with clinically meaningful accuracy when models are personalized but does not generalize across individuals. These findings support a 2-stage clinical architecture, coarse lexical screening for new patients, with personalized VLM-based monitoring after a calibration period. On-device inference may enable privacy-preserving deployment.

Background: Self‐disclosure of suicidal thoughts and behaviors (STBs) is integral for risk assessment and intervention. However, limited research elucidates the nuanced characteristics of first disclosure experiences for bisexual individuals, who are disproportionately impacted by suicide yet remain underrepresented in the literature. This study examined the features, motivations, and outcomes of an initial STB disclosure among heterosexual and bisexual individuals. Methods: Self‐report data from 259 adults ( M _age = 35.40 years; 35.09% bisexual) with a history of STB disclosure recruited through Amazon Mechanical Turk (mTurk) were analyzed with univariate and non‐parametric tests and binary logistic regression models. Results: Compared to heterosexuals, bisexual participants reported a higher prevalence of suicidal behavior disclosure, seeking formal disclosure recipients, and disclosing through online platforms. They were also more motivated to address physical safety concerns and obtain professional help and more frequently engaged in help‐seeking behaviors post‐disclosure. On average, both groups rated disclosure as helpful, with heterosexual individuals reporting it as more helpful. There was a significant main effect of help‐seeking encouragement from recipients in predicting post‐disclosure help‐seeking engagement. Conclusions: These findings underscore the necessity of considering the impact of sexual orientation differences in initial STB disclosure processes, which may set the benchmark for subsequent disclosure and help‐seeking trajectories.

Background/Objectives: Computational studies using drift diffusion models on go/no-go escape tasks consistently show that individuals with suicidal ideation (SI) preferentially engage in active escape from negative emotional states. This study extends these findings by examining how individuals with SI update beliefs about action–outcome contingencies and uncertainty when trying to escape an aversive state. Methods: Undergraduate students with (n = 58) and without (n = 62) a lifetime history of SI made active (go) or passive (no-go) choices in response to stimuli to escape or avoid an unpleasant state in a laboratory-based negative reinforcement task. A Hierarchical Gaussian Filter (HGF) was used to estimate trial-by-trial trajectories of contingency and volatility beliefs, along with their uncertainties, prediction errors (precision-weighted), and dynamic learning rates, as well as fixed parameters at the person level. Bayesian mixed-effects models were used to examine the relationship between trial number, SI history, trial type, and all two-way interactions on HGF parameters. Results: We did not find an effect of SI history, trial type, or their interactions on perceived volatility of reward contingencies. At the trial level, however, participants with a history of SI developed progressively stronger contingency beliefs while simultaneously perceiving the environment as increasingly stable compared to those without SI experiences. Despite this rigidity, they maintained higher uncertainty during escape trials. Participants with an SI history had higher dynamic learning rates during escape trials compared to those without SI experiences. Conclusions: Individuals with an SI history showed a combination of cognitive inflexibility and hyper-reactivity to prediction errors in escape-related contexts. This combination may help explain difficulties in adapting to changing environments and in regulating responses to stress, both of which are relevant for suicide risk.

Background: Intensive longitudinal designs support temporally granular study of processes, making methods like ecological momentary assessment (EMA) increasingly common in medical and behavioral science. However, the repetitive and intensive measurement strategies associated with these designs increase participant burden, which limits the breadth and precision of EMA surveys. This is particularly problematic for complex clinical phenomena, such as suicide risk, which research has shown is multidimensional and fluctuates over narrow time intervals (eg, hours). To overcome this limitation, we proposed the Computerized Adaptive Test for Suicide Risk Pathways (CAT-SRP), which supports the simultaneous assessment of multiple empirically informed risk domains and facilitates personalized survey content. Objective: The objective of this study is to develop, calibrate, and pilot the first multidimensional computerized adaptive test for suicidal thoughts and related psychosocial risk factors in intensive longitudinal designs like EMA. Methods: A web-based assessment platform was developed to adaptively administer the CAT-SRP. CAT-SRP items were modified from existing validated instruments to support administration in intensive longitudinal designs. The item bank was developed in line with major ideation-to-action theories of suicide and consultation with experts outside the study team. Exploratory item factor analysis was used to identify dimensionality of the item bank. Item parameters were calibrated using a multidimensional graded response model in a large cross-sectional community sample (n=1759, 36.33% with a history of suicidal thoughts). Following calibration, the CAT-SRP was evaluated in an EMA study of participants with a past month history of suicidal thoughts (n=29 across 2134 observations). Adaptive testing used D-optimal item selection, a dual variable-length stopping criterion, and Maximum a Posteriori (MAP) scoring. Descriptive statistics and mixed effects models were used to examine CAT-SRP performance (eg, efficiency and survey overlap) and relationships among CAT-SRP domain scores. Results: The calibration study identified 2 suicidal thought domains (active and passive thoughts) and 12 risk factor domains: humiliation, loneliness, anger, pain, defeat, impulsivity (ie, negative urgency), entrapment, distress tolerance, perceived burdensomeness, thwarted belongingness, aggression, and a positively valenced method factor. Domain information was the highest between average to high levels of domain scores. Study 2 showed that the CAT-SRP (1) administered surveys with low to moderate item overlap, (2) incurred low participant burden, and (3) may improve near-term prediction of suicidal thoughts relative to traditional EMA measurement. Most EMA surveys reached the maximum length, 50 questions, highlighting a need to refine selection and stopping rules. Conclusions: The CAT-SRP effectively personalized EMA survey content to respondents, which reduces the repetitiveness and perceived burden of intensive longitudinal research designs. Continuous domain scores from multidimensional computerized adaptive testing (MCAT) also provided more nuanced measurement compared to traditional approaches that struggle with zero-inflation in EMA and appeared to produce stronger predictive relationships. Overall, the CAT-SRP demonstrated strong methodological advantages to use CAT for intensive longitudinal data collection.

Suicide rates are increasing rapidly among Black children and adolescents, calling for novel approaches to understanding their unique risk factors. The Structural Racism and Suicide Prevention Systems Framework offers a new culturally responsive theory that structural racism is an underlying mechanism for disparities in suicide among ethnoracial marginalized youth. Thus, a deeper analysis of the intersection of racism and systems to better understand suicide risk and create more effective targeted interventions for Black youth is imperative. The current systematic review comprehensively evaluated and synthesized the empirical literature regarding the relationship between structural racism and suicide risk among Black youth. 17 studies from 3 database searches, published between 2013 and 2024 are presented. Results revealed a positive relationship between structural racism and suicidal thoughts and behaviors among Black youth. Systems that particularly facilitate the perpetration of racism toward Black youth include schools, criminal justice, and income inequality. Findings serve as a call to action to incorporate more socioecological models into suicide prevention research focused on Black youth. Understanding the depth and scope of how racism contributes to suicide risk provides key targets for prevention and intervention strategies that are specific to individuals belonging to this group at disparate risk for suicide.