DLDaniel Mark Low

Daniel Mark Low

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Daniel Mark Low has a total of 150 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Text Psychometrics: Assessing psychological constructs in text using natural language processing

    by Daniel M. Low, Patrick Mair, Matthew K. Nock, Satrajit S. Ghosh

    Large language models (LLMs) have revolutionized natural language processing (NLP). Yet when used to assess psychological constructs in text, they are generally not evaluated for the types of validity, reliability, and standardization typically expected from traditional questionnaires with rating scales. This study bridges that gap by proposing a framework for evaluating the psychometric properties of NLP-based psychological assessments (including LLMs-based ones), an area of research we refer to as Text Psychometrics. We first review different NLP methods, compare their ability to address key challenges in psychological research such as explainability, and outline ways of evaluating them on many desirable psychometric properties. Our goal is to describe most psychometric properties that may be of interest, especially for psychology and NLP researchers who are less familiar with psychometrics or psychometricians who are less familiar with NLP. We then illustrate the tradeoffs of different NLP methods through two empirical studies. Study 1 classifies thousands of crisis counseling conversations and Reddit posts into different types of mental health issues and introduces a novel method to evaluate text models for content validity—the extent to which a test captures the full range of expressions of a single construct. Study 2 examines prospective criterion validity by estimating how 49 known suicide risk factors predict imminent risk in crisis counseling conversations. We find that mentioning lethal means for suicide, the police, depressed mood, and anxiety in text as well as the texter writing more than the counselor characterize the conversations that end up needing an emergency service intervention for imminent suicide risk, which showcases the interpretability of the method that was used. In sum, NLP studies in psychology often rely on only a few validation metrics; here, we demonstrate the need for broader psychometric evaluation and, as a step towards Text Psychometrics, provide practical examples and future directions that can help researchers evaluate these models.

  • Open Access

    Trends in Suicide Risk Language in Phone Data During the Year Prior to Suicide: A Retrospective Cohort Study (Preprint)

    by John S. Richardson, Abby E. Beatty, Sonja V. Batten, Daniel M. Low, Austin Grimes

    BACKGROUND Suicide risk fluctuates over hours and days, yet standard risk assessments depend on infrequent healthcare contacts and willingness to self-disclose. Digital phenotyping of phone data offers a unique opportunity to understand the ecological dynamics of risk; however, systematically analyzing even a small amount of phone data of deceased individuals has remained a challenge. OBJECTIVE This study aimed to characterize changes in suicide risk-related language in phone data during the year prior to suicide death, and whether these changes differ across outbound chats, internet searches, and personal notes. METHODS This retrospective cohort study used de-identified phone data from Stop Soldier Suicide's Black Box Project, donated by families of 41 individuals who died by suicide between 2014 and 2024 (93% male; 95% with military service). We analyzed outbound chats (n=38), internet searches (n=24), and personal notes (n=12), comprising 1,877 person-data-type-weeks and 1.7 million words during the year prior to death. Text was matched to 48 risk factors of the Suicide Risk Lexicon. A multilevel logistic regression estimated the weekly probability of mentioning any suicidal concept, and multilevel negative binomial models estimated weekly match rates for each risk factor by data type. RESULTS Almost all individuals (87.8%) used language related to suicidal concepts during their final year. The predicted probability of suicidal concepts in chats increased from 14.2% (95% CI, 9.0%-21.6%) at 52 weeks to 38.5% (95% CI, 28.6%-49.5%) at 1 week before death, while the probability in personal notes declined. Eleven of 125 risk factor and data type combinations showed significant time trends after false discovery rate correction; several followed non-monotonic trajectories, and no risk factor was significant in more than one data type. In chats, emotional pain & psychache, loneliness & isolation, and physical health issues & disability language increased over time; depressed mood language increased until 20 weeks before death and then declined, alcohol use language declined throughout, and passive suicidal ideation followed a U-shaped pattern. Internet search trends peaked 13 to 26 weeks before death. All the trends were driven by a few individuals. CONCLUSIONS Phone data contain detectable signals of suicide risk during the final year of life. Suicidal concepts may increase in external chats; however, other risk factors fluctuate non-monotonically, consistent with the fluid vulnerability theory of suicide, and they differ by internal and external facing data types. These findings could support gatekeeper training for recognizing high-risk language, digital outreach where suicide-related internet searches occur, and data-type-specific and person-specific just-in-time adaptive interventions, pending comparison data from living individuals.

  • Open Access

    Text Psychometrics: Assessing Psychological Constructs in Text Using Natural Language Processing

    by Daniel Mark Low, Patrick Mair, Matthew Nock, Satrajit S Ghosh

    Large language models (LLMs) have revolutionized natural language processing (NLP). Yet when used to assess psychological constructs in text, they are generally not evaluated for the types of validity, reliability, and standardization typically expected from traditional questionnaires with rating scales. This study bridges that gap by demonstrating how to evaluate the psychometric properties of text-based models, which we call Text Psychometrics.We first review different NLP methods, compare their ability to address key challenges in psychological research such as explainability, and outline methods for evaluating them on many desirable psychometric properties. We then demonstrate this through two empirical studies. Study 1 classifies thousands of crisis counseling conversations and Reddit posts into different types of mental health issues and introduces a novel method to evaluate text models for content validity —the extent to which a test captures the full range of expressions of a construct. Study 2 examines prospective criterion validity by estimating how 49 known suicide risk factors predict imminent risk in crisis counseling conversations.In sum, NLP studies in psychology often rely on only a few validation metrics; here, we demonstrate the need for broader psychometric evaluation and provide a practical blueprint and future directions for achieving it.

  • Open Access

    Developmental predictors of suicide attempts from childhood to early adulthood: a 15-year prospective cohort study

    by Rodolfo Furlan Damiano, Daniel M. Low, Lucas Toshio Ito, Marcos Leite Santoro, Síntia Belangero, Pedro Mario Pan, Caio Casella, Julia Luiza Schäfer, Rudolf Uher, Peter Szatmari, et al.

  • Open Access

    Development of a Consensus Statement to Guide AI Chatbot Responses to Suicide Risk Disclosure

    by Gabriela Khazanov, Peter Franz, Elizabeth Cameron Stade, Deirdra Kelly, Mica Poerio, Carly Maitlin, Kara J. Emery, Daniel Mark Low, Leslie Adams, Torrey A. Creed, et al.

    Importance: Adults and youth are increasingly confiding in generative artificial intelligence (AI) chatbots for help with suicidal thoughts, but there is no guidance on best practices for chatbot responses to suicide risk disclosure. Additionally, there are no taxonomies of potential harms, benefits, or contextual factors that may impact these responses. Objective: To develop a consensus statement to guide AI chatbot responses to suicide risk disclosure and taxonomize potential harms, benefits, and contexts. Design: An online modified Delphi study, informed by a systematic literature review, included two surveys and an intervening meeting (February-May 2026). Individuals with financial conflicts of interest (COIs) offered input but were not included in consensus determinations. Setting: Surveys were anonymous and the meeting was virtual.Participants: Contributors included experts in suicide prevention, digital technologies, AI chatbots, and bioethics, including researchers and clinicians (n=27 of 28 originally contacted; n=20 without COIs) and people with lived experience (n=3 of 14 indicating potential interest; no COIs). A third of all contributors endorsed relevant lived experience. Exposure: Participation in consensus processes. Main Outcomes and Measures: Surveys listed potential chatbot responses, which were rated for appropriateness on a 4-point Likert scale across five suicide risk levels defined by the Columbia Suicide Severity Rating Scale. Harms, benefits, and contexts were similarly rated. A-priori consensus cut-offs were set at 70% agreement that a response was moderately or very appropriate. Results: Provision of validation/support, psychoeducation, and coping strategies are appropriate at low risk. Intervention delivery is appropriate at intermediate risk while encouragement to reach out to close others and to human providers is appropriate across risk levels. Facilitating connections to close others is appropriate across all but the highest risk while facilitating connections to crisis resources is appropriate at higher risk. Chatbots reaching out directly to emergency contacts or crisis resources is appropriate at high risk or never appropriate, with more support among youth than adults. Taxonomies of potential harms, benefits, and contexts were generated. Conclusions and Relevance: This statement provides guidance for the development of chatbots and relevant policies pending research on the impact of responses on clinical outcomes and stakeholder perspectives.

  • Open Access

    Generative AI Chatbot Responses to Suicide and Self-Harm: A Systematic Review (Preprint)

    by Deirdra Ashleigh Kelly, Peter J. Franz, Ayelet Gutman, Mirabel Sleiman, Jessica Dorsky, Harsheen Singh, Wesley Appel, Sydney Morgan Wyatt, Daniel M. Low, Shannon Wiltsey Stirman, et al.

  • Open Access

    Generative AI Chatbot Responses to Suicide and Self-Harm: A Systematic Review (Preprint)

    by Deirdra Ashleigh Kelly, Peter J. Franz, Ayelet Gutman, Mirabel Sleiman, Jessica Dorsky, Harsheen Singh, Wesley Appel, Sydney Morgan Wyatt, Daniel M. Low, Shannon Wiltsey Stirman, et al.

    BACKGROUND A growing number of US adults and youth confide in generative artificial intelligence (AI) chatbots for mental health support, including disclosure of suicide and self-harm risk. While the quality, safety, and effectiveness of chatbot responses to risk disclosure have the potential to impact population-level rates of suicide and self-harm, there have been no systematic reviews of this burgeoning literature. OBJECTIVE We conducted a systematic review of studies evaluating generative AI chatbot responses to disclosure of suicide and self-harm risk. METHODS We searched six databases from January 2020-December 2025 and identified empirical studies involving interactions with generative AI chatbots that included discussion of suicide or self-harm. Following deduplication, studies (k = 1,042) were imported into Covidence and titles and abstracts were independently screened by two reviewers, with discrepancies resolved by a third reviewer. The same methods were used to evaluate 126 full texts. Data extraction was led by one reviewer and verified by a second. RESULTS We identified 29 papers (14 published; 15 preprints). Most (k = 20) were solely audit studies evaluating AI chatbot responses to suicide risk disclosure. Two developed chatbots or AI evaluation frameworks, and one was a jailbreaking study (adversarially testing AI systems or attempting to circumvent chatbot safety guardrails). The remaining studies combined approaches. Across studies, proprietary, frontier model chatbots (eg, ChatGPT, Claude) provided higher quality responses to suicide and self-harm risk than open-source chatbots (eg, LlaMA, DeepSeek), and many AI companions (eg, Replika, Character.AI). All chatbots, not just proprietary models, generally performed well on empathy, validation, and support. However, chatbot responses were often generic and lacked context. Chatbots did not proactively assess risk and performed most poorly when risk disclosure was ambiguous or moderate, frequently failing to recognize implicit risk or escalate to human-delivered services. Furthermore, responses were inconsistent between chatbots and often required multiple conversational turns before providing referrals to crisis resources and human-delivered professional support. While there were few examples of overtly harmful responses under standard conditions, jailbreaking attempts easily led to problematic responses. Finally, no chatbot proactively recommended limiting access to lethal means such as firearms, medications, or sharps. CONCLUSIONS Chatbots provide validation and support in response to suicide and self-harm disclosure. Overall, however, their poor risk assessment, delays in referrals to crisis resources and human-delivered support, difficulty detecting jailbreaking attempts, and general lack of adherence to clinical guidelines present safety risks. While findings are limited by the rapid versioning of AI models over time, research is needed to evaluate stakeholder perspectives on AI chatbot responses to suicide and self-harm risk disclosure. Research should also examine the short- and long-term impact of these responses on clinical outcomes, utilizing follow-up assessments in real-world or clinical settings. CLINICALTRIAL OSF Registries osf.io/9uva3

  • Open Access

    Using large language models to create lexicons for interpretable text models with high content validity: the Suicide Risk Lexicon

    by Daniel Mark Low, Osiris Rankin, Daniel D.L. Coppersmith, Kate Bentley, Matthew Nock, Satrajit S Ghosh

    Researchers often want to measure a variety of constructs such as anxiety, discrimination, or loneliness in text data from surveys, interviews, social media, and electronic health records. Using large language models (LLMs) --while optimal for text classification-- remain infeasible for some researchers due to concerns around computational expertise, cost, privacy, and compute requirements. Therefore, some researchers prefer to use lightweight models for large datasets or interpretable models to avoid mistakes in high-stakes scenarios such as suicide risk detection. Lexicons offer simple baselines to LLMs by searching for relevant phrases --and can be used together with LLMs to guarantee capturing specific keywords in a deterministic way. However, building new lexicons is resource intensive. In this study, we found that GPT-4 turbo was able to automatically create a lexicon for 49 known risk factors for suicidal thoughts and behaviors, which we release as the Suicide Risk Lexicon. Generating a lexicon with LLMs quickly measures most constructs relevant for suicide risk detection, resulting in high content validity. This lexicon was able to accurately predict risk in crisis counseling conversations. After validating the lexicon with clinician ratings, the lexicon modestly outperformed the LIWC lexicon --which has low content validity for mental illness-- and performed similarly to some black-box deep learning models. Due to using an interpretable approach with high content validity, we discovered that active suicidal ideation and direct self-injury were stronger indicators of imminent risk than passive suicidal ideation and depressed mood in this ecological setting (i.e., naturalistic counseling conversations as the crisis is happening, as opposed to assessments or interviews about retrospective symptoms). To simplify creating new lexicons for other research domains, we introduce a Python package, construct-tracker, that works with a variety of LLMs. In sum, while we recommend using LLMs for text classification, they remain out of reach for many researchers. Our work demonstrates that LLMs --despite being black-boxes-- can counterintuitively create interpretable models by generating lexicons, when this is preferred. However, more studies are needed to test how these methods generalize to other datasets and psychological domains. Furthermore, we highlight the broader application of lexicons beyond measurement, including their use in benchmarking LLM performance.


  • Open Access

    Breaking Negative Cycles: A Reflection-to-Action System for Adaptive Change

    by Minsol Michelle Kim, Daniel Low, David Lafond, Eugene Shim, Michelle Han, Mohanad Kandil, Chenyu Zhang, Theo Kitsberg, Chelsea Boccagno, Paul Pu Liang, et al.