HAI Postdoc Fellow
Stanford University, Institute for Human-Centered Artificial Intelligence (HAI), CA, USA
I am a postdoc fellow at Stanford University's Institute for Human-Centered Artificial Intelligence (HAI). I completed my PhD in Computer Science at Dartmouth College where I worked under the supervision of Prof. Andrew T. Campbell in the amazing HealthX lab (prev: DartNets lab). My research interests span the field of ubiquitous computing and applied machine learning, particularly focussing on the use of passive and mobile sensing in conjunction with AI to assess health and wellbeing of individuals. I'm also interested in understanding and modeling human behavior based on passive sensing data (such as those collected from mobile phones, wearables and other ubiquitous technologies).
I am currently on the job market (for both academic and industry openings) -- please reach out if you have any pointers! Here's my Google Scholar for reference. My CV (long) is available here and here's my résumé (short). You can access my research statement/agenda from here.
Stanford University, Institute for Human-Centered Artificial Intelligence (HAI), CA, USA
Dartmouth College, NH, USA
Microsoft Research, Cambridge, MA
Microsoft Research, Redmond, WA
Remote Internship
Deerwalk Institute of Technology, Kathmandu, Nepal
TechLekh Services Pvt. Ltd., Kathmandu, Nepal
Ph.D. in Computer Science
Dartmouth College, USA
Bachelors in Computer Science & Information Technology
Tribhuwan University, Nepal
CIE A Level
Trinity International College, Nepal
Affiliated to Cambridge University, UK
Several of my work has been published in top-tier journals and conferences. Below listed are some of the selected publications. Please check my Google Scholar for recent updates.
As concerns about employee burnout and skilled staff shortages in cybersecurity grow, our study aims to better understand the contributing factors to burnout in this field. Utilizing a mixed-methods approach, we analyze self-reported job and personal characteristics, along with digital activity data from 35 incident responders, identifying several factors such as high workload, time pressure, and lack of support from management. Our findings reveal that over half of the participants experience burnout (N=19), which is linked to increased workload, limited control, poor teamwork, and inadequate recognition. Burned-out responders often work more than 40 hours per week, have poor sleep quality, and engage in more email activities, meetings, and after-hour collaborations. Through our research, we also identify coping strategies individuals use to mitigate these stressors. Based on our findings, we provide practical recommendations to help organizations better support their cybersecurity incident response teams. While our study acknowledges limitations and suggests future research directions, it contributes significantly to understanding the challenges faced by cybersecurity incident responders. Our insights offer a comprehensive understanding of burnout factors in this domain and have broader implications for other high-stress work environments consistent with the interdisciplinary nature of CSCW.
Social isolation is a common problem faced by individuals with serious mental illness (SMI), and current intervention approaches have limited effectiveness. This paper presents a blended intervention approach, called mobile Social Interaction Therapy by Exposure (mSITE), to address social isolation in individuals with serious mental illness. The approach combines brief in-person cognitive-behavioral therapy (CBT) with context-triggered mobile CBT interventions that are personalized using mobile sensing data. Our approach targets social behavior and is the first context-aware intervention for improving social outcomes in serious mental illness.
Understanding the dynamics of mental health among undergraduate students across the college years is of critical importance, particularly during a global pandemic. In our study, we track two cohorts of first-year students at Dartmouth College for four years, both on and off campus, creating the longest longitudinal mobile sensing study to date. Using passive sensor data, surveys, and interviews, we capture changing behaviors before, during, and after the COVID-19 pandemic subsides. Our findings reveal the pandemic's impact on students' mental health, gender based behavioral differences, impact of changing living conditions and evidence of persistent behavioral patterns as the pandemic subsides. We observe that while some behaviors return to normal, others remain elevated. Tracking over 200 undergraduate students from high school to graduation, our study provides invaluable insights into changing behaviors, resilience and mental health in college life. Conducting a long-term study with frequent phone OS updates poses significant challenges for mobile sensing apps, data completeness and compliance. Our results offer new insights for Human-Computer Interaction researchers, educators and administrators regarding college life pressures. We also detail the public release of the de-identified College Experience Study dataset used in this paper and discuss a number of open research questions that could be studied using the public dataset.
MindScape aims to study the benefits of integrating time series behavioral patterns (e.g., conversational engagement, sleep, location) with Large Language Models (LLMs) to create a new form of contextual AI journaling, promoting self-reflection and well-being. We argue that integrating behavioral sensing in LLMs will likely lead to a new frontier in AI. In this Late-Breaking Work paper, we discuss the MindScape contextual journal App design that uses LLMs and behavioral sensing to generate contextual and personalized journaling prompts crafted to encourage self-reflection and emotional development. We also discuss the MindScape study of college students based on a preliminary user study and our upcoming study to assess the effectiveness of contextual AI journaling in promoting better well-being on college campuses. MindScape represents a new application class that embeds behavioral intelligence in AI.
As emerging technologies increasingly integrate into all facets of our lives, the workplace stands at the forefront of potential transformative changes. A notable development in this realm is the advent of passive sensing technology, designed to enhance both cognitive and physical capabilities by monitoring human behavior. This paper reviews current research on the application of passive sensing technology in the workplace, focusing on its impact on employee wellbeing and productivity. Additionally, we explore unresolved issues and outline prospective pathways for the incorporation of passive sensing in future workplaces.
MoodCapture presents a novel approach that assesses depression based on images automatically captured from the front-facing camera of smartphones as people go about their daily lives. We collect over 125,000 photos in the wild from N=177 participants diagnosed with major depressive disorder for 90 days. Images are captured naturalistically while participants respond to the PHQ-8 depression survey question: "I have felt down, depressed, or hopeless"". Our analysis explores important image attributes, such as angle, dominant colors, location, objects, and lighting. We show that a random forest trained with face landmarks can classify samples as depressed or non-depressed and predict raw PHQ-8 scores effectively. Our post-hoc analysis provides several insights through an ablation study, feature importance analysis, and bias assessment. Importantly, we evaluate user concerns about using MoodCapture to detect depression based on sharing photos, providing critical insights into privacy concerns that inform the future design of in-the-wild image-based mental health assessment tools.
Speech-based diaries from mobile phones can capture paralinguistic patterns that help detect mental illness symptoms such as suicidal ideation. However, previous studies have primarily evaluated machine learning models on a single dataset, making their performance unknown under distribution shifts. In this paper, we investigate the generalizability of speech-based suicidal ideation detection using mobile phones through cross-dataset experiments using four datasets with N=786 individuals experiencing major depressive disorder, auditory verbal hallucinations, persecutory thoughts, and students with suicidal thoughts. Our results show that machine and deep learning methods generalize poorly in many cases. Thus, we evaluate unsupervised domain adaptation (UDA) and semi-supervised domain adaptation (SSDA) to mitigate performance decreases owing to distribution shifts. While SSDA approaches showed superior performance, they are often ineffective, requiring large target datasets with limited labels for adversarial and contrastive training. Therefore, we propose sinusoidal similarity sub-sampling (S3), a method that selects optimal source subsets for the target domain by computing pair-wise scores using sinusoids. Compared to prior approaches, S3 does not use labeled target data or transform features. Fine-tuning using S3 improves the cross-dataset performance of deep models across the datasets, thus having implications in ubiquitous technology, mental health, and machine learning.
Mobile phone sensing is increasingly being used in clinical research studies to assess a variety of mental health conditions (e.g., depression, psychosis). However, in-the-wild speech analysis -- beyond conversation detecting -- is a missing component of these mobile sensing platforms and studies. We augment an existing mobile sensing platform with a daily voice diary to assess and predict the severity of auditory verbal hallucinations (i.e., hearing sounds or voices in the absence of any speaker), a condition that affects people with and without psychiatric or neurological diagnoses. We collect 4809 audio diaries from N=384 subjects over a one-month-long study period. We investigate the performance of various deep-learning architectures using different combinations of sensor behavioral streams (e.g., voice, sleep, mobility, phone usage, etc.) and show the discriminative power of solely using audio recordings of speech as well as automatically generated transcripts of the recordings; specifically, our deep learning model achieves a weighted f-1 score of 0.78 solely from daily voice diaries. Our results surprisingly indicate that a simple periodic voice diary combined with deep learning is sufficient enough of a signal to assess complex psychiatric symptoms (e.g., auditory verbal hallucinations) collected from people in the wild as they go about their daily lives.
Rare life events significantly impact mental health, and their detection in behavioral studies is a crucial step towards health-based interven- tions. We envision that mobile sensing data can be used to detect these anomalies. However, the human-centered nature of the problem, com- bined with the infrequency and uniqueness of these events makes it challenging for unsuper- vised machine learning methods. In this pa- per, we first investigate granger-causality be- tween life events and human behavior using sensing data. Next, we propose a multi-task framework with an unsupervised autoencoder to capture irregular behavior, and an auxil- iary sequence predictor that identifies transi- tions in workplace performance to contextu- alize events. We perform experiments using data from a mobile sensing study comprising N=126 information workers from multiple in- dustries, spanning 10106 days with 198 rare events (< 2%). Through personalized inference, we detect the exact day of a rare event with an F1 of 0.34, demonstrating that our method out- performs several baselines. Finally, we discuss the implications of our work from the context of real-world deployment.
Anhedonia and amotivation are symptoms of many different mental health disorders that are frequently associated with functional disability, but it is not clear whether the same processes contribute to motivational impairments across disorders. This study focused on one possible factor, the willingness to exert cognitive effort, referred to as cognitive effort–cost decision making. We examined performance on the deck choice task as a measure of cognitive effort–cost decision making, in which people choose to complete an easy task for a small monetary reward or a harder task for larger rewards, in 5 groups: healthy control (n = 80), schizophrenia/schizoaffective disorder (n = 50), bipolar disorder with psychosis (n = 58), current major depression (n = 60), and past major depression (n = 51). We examined cognitive effort–cost decision making in relation to clinician and self-reported motivation symptoms, working memory and cognitive control performance, and life function measured by ecological momentary assessment and passive sensing.
Regularity in daily activities has been linked to positive well-being outcomes, but previous studies have mostly focused on clinical populations and traditional daily activities such as sleep and exercise. This research extends prior work by examining the regularity of both self-reported and digital activities of 49 information workers in a 4-week naturalistic study. The data was analyzed in conjunction with self-reported stress, anxiety, and depression scores. The findings suggest that greater variability in self-reported mood, job demands, lunch time, and sleep quality may be associated with increased stress, anxiety, and depression. However, for digital activity-based measures, we find that the opposite is true---greater variability in digital rhythm is related with reduced emotional distress. This study expands our understanding of workers and the potential insights that can be gained from analyzing technology interactions and well-being.
There is a growing body of research revealing that longitudinal passive sensing data from smartphones and wearable devices can capture daily behavior signals for human behavior modeling, such as depression detection. Most prior studies build and evaluate machine learning models using data collected from a single population. However, to ensure that a behavior model can work for a larger group of users, its generalizability needs to be verified on multiple datasets from different populations. We present the first work evaluating cross-dataset generalizability of longitudinal behavior models, using depression detection as an application. We collect multiple longitudinal passive mobile sensing datasets with over 500 users from two institutes over a two-year span, leading to four institute-year datasets. Using the datasets, we closely re-implement and evaluated nine prior depression detection algorithms. Our experiment reveals the lack of model generalizability of these methods. We also implement eight recently popular domain generalization algorithms from the machine learning community. Our results indicate that these methods also do not generalize well on our datasets, with barely any advantage over the naive baseline of guessing the majority. We then present two new algorithms with better generalizability. Our new algorithm, Reorder, significantly and consistently outperforms existing methods on most cross-dataset generalization setups. However, the overall advantage is incremental and still has great room for improvement. Our analysis reveals that the individual differences (both within and between populations) may play the most important role in the cross-dataset generalization challenge. Finally, we provide an open-source benchmark platform GLOBEM- short for Generalization of Longitudinal BEhavior Modeling - to consolidate all 19 algorithms. GLOBEM can support researchers in using, developing, and evaluating different longitudinal behavior modeling methods. We call for researchers' attention to model generalizability evaluation for future longitudinal human behavior modeling studies.
The transition from high school to college is a taxing time for young adults. New students arriving on campus navigate a myriad of challenges centered around adapting to new living situations, financial needs, academic pressures and social demands. First-year students need to gain new skills and strategies to cope with these new demands in order to make good decisions, ease their transition to independent living and ultimately succeed. In general, first-generation students are less prepared when they enter college in comparison to non-first-generation students. This presents additional challenges for first-generation students to overcome and be successful during their college years. We study first-year students through the lens of mobile phone sensing across their first year at college, including all academic terms and breaks. We collect longitudinal mobile sensing data for N=180 first-year college students, where 27 of the students are first-generation, representing 15% of the study cohort and representative of the number of first-generation students admitted each year at the study institution, Dartmouth College. We discuss risk factors, behavioral patterns and mental health of first-generation and non-first-generation students. We propose a deep learning model that accurately predicts the mental health of first-generation students by taking into account important distinguishing behavioral factors of first-generation students. Our study, which uses the StudentLife app, offers data-informed insights that could be used to identify struggling students and provide new forms of phone-based interventions with the goal of keeping students on track.
The COVID-19 pandemic continues to affect the daily life of college students, impacting their social life, education, stress levels and overall mental well-being. We study and assess behavioral changes of N=180 undergraduate college students one year prior to the pandemic as a baseline and then during the first year of the pandemic using mobile phone sensing and behavioral inference. We observe that certain groups of students experience the pandemic very differently. Furthermore, we explore the association of self-reported COVID-19 concern with students’ behavior and mental health. We find that heightened COVID-19 concern is correlated with increased depression, anxiety and stress. We evaluate the performance of different deep learning models to classify student COVID-19 concerns with an AUROC and F1 score of 0.70 and 0.71, respectively. Our study spans a two-year period and provides a number of important insights into the life of college students during this period.
People with serious mental illness (SMI) have significant unmet mental health needs. Development and testing of digital interventions that can alleviate the suffering of people with SMI is a public health priority. The aim of this study is to conduct a fully remote randomized waitlist-controlled trial of CORE, a smartphone intervention that comprises daily exercises designed to promote reassessment of dysfunctional beliefs in multiple domains.
We hypothesize that behavioral patterns of people are reflected in how they interact with their mobile devices and that continuous sensor data passively collected from their phones and wearables can infer their job performance. Specifically, we study day-today job performance (improvement, no change, decline) of N=298 information workers using mobile sensing data and offer data-driven insights into what data patterns may lead to a high-performing day. Through analyzing workers' mobile sensing data, we predict their performance on a handful of job performance questionnaires with an F-1 score of 75%. In addition, through numerical analysis of the model, we get insights into how individuals must change their behavior so that the model predicts improvements in their job performance. For instance, one worker may benefit if they put their phone down and reduce their screen time, while another worker may benefit from getting more sleep.
Pandemics significantly impact human daily life. People throughout the world adhere to safety protocols (e.g., social distancing and self-quarantining). As a result, they willingly keep distance from workplace, friends and even family. In such circumstances, in-person social interactions may be substituted with virtual ones via online channels, such as, Instagram and Snapchat. To get insights into this phenomenon, we study a group of undergraduate students before and after the start of COVID-19 pandemic. Specifically, we track N=102 undergraduate students on a small college campus prior to the pandemic using mobile sensing from phones and assign semantic labels to each location they visit on campus where they study, socialize and live. By leveraging their colocation network at these various semantically labeled places on campus, we find that colocations at certain places that possibly proxy higher in-person social interactions (e.g., dormitories, gyms and Greek houses) show significant predictive capability in identifying the individuals’ change in social media usage during the pandemic period. We show that we can predict student’s change in social media usage during COVID-19 with an F1 score of 0.73 purely from the in-person colocation data generated prior to the pandemic.
Commuting to and from work presents daily stressors for most workers. It is typically demanding in terms of time and cost, and can impact people’s mental health, job performance, and, broadly speaking, personal life. We use mobile phones and wearable sensing to capture location-related context, physiology, and behavioral patterns of N=275 information workers while they commute, mainly by driving, between home and work locations spread across the United States for a one-year period. We assess the impact of commuting on participant’s workplace performance, showing that we can predict self-reported workplace performance metrics based on passively collected mobile-sensing features captured during commute periods.
The ubiquity of smartphones and wearables makes it an attractive option to passively study human behavior. We explore the current practices of using passive sensing devices to assess mental health and wellbeing, including the limitations and future directions.
In the initial lockdown phase of the COVID-19 pandemic, people spent more time on their phones, were more sedentary, visited fewer locations, and exhibited increased symptoms of anxiety and depression. As the pandemic persisted through the spring, people continued to exhibit very similar changes in both mental health and behaviors. Although these large-scale shifts in mental health and behaviors are unsurprising, understanding them is critical in disrupting the negative consequences to mental health during the ongoing pandemic.
Most people desire promotions in the workplace. Typically, rising through the ranks comes with increased demands, better salary and higher status among peers. However, promoted workers have to deal with new challenges, such as, adjusting to new roles and responsibilities, which can in turn impact their physical and mental wellbeing. In this year long study, we use mobile sensing to track physiological and behavioral patterns of N=141 information workers who are promoted. We show that the workers experience a change in their physiological and behavioral patterns after promotion captured by passive sensing from phones, wearables and Bluetooth beacons. Furthermore, we use a random convolutions based approach to extract patterns from multivariate time series signals and evaluate the performance of different models to classify a worker’s mobile sensing data as belonging to a promoted or non-promoted period with an AUC of 0.72. As a result, we report for the first time that mobile sensing can detect job promotion events by modeling physiological and behavioral changes of information workers in an objective manner.
Assessing performance in the workplace typically relies on subjective evaluations, such as, peer ratings, supervisor ratings and self assessments, which are manual, burdensome and potentially biased. We use objective mobile sensing data from phones, wearables and beacons to study workplace performance and offer new insights into behavioral patterns that distinguish higher and lower performers when considering roles in companies (i.e., supervisors and non-supervisors) and different types of companies (i.e., high tech and consultancy). We present initial results from an ongoing year-long study of N=554 information workers collected over a period ranging from 2-8.5 months. We train a gradient boosting classifier that can classify workers as higher or lower performers with AUROC of 0.83. Our work opens the way to new forms of passive objective assessment and feedback to workers to potentially provide week by week or quarter by quarter guidance in the workplace.
Several psychologists posit that performance is not only a function of personality but also of situational contexts, such as day-level activities. Yet in practice, since only personality assessments are used to infer job performance, they provide a limited perspective by ignoring activity. However, multi-modal sensing has the potential to characterize these daily activities. This paper illustrates how empirically measured activity data complements traditional effects of personality to explain a worker’s performance. We leverage sensors in commodity devices to quantify the activity context of 603 information workers. By applying classical clustering methods on this multisensor data, we take a person-centered approach to describe workers in terms of both personality and activity. We encapsulate both these facets into an analytical framework that we call organizational personas. On interpreting these organizational personas we find empirical evidence to support that, independent of a worker’s personality, their activity is associated with job performance. While the effects of personality are consistent with the literature, we find that the activity is equally effective in explaining organizational citizenship behavior and is less but significantly effective for task proficiency and deviant behaviors. Specifically, personas that exhibit a daily-activity pattern with fewer location visits, batched phone-use, shorter desk-sessions and longer sleep duration, tend to perform better on all three performance metrics. Organizational personas are a descriptive framework to identify the testable hypotheses that can disentangle the role of malleable aspects like activity in determining the performance of a worker population.
The vast majority of people worldwide have been impacted by coronavirus disease (COVID-19). In addition to the millions of individuals who have been infected with the disease, billions of individuals have been asked or required by local and national governments to change their behavioral patterns. Previous research on epidemics or traumatic events suggests that this can lead to profound behavioral and mental health changes; however, researchers are rarely able to track these changes with frequent, near-real-time sampling or compare their findings to previous years of data for the same individuals. By combining mobile phone sensing and self-reported mental health data among college students who have been participating in a longitudinal study for the past 2 years, we sought to answer two overarching questions. First, have the behaviors and mental health of the participants changed in response to the COVID-19 pandemic compared to previous time periods? Second, are these behavior and mental health changes associated with the relative news coverage of COVID-19 in the US media?
Over the past few years, an incredible diversity of consumer-grade wearables has emerged with tremendous breadth in capabilities, form factor, and cost. These wearable devices show significant promise for researchers to conduct expansive research studies in terms of scale, scope, and duration. Unfortunately, there is limited public data shared with respect to the data quality, device longevity, and scaling issues that emerge when trying to execute such studies. To that end, we share real-world results with respect to data quality, participant compliance, and device efficacy on a large scale, longitudinal study involving over seven hundred and fifty working professionals over the period of an entire year. In this paper, we present analyses with respect to the different types of data being collected including sleep, heart rate, physical activity, and stress. Furthermore, we explore participants behavior regarding charging frequency, and device robustness to further aid researchers considering large scale wearable studies.
Commercial grade activity trackers and phone agents are increasingly being deployed as sensors for sleep in large scale, longitudinal designs. In general, wearables detect sleep through diminished movement and decreased heart rate (HR), while phone agents look for lack of user input, movement, sound or light. However, recent literature suggests that commercial-grade wearables and phone apps vary greatly in the accuracy of sleep predictions. Constant innovation in wearables and proprietary algorithms further make it difficult to evaluate their efficacy for scientific study, especially outside of the laboratory. In a longitudinal study, we find that wearables cannot detect when a person is laying still but using their phones, a common behavior, overestimating sleep when compared to self-reports. Therefore, we propose that fusing wearables and phone sensors allows for more accurate sleep detection by capitalizing on the benefits of both streams: combining the movement detection of wearables with the technology usage detected by cell phones. We determine that fusing phone activity to wearables can generate better models of self-reported sleep than either stream alone, and test models in two separate datasets.
Impaired social functioning is a symptom of mental illness (e.g., depression, schizophrenia) and a wide range of other conditions (e.g., cognitive decline in the elderly, dementia). Today, assessing social functioning relies on subjective evaluations and self assessments. We propose a different approach and collect detailed social functioning measures and objective mobile sensing data from N=55 outpatients living with schizophrenia to study new methods of passively accessing social functioning. We identify a number of behavioral patterns from sensing data, and discuss important correlations between social function sub-scales and mobile sensing features. We show we can accurately predict the social functioning of outpatients in our study including the following sub-scales: prosocial activities (MAE = 7.79, r = 0.53), which indicates engagement in common social activities; interpersonal behavior (MAE = 3.39, r = 0.57), which represents the number of friends and quality of communications; and employment/occupation (MAE = 2.17, r = 0.62), which relates to engagement in productive employment or a structured program of daily activity. Our work on automatically inferring social functioning opens the way to new forms of assessment and intervention across a number of areas including mental health and aging in place.
I have taught a few undergraduate classes after I obtained my Bachelors degree in Nepal. During my PhD, I have had experience of TAing in a couple more. I enjoy teaching! I love breaking down the concepts into easy-to-understand bits in order to help students see a clearer picture of how and why things function the way they do.
Dartmouth College, USA.
Dartmouth College, USA.
Dartmouth College, USA.
Dartmouth College, USA.
Dartmouth College, USA.
Dartmouth College, USA.
Deerwalk Institute of Technology, Kathmandu, Nepal
Deerwalk Institute of Technology, Kathmandu, Nepal
Please feel free to reach out to me in any of the social handles mentioned here.