About

Daniel Kang is an Assistant Professor at the Siebel School of Computing and Data Science at the University of Illinois Urbana-Champaign. His research focuses on making analytics with machine learning (ML) accessible for scientists and analysts. His work has included a particular emphasis on video analytics and data systems for deploying ML. His research has been supported by organizations such as Google and the Open Philanthropy project. Daniel is actively recruiting PhD students for fall 2023 and can be contacted via email for those interested.

Research topics

• Medicine
• Artificial Intelligence
• Computer Science
• Chemistry
• Machine Learning
• Pharmacology
• Psychology
• Internal medicine
• Medical emergency
• Biology
• Simulation
• Neuroscience
• Psychiatry
• Physics

Selected publications

• Seven Symptoms Over Nearly 4,000 Days: Item-Level Variability in the Psychometric Properties of Daily Alcohol Use Disorder Symptoms in Young Adult Drinkers

  Assessment · 2026-01-12

  article

  People experience symptoms of alcohol use disorders (AUD) in their daily lives, including more impairment-based symptoms (e.g., hazardous use, interpersonal problems) and symptoms based on the intensity or frequency of consumption (e.g., cravings or limit violations). We estimated the psychometric properties of seven manifestations of daily AUD symptoms in a high-risk sample of regularly drinking young adults and identified the optimal operationalization and thresholds for those symptoms. We estimated item response theory (IRT) models in ecological momentary assessment data ( n = 527, age 18–22, 45% female) assessed over 3,963 alcohol use days. Symptoms were relatively common on drinking days, especially symptoms related to consumption (such as time spent drinking or consuming larger among than intended). The specific threshold or item used to define each AUD symptom could have a substantial impact on item parameters. Tolerance was best loaded onto a factor of daily AUD symptoms when operationalized as sensitivity to the effects of alcohol, while larger/longer was best reflected as drinking much more than intended (e.g., 3+ drinks). Daily life research focusing only on alcohol-related consequences misses important information about common experiences of AUD symptoms in daily life. Refinement of daily measures of AUD symptoms could help researchers understand how the disorder develops over time.

  PublisherDOI

• The Assessment of Alcohol Use Disorder Symptoms in Daily Life

  2025-07-14

  preprintOpen access1st authorCorresponding

  Understanding how alcohol use disorder (AUD) symptoms manifest in daily life is criticalfor improving assessment and intervention strategies. This study examined the correspondence between retrospective self-reports of AUD symptoms and their daily manifestations using ecological momentary assessment (EMA) in a community sample of young adults (N = 496, Mage=20.3, 45% female). Participants completed daily reports of AUD symptoms over an 8-week EMA period and follow-up retrospective assessments six months later. Results showed significant convergence between retrospective and daily reports for several symptoms (e.g., hazardous use, social/occupational problems, failure to fulfill obligations, time spent obtaining alcohol). Among these, some also demonstrated predictive validity, with daily experiences of symptoms (e.g., social/occupational problems, time spent obtaining alcohol) significantly predicting retrospective AUD severity at follow-up. These findings provide novel support for the convergent and predictive validity of daily assessments of AUD symptoms and highlight areas where retrospective and momentary reports diverge.

  PublisherDOI

• Exploring problematic alcohol use patterns in mood disorders through network analysis

  Journal of Substance Use · 2025-05-29

  article
  PublisherDOI

• When less is more: How attentiveness impacts the efficacy of online personalized feedback interventions for college student alcohol use

  Alcohol Clinical and Experimental Research · 2025-05-01

  article1st authorCorresponding

  BACKGROUND: Personalized feedback-based interventions for reducing high-risk alcohol use among college students vary in length and intensity. Comprehensive multicomponent personalized feedback interventions (PFIs) include more material and have greater intensity compared to briefer, single-component interventions such as personalized normative feedback (PNF). However, while PFIs may offer more comprehensive support, their lengthiness can potentially reduce attention and engagement with the intervention content, impacting their overall efficacy. This study examines how attentiveness-the degree to which participants engage with and process the intervention material-differs between single- and multicomponent interventions and how this variation moderates the efficacy of PFIs. METHODS: A secondary analysis of a longitudinal randomized clinical trial was conducted, involving 1137 undergraduates reporting past-month heavy episodic drinking (63% female; mean age = 20.1 years). Assessments occurred at baseline and 3, 6, and 12 months postintervention, with primary outcomes including drinks per week and negative alcohol-related consequences. Intervention conditions included (a) assessment-only control (AOC), (b) multicomponent PFI, and (c) single-component PNFs. Generalized linear mixed models were used to evaluate attentiveness as a moderator of treatment efficacy across multicomponent PFI, single-component PNFs, and AOC conditions. RESULTS: Analysis detected significantly higher attentiveness levels in single-component PNFs compared to the multicomponent PFI (b = 0.35, p < 0.001). A three-way interaction (Time × Condition × Attentiveness) indicated that the efficacy of multicomponent PFI versus AOC on drinks per week was only significant for those reporting moderate-to-high attentiveness levels, not for those with low attentiveness. When comparing multicomponent PFI to single-component PNF, multicomponent PFI outperformed single-component PNF only when attentiveness was high; conversely, when attentiveness was low, single-component PNF outperformed multicomponent PFI. CONCLUSIONS: While the simplicity of PNF allows for easy implementation with minimal cognitive effort, multicomponent PFI demonstrates greater efficacy potential, particularly when comprehended thoroughly. Future research could explore strategies to enhance attentiveness with multicomponent PFI, such as sequential delivery across multiple sessions to optimize its benefits.

  PublisherDOI

• Alcohol use disorder symptoms are kinda detectable in daily life (but it depends on how they are operationalized)

  2025-08-19 · 1 citations

  articleOpen access

  Contemporary theoretical models articulate how people experience symptoms of alcohol use disorders (AUD) in their daily lives, but few studies have modeled these symptoms at the daily level. We estimated Rasch and 2PL IRT models in two ecological momentary assessment samples of regularly drinking young adults (n = 527, age 18 – 22, 45% female) assessed over 3,963 alcohol use days. Symptoms were relatively common on drinking days, especially symptoms related to consumption (such as time spent or larger/longer). Items related to consumption (e.g. tolerance, larger/longer) were more common than alcohol-related consequences (e.g. interpersonal consequences). The specific threshold or item used to define each AUD symptom could have a substantial impact on item parameters. Tolerance best loaded onto a factor of daily AUD symptoms when operationalized as sensitivity to the effects of alcohol, while larger/longer was best reflected as drinking much more than intended (e.g. 3+). Daily life research focusing only on alcohol-related consequences misses important information about common experiences of AUD symptoms in daily life. Further refinement of daily measures of AUD symptoms could help researchers understand how the disorder develops over time.

  PublisherDOI

• Objective Assessment in Clinical Psychological Science: Progress in Wearable Alcohol Biosensors

  2025-10-24

  articleOpen access

  Clinical psychology is a discipline reliant on self-reports, but uniquely susceptible to specific biases associated therewith. Here we review progress in objective behavioral assessment in the domain of alcohol research, introducing an emerging transdermal class of wearable alcohol biosensor. We note challenges of transdermal assessment, together with recent performance gains from updated devices and analytic tools, including machine learning. We indicate unanswered questions for transdermal technology, including whether devices might ultimately produce fine-grained drinking quantity estimates and device longevity. We further identify factors that can impede development of new objective measures, including the tendency to judge new tools against an implicit ideal and consider scientific findings divorced from methodological details. Finally, in evaluating transdermal and other objective measurement tools, we argue for consideration of the specific error type (random vs systematic) generally linked with novel vs existing tools, identifying measurement diversification as a priority for clinical psychology moving forward.

  PublisherDOI

• Alcohol use disorder symptoms are kinda detectable in daily life (but it depends on how they are operationalized)

  2025-03-26

  preprintOpen access

  Contemporary theoretical models articulate how people experience symptoms of alcohol use disorders (AUD) in their daily lives, but few studies have modeled these symptoms at the daily level. We estimated Rasch and 2PL IRT models in two ecological momentary assessment samples of regularly drinking young adults (n = 527, age 18 – 22, 45% female) assessed over 3,963 alcohol use days. Symptoms were relatively common on drinking days, especially symptoms related to consumption (such as time spent or larger/longer). Items related to consumption (e.g. tolerance, larger/longer) were more common than alcohol-related consequences (e.g. interpersonal consequences). The specific threshold or item used to define each AUD symptom could have a substantial impact on item parameters. Tolerance best loaded onto a factor of daily AUD symptoms when operationalized as sensitivity to the effects of alcohol, while larger/longer was best reflected as drinking much more than intended (e.g. 3+). Daily life research focusing only on alcohol-related consequences misses important information about common experiences of AUD symptoms in daily life. Further refinement of daily measures of AUD symptoms could help researchers understand how the disorder develops over time.

  PublisherOA PDFDOI

• Social Drinking and Addiction: A Social-Cognitive Model for Understanding Alcohol Use Disorder Risk

  Current Directions in Psychological Science · 2025-04-06 · 7 citations

  articleOpen accessSenior author

  Scientists have long focused on intrapersonal factors and solitary drinking settings in researching addiction etiology. Yet evidence has accumulated to indicate a key role for social contexts in alcohol use disorder development. Here we review four core characteristics of social drinking contexts relevant for the understanding of disordered drinking, including prevalence, developmental timing, negative consequences, and reward value. We present a social-cognitive model aimed at elucidating reinforcement from alcohol in social context, proposing a role for alcohol in inhibiting higher-order cognitive processes that otherwise dampen the experience of social reward. Finally, we review a series of empirical studies providing evidence for the role of social context in alcohol use disorder development, highlighting methodological challenges and indicating directions for future research.

  PublisherOA PDFDOI

• Alcohol and emotion: Analyzing convergence between facially expressed and self-reported indices of emotion under alcohol intoxication.

  Psychology of Addictive Behaviors · 2025-01-09

  articleOpen access

  OBJECTIVE: Emotion measurement is central to capturing acute alcohol reinforcement and so to informing models of alcohol use disorder etiology. Yet our understanding of how alcohol impacts emotion as assessed across diverse response modalities remains incomplete. The present study leverages a social alcohol-administration paradigm to assess drinking-related emotions, aiming to elucidate impacts of intoxication on self-reported versus behaviorally expressed emotion. METHOD: = 22.5; 50% male; 55% White) attended two counterbalanced laboratory sessions, on one of which they were administered an alcoholic beverage (target blood alcohol content .08%) and on the other a nonalcoholic control beverage. Participants in both conditions were accurately informed of beverage contents and consumed study beverages in assigned groups of three while their behavior was videotaped. Emotion was assessed via self-report as well as continuous coding of facial muscle movements. RESULTS: = .021. Specifically, self-reports and behavioral displays converged among sober but not intoxicated participants. Further, alcohol's effects on positive facial displays remained significant in models controlling for self-reported positive and negative emotion, with alcohol enhancing Duchenne smiles 20% beyond effects captured via self-reports, pointing to unique effects of alcohol on behavioral indicators of positive emotion. CONCLUSIONS: Findings highlight effects of acute intoxication on the convergence and divergence of emotion measures, thus informing our understanding of measures for capturing emotions that are most proximal to drinking and thus most immediately reinforcing of alcohol consumption. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

  PublisherDOI

• Exploring associations between drinking contexts and alcohol consumption: An analysis of photographs.

  Journal of Psychopathology and Clinical Science · 2025-03-06 · 3 citations

  articleOpen access

  = 60). Participants wore transdermal alcohol biosensors during an ambulatory assessment period, while also taking photographs of their surroundings in response to random prompts. Computer vision methods were employed to extract contextual features from photographs. Results indicated numerous and often potent links between contextual features and patterns of consumption across SPAIS dimensions. Specifically, evening and weekend drinking, drinking during celebrations, drinking in bars, the presence of alcohol-related cues, distracting activity, and crowded, mixed-gender spaces were all associated with elevated levels of consumption. Results represent a step toward the identification of behavioral and structural change targets for alcohol use intervention programs, while at the same time providing new methods for capturing context in the field of addiction science. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

  PublisherDOI

Frequent coauthors

• Catharine E. Fairbairn

  University of Illinois Urbana-Champaign

  14 shared

• I. G. Rosen

  Temple University

  8 shared

• Susan E. Luczak

  Warsaw University of Technology

  8 shared

• Melike Sirlanci

  University of Colorado Boulder

  6 shared

• Konrad Bresin

  University of Louisville

  6 shared

• Yeonjae Kim

  Daejeon University

  4 shared

• Talia Ariss

  University of Illinois Urbana-Champaign

  3 shared

• Nigel Bosch

  University of Illinois Urbana-Champaign

  2 shared

Labs

• Siebel School of Computing and Data SciencePI

Education

• Ph.D., Computer Science

  University of Illinois at Urbana-Champaign

  2005

• M.S., Computer Science

  University of Illinois at Urbana-Champaign

  2001

• B.S., Computer Science

  University of Illinois at Urbana-Champaign

  1999

Awards & honors

• Google Junior Faculty Award