About

Saeed Abdullah is an Associate Professor of Human-Centered Computing and Social Informatics at Penn State College of Information Sciences and Technology. He is also a Faculty Affiliate at the Center for Socially Responsible Artificial Intelligence. His areas of expertise include Digital Health, Human-Computer Interaction, Human-Centered AI for Health, Data Science, and Education. Abdullah holds a Ph.D. in Information Science from Cornell University, an M.S. in Computer Science from the University of Vermont, and a B.S. in Computer Science and Engineering from Bangladesh University of Engineering and Technology. His research focuses on wellbeing and health innovation, emphasizing the development of human-centered AI systems and digital health solutions. He is involved in research projects and labs dedicated to these fields, contributing to advancing interdisciplinary approaches in informatics and health technology.

Research topics

• Computer Science
• Artificial Intelligence
• Computer Security
• Data Mining
• Human–computer interaction
• Machine Learning
• World Wide Web
• Psychology
• Communication
• Pedagogy
• Database
• Applied psychology
• Internet privacy
• Social psychology
• Data science
• Geography

Selected publications

• Prompt Coaching for Inclusiveness: A Media Literacy Approach to Increase Users’ Awareness of Algorithmic Bias and Prompting Efficacy

  2026-04-13 · 2 citations

  articleOpen access

  Large language models often produce biased or stereotypical outputs. One way to reduce this possibility is to be more inclusive in our prompts, but doing so may not come naturally to most users. Therefore, we designed a tool that coaches users to write more inclusive prompts—a strategy that leverages design friction to provide a media literacy intervention. Data from a user study (N=344) show that compared to no coaching, inclusive prompt coaching directly increased users’ awareness of algorithmic bias and their perceived prompting efficacy. It also indirectly enhanced their trust in the system and perceived trust calibration through cognitive elaboration. However, inclusive prompt coaching resulted in a less satisfying user experience. These findings have implications for ethical interventions in prompting for better communicating and combating algorithmic bias. We discuss the benefits and limitations of inclusive prompt coaching, as well as ways to balance usability for long-term adoption of generative AI systems.

  PublisherDOI

• Characterizing middle-aged and older adults’ perceptions of the cultural sensitivity and quality of generative artificial intelligence-authored text messages to promote physical activity

  The Journals of Gerontology Series B · 2026-04-03

  articleOpen access

  OBJECTIVES: Generative artificial intelligence (GenAI) could be used to write text message content in physical activity behavior change interventions for middle-aged and older adults. Yet, biases in GenAI systems could lead to culturally insensitive or low-quality messages. Evaluating the acceptability of GenAI-authored messages is crucial before use in interventions. This research examined middle-aged and older adults' perceptions of the cultural sensitivity and quality of GenAI-authored messages for promoting physical activity, and the person- and message-level factors influencing these perceptions. METHODS: In a cross-sectional survey, middle-aged and older adults (≥40 years of age; N = 630; mean age = 56.8 years; SD = 10.1) read 80 text messages written by GenAI and identified those that were culturally insensitive or had other problems. Descriptive statistics identified the proportion of GenAI-authored messages labeled as having issues. Separate negative binomial regressions examined the participant (zero-inflated) and message factors associated with message issues. RESULTS: Of 49,859 cultural sensitivity and 49,894 quality message ratings, only 4.9% and 6.1% of the messages, respectively, were labeled as having issues. Knowledge of AI-authorship and more favorable attitudes toward AI were associated with identifying more messages as culturally insensitive. Messages generated by prompts that targeted sitting less (compared to moving more) or that described preparing for activity (compared to performing physical activity) received more labels as containing quality issues. DISCUSSION: GenAI can be prompted to write high-quality, culturally sensitive text messages for promoting physical activity for middle-aged and older adults. Message content and participants' knowledge of AI use could influence perceptions.

  PublisherOA PDFDOI

• Efficacy of digital cognitive stimulation therapy for people with dementia: a systematic review and meta-analysis

  BMC Geriatrics · 2026-04-06

  articleOpen accessSenior authorCorresponding

  There is growing interest in using digital health technologies to deliver Cognitive Stimulation Therapy (CST) for people with dementia, given the potential to improve access and outcomes. However, evidence for the effectiveness of delivering CST using digital technologies (“digital CST”) remain limited, making it difficult to determine best practices. This study systematically reviews and evaluates the efficacy of digital CST on cognitive and psychosocial outcomes in people with dementia. We conducted a comprehensive literature search in CINAHL, Cochrane Library, Embase, MEDLINE, PubMed, and Web of Science. We included randomized controlled trials and quasi-experimental studies that examined the effects of digital CST in people with dementia. We assessed methodological quality using the Cochrane Risk of Bias tool (RoB 2) and ROBINS-I. We applied a DerSimonian and Laird random-effects model to estimate pooled standardized mean differences (SMDs). Eighteen studies comprising 756 participants met inclusion criteria. Digital CST significantly improved cognitive function (k = 559; SMD = 0.33; 95% CI: 0.04–0.62; p = 0.03; I2 = 61.39%) and semantic fluency (k = 181; SMD = 0.43; 95% CI: 0.13–0.73; p < 0.001; I2 = 0.00%). It also significantly reduced depressive symptoms (k = 273; SMD = –0.66; 95% CI: –1.12 to –0.20; p = 0.01; I2 = 68.94%). The greatest cognitive improvements were observed when interventions were delivered biweekly in 45-minute sessions over fewer than eight weeks. This systematic review and meta-analysis demonstrates that digital CST is effective in enhancing cognitive performance and supporting aspects of psychosocial functioning in people with dementia. Improvements were observed particularly in global cognition, semantic fluency, and depressive symptoms, indicating that digital CST offers a beneficial non-pharmacological option within dementia care. Further research is warranted to explore its applicability across different dementia stages and to determine the sustainability of these effects over time. CRD420250653224.

  PublisherDOI

• Fine-Tuning Large Audio-Language Models with Lora for Precise Temporal Localization of Prolonged Exposure Therapy Elements

  2026-04-21

  articleOpen accessSenior author

  Prolonged Exposure (PE) therapy is an effective treatment for post-traumatic stress disorder (PTSD), but evaluating therapist fidelity remains labor-intensive due to the need for manual review of session recordings. We present a method for the automatic temporal localization of key PE fidelity elements, identifying their start and stop times, directly from session audio and transcripts. Our approach fine-tunes a large pre-trained audio-language model, Qwen2-Audio, using Low-Rank Adaptation (LoRA) to process focused 30-second windows of audio-transcript input. Fidelity labels for three core protocol phases, therapist orientation (P1), imaginal exposure (P2), and post-imaginal processing (P3), are generated via LLM-based prompting and verified by trained raters. The model is trained to predict normalized boundary offsets using soft supervision guided by task-specific prompts. On a dataset of 308 real PE sessions, our best configuration (LoRA rank 8, 30s windows) achieves a mean absolute error (MAE) of 5.3s across tasks, within typical rater tolerance for timestamp review, enabling practical fidelity QC. We further analyze the effects of window size and LoRA rank, highlighting the importance of context granularity and model adaptation. This work introduces a privacy-preserving, scalable framework for fidelity tracking in PE therapy, with potential to support clinician training, supervision, and quality assurance.

  PublisherOA PDFDOI

• Voice Assistants to Deliver Cognitive Stimulation Therapy for Persons Living with Dementia

  2025-04-23 · 6 citations

  articleOpen accessSenior author

  Dementia is a global public health concern, with approximately 55 million individuals worldwide currently living with it. Individual cognitive stimulation therapy (iCST) has been shown to improve quality of life for persons living with dementia (PLwDs). However, providing iCST at scale remains a serious challenge, specifically as it can add considerable burden on care partners. This project focuses on a voice assistant (VA) to support care partners in delivering iCST. We developed a VA prototype and conducted a qualitative study with care partners (N=5). Our preliminary findings show that using a VA to deliver iCST is feasible and acceptable. We have also identified design requirements for the VA to effectively provide iCST, including need for personalization reflecting dementia severity and individual interests, collaboration between care partners and PLwDs, and accessible interactions to minimize frustration and distress. These findings can inform the future design of inclusive and accessible VAs.

  PublisherOA PDFDOI

• Technoskepticism or Justified Caution? The Future of Human-Centered AI in Mental Health Care

  2025-04-23

  article
  PublisherDOI

• 10$th$ International Workshop on Mental Health and Well-being: From Research to Practice in Mental Healthcare

  2025-10-12

  articleOpen access

  Ubiquitous computing technologies (UbiComp) are emerging as crucial tools for collecting behavioral, physiological, social, and environmental data to enable early symptom detection, deliver preventative interventions, and support ongoing symptom management. With decades of success in demonstrating the feasibility of using UbiComp technologies to support well-being and mental health in general populations, researchers are exploring the use of these technologies for clinical populations living with mental illness, such as schizophrenia. However, designing, implementing, and validating these technologies in a clinical setting is complex and faces multiple challenges, including ensuring clinical relevance, developing novel analytics systems, integration into existing care systems, user engagement, ethical considerations, and long-term feasibility. This workshop aims to bring together researchers, service providers, practitioners, and industry professionals to collaboratively explore these challenges and discuss strategies for evaluating and validating these technologies in real-world clinical settings. We are calling for papers that inspire new research directions, including co-designing systems with multiple healthcare stakeholders. Building on nine years of success, we continue to support the UbiComp community in advancing reliable, responsible, and effective mental health technologies that can potentially extend UbiComp technologies to support improving patient outcomes in clinical settings at scale.

  PublisherDOI

• The Pursuit of Empathy: Evaluating Small Language Models for PTSD Dialogue Support

  ArXiv.org · 2025-05-21

  preprintOpen accessSenior author

  This paper investigates the capacity of small language models (0.5B-5B parameters) to generate empathetic responses for individuals with PTSD. We introduce Trauma-Informed Dialogue for Empathy (TIDE), a novel dataset comprising 10,000 two-turn conversations across 500 diverse, clinically-grounded PTSD personas (https://huggingface.co/datasets/yenopoya/TIDE). Using frontier model outputs as ground truth, we evaluate eight small LLMs in zero-shot settings and after fine-tuning. Fine-tuning enhances empathetic capabilities, improving cosine similarity and perceived empathy, although gains vary across emotional scenarios and smaller models exhibit a "knowledge transfer ceiling." As expected, Claude Sonnet 3.5 consistently outperforms all models, but surprisingly, the smaller models often approach human-rated empathy levels. Demographic analyses showed that older adults favored responses that validated distress before offering support (p = .004), while graduate-educated users preferred emotionally layered replies in specific scenarios. Gender-based differences were minimal (p &gt; 0.15), suggesting the feasibility of broadly empathetic model designs. This work offers insights into building resource-efficient, emotionally intelligent systems for mental health support.

  PublisherOA PDFDOI

• DOSE: An open-source, iOS watch-based tool for experience sampling

  medRxiv · 2025-10-14

  preprintOpen access

  Smartwatches facilitate low-burden rapid-access micro-interactions, making them ideal for Experience Sampling Methods (ESMs). Despite the Apple Watch being the most popular smartwatch in the U.S., it has yet to be utilized in ESM studies due to a lack of accessible frameworks that enable deployment without technical expertise. We developed DOSE, an open-source ESM framework tailored for the Apple Watch. It includes tools and documentation that allow researchers to configure surveys, build custom apps, deploy studies, and stream data to servers without programming skills. We evaluated the framework's feasibility in a 28-day field study with 18 participants (mean age = 55.3 ± 9.2). Results showed reliable prompt delivery and high response rates (>80% overall), with median interaction times under 10 seconds. Participants demonstrated increasing efficiency in responses over time. These findings establish the DOSE framework as a practical, scalable solution for Apple Watch-based ESMs and a foundation for future smartwatch research.

  PublisherOA PDFDOI

• The Currency of Mood: Assessing Acceptance and Privacy Preferences of Third-party Financial Data Sharing in Bipolar Disorder

  2025-10-07

  preprintOpen accessSenior author

  Objectives: Bipolar disorder is strongly associated with financial instability. We examine how different interventions motivate individuals with bipolar disorder to share financial data with others. This approach can inform the development of tools for digital monitoring and intervention designed to promote financial stability in this population.Methods: 500 individuals with BD completed a pre-registered factorial vignette survey to examine level of comfort with hypothetical scenarios involving third-party financial interventions during symptomatic and euthymic periods. Scenario components were systematically varied between third-party actors, mood states, and intervention types. Participants rated sharing comfort on a 0-10 point scale. Multilevel models tested differences alongside clinical and financial histories, relational trust, and personality.Results: Participants were most comfortable involving care partners in financial planning. They were more comfortable with temporary spending restrictions during symptomatic states than euthymic periods, underscoring the importance of accurate mood detection for intervention delivery. Prior financial help-seeking behavior and higher relational trust predicted greater comfort. Bankruptcy experience — declared by 11.4% and considered by 31.7% — was associated with increased comfort with spending restrictions. Individuals with psychiatric advance directives (8%) were significantly more comfortable sharing spending behaviors than those without. Conclusions: Comfort with financial interventions was higher among those with prior financial challenges or help-seeking histories. Participants distinguished between symptomatic and euthymic periods, favoring targeted, time-limited restrictions over general monitoring. These findings extend prior work on financial data sharing for illness self-management, highlighting the role of trusted third parties in designing acceptable, effective interventions.

  PublisherOA PDFDOI

Recent grants

• CRII: SCH: Designing for Supportive Accountability: Using Conversational Agents to Sustain Patient Engagement in Post-Traumatic Stress Disorder

  NSF · $175k · 2019–2023

Frequent coauthors

• Tanzeem Choudhury

  Cornell University

  31 shared

• Mark Matthews

  18 shared

• Elizabeth L. Murnane

  Dartmouth College

  16 shared

• Geri Gay

  14 shared

• Johnna Blair

  Pennsylvania State University

  11 shared

• Sahiti Kunchay

  Pennsylvania State University

  10 shared

• Jakob E. Bardram

  Technical University of Denmark

  9 shared

• Jason E. Owen

  National Center for PTSD

  9 shared