Corina Sas

Background: AI technologies have considerable potential to support mental health practitioners in identifying depression symptoms. However, existing tools are often developed with limited practitioner involvement and evaluated primarily on technical performance, leaving questions of clinical relevance, ethical design, and practitioner trust insufficiently addressed. Objective: This study aimed to examine how iterative co-design with mental health practitioners can inform the development of an AI-based model for identifying depression symptoms, focusing on how practitioner input shaped its clinical relevance and ethical foundations across 2 design cycles. Method: We conducted a 2-stage qualitative co-design study. In stage 1, we developed DepressionSymp 1.0, a fine-tuned GPT-3.5 model trained on the CounselChat dataset, identifying 5 symptom categories based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition ( DSM-5 ) criteria and the Linguistic Inquiry and Word Count (LIWC-22) psycholinguistic framework, reviewed and discussed by 6 mental health practitioners during online co-design workshops. In stage 2, this feedback informed DepressionSymp 2.0, expanded to 10 symptom categories and reviewed by 8 further practitioners in a second round of workshops. Results: Practitioners valued the structured symptom outputs for reducing cognitive workload and supporting intervention planning. In stage 1, indirect indicators such as social withdrawal, financial difficulties, and major life changes were often classified as lifestyle factors rather than symptoms; this informed the expansion to 10 categories in stage 2, described as clearer and more useful for preassessment. Risk categorization was identified as the most ethically sensitive aspect across both stages, alongside broader concerns including the lack of nonverbal information in text-based assessments, cultural and linguistic differences in symptom expression, and the potential for automation bias without sufficient human oversight. Conclusions: Iterative co-design with mental health practitioners is valuable for developing AI-based tools that are clinically relevant and ethically grounded, demonstrating how qualitative practitioner feedback can inform iterative model refinement and ethical design. Findings inform design implications around large language model (LLM) features, practitioner-in-the-loop oversight, and ethical considerations covering clinical standards, informed consent, and data governance. Future work should prioritize larger-scale independent validation, formal safety evaluation, and culturally diverse dataset development.

Background: Design research on AI technologies for mental health has expanded rapidly, yet existing work has largely focused on practitioners’ perspectives within isolated therapeutic stages or specific services. As a result, there is limited understanding of how practitioners view the role of AI across the full therapeutic journey, including assessment, diagnosis, and treatment, or how these perspectives can be translated into actionable and ethically grounded design guidance. This lack of stage-specific guidance introduces several risks: (1) the development of AI tools that practitioners may not trust or adopt, (2) heightened ethical concerns, particularly in diagnosis where bias and erosion of human judgment are most likely, and (3) missed opportunities to leverage AI to support ongoing assessment and treatment adaptation. Objective: This study aimed to explore mental health practitioners’ perspectives on how AI may support or hinder therapeutic practice across all 3 therapeutic stages, with particular attention to their ethical considerations. Methods: We conducted semistructured interviews with 18 mental health practitioners to explore how AI may support or hinder therapeutic practice across assessment, diagnosis, and treatment. Data were analyzed thematically, and ethical challenges were organized using 4 biomedical principles. This study received approval from the Lancaster University Institutional Ethics Committee (ID 3530). Results: Findings suggest a stage-specific pattern: practitioners identified valuable roles for AI in supporting clinical assessment and treatment, while expressing strong resistance to autonomous AI diagnosis, which was regarded as the most ethically sensitive stage. Across contexts, AI was seen as supporting and validating, rather than replacing, clinical judgment. Ethical concerns were organized using these principles, with key issues including data validity, inclusivity, privacy, transparency, and explainability. Grounded in these findings, four design directions are proposed: (1) AI-enabled clinical assessment hubs for analyzing multimodal patient data to support transparent decision-making; (2) natural language processing-based tools to enable richer understanding of patients’ lived experiences; (3) new models of trust to support ethically sensitive AI-assisted diagnostic decision-making; and (4) large language model–powered personalized and adaptive mental health support systems that maintain human-AI partnership and preserve practitioners’ central role in care. Conclusions: This study provides a stage-specific understanding of practitioners’ perspectives on AI use across assessment, diagnosis, and treatment and systematically frames ethical challenges using biomedical principles. The findings translate practitioner insights into actionable design guidance for ethically informed AI-supported mental health practice.

Personal informatics research has predominantly explored the setting and tracking of quantitative goals, with less focus on qualitative goals. To address this gap, we report staged participatory design sessions with 26 stakeholders, including goal-oriented work practitioners, self-trackers, and designers. Leveraging practitioners’ expertise, we co-designed a novel multi-session, 4-week intervention for goal identifying and goal setting, consisting of sequentially scheduled tools from goal-oriented work. We also explored the intervention’s pre-use acceptability, the perception of user interaction, and usability of the digital format. Our findings informed four design implications to support identification of ambitious yet realistic goals, structured and customizable support for goal identification and setting, iterative goal revision, and reduced reliance on quantitative tools. We also reflect on the value of engaging multiple stakeholders in participatory design.

Much HCI research has explored user experience and, in particular, emotional ones, reflected in the increasing interest in well-being and affective health technologies. While much such research has focused on emotional awareness and regulation, less work has explored how to design for emotional appraisal. We report a 2-week diary study informed by the conceptual model of Lazarus’s appraisal theory to explore emotional appraisal in the wild by proposing two novel design research methods: the Emotional Appraisal Kit and the Emotional Appraisal Diary. Findings indicate that these tools support this shifting in paradigm by providing a holistic method that considers emotional, physical, and computational elements. Findings also highlight the value of such tools in capturing the emotional process, going beyond emotional recognition to provide a deeper insights into what triggers emotions.

Emotion regulation skills are crucial for wellbeing, yet mastering them is not trivial. Because of wearability, private passthrough display, and ability to keep users connected to their surroundings, Augmented Reality (AR) technologies have potential for supporting on-the-spot emotion regulation. We designed TouchEmotion, an AR design exemplar consisting of three prototypes aimed to support metaphorical emotional trajectories for regulating anxiety, anger, and boredom. We employed Snap Spectacles to design interactive animations and touchpad interactions, which we examined through an exploratory study with 18 participants. Findings indicate potential of AR technologies and in particular of metaphorical emotional trajectories coupled with touch-based interactions for on-the-spot regulation of three discrete negative emotions. We advanced a design space for on-the-spot emotion regulation technologies, and proposed five design implications: nature-inspired slow transitions, frictions to support frustration tolerance, layered randomness to support surprise, and finger-worn devices integrated with AR technologies for increased acceptance in public contexts.

Eating is complex, as is people's relationship to food, which significantly impacts wellbeing and health. Mindful eating interventions have been shown to be effective in addressing problematic eating, and mindful eating has also gained increasing interest as a research topic in HCI, albeit with limited grounding in health research, so that we know little about how HCI work could be informed by health research on mindful eating and its interventions. Such a theoretical foundation grounded in health research is crucial to ensure the design of safe and effective mindful eating technologies and interventions. To address this gap, we present an analysis of health research on mindful eating principles, measurement scales, and therapeutic interventions such as the MB-EAT program, which helped us identify the main aspects of mindful eating. Then, we conducted a scoping review on technologies targeting these aspects, from which we curated 16 design exemplars representing the breadth of these technologies and generated a new SmartPlate conceptual design. Then, we designed the Mindful Eating Design Critique (MEDEC) cards and reported workshops with 36 mindful eating practitioners who used the MEDEC cards to critique the set of design exemplars. Our main contributions include a solid theoretical foundation grounded in health research for the design of mindful eating technologies, the MEDEC cards as a novel critique tool consisting of 29 cards improved based on practitioners’ feedback, and a design framework to further support research and development of mindful eating technologies.

BACKGROUND Design research on AI technologies for mental health has grown substantially, but prior work often focuses on practitioners’ perspectives within a single therapeutic stage or specific service. Less research has mapped how practitioners view AI across the main stages of therapy (assessment, diagnosis, and treatment) or translated their perspectives into actionable, ethically grounded design guidance. To address this gap, we conducted interviews with 18 mental health practitioners to explore how AI may support or hinder therapeutic practice, with attention to ethical considerations. Findings show that practitioners see potential for AI across all therapy stages to aid data analysis and transparent decision-making, under human oversight. Notably, there is a stage-specific pattern: high value for assessment (38%) and treatment (55%), but strong resistance to autonomous AI diagnosis (6%), which is considered the most ethically sensitive stage. Practitioners emphasized that AI should validate and support decisions rather than replace human judgment. Ethical challenges were organized using four biomedical principles: non-maleficence, beneficence, justice, and autonomy. Key concerns include validity, inclusivity, accessibility, privacy, transparency, and explainability. Practitioners highlighted depression-specific challenges, such as gender-based diagnostic bias, age-related variations, and ethnic disparities. Our paper addresses these gaps by: (1) providing a stage-specific analysis of AI support within and across assessment, diagnosis, and treatment; (2) systematically framing ethical concerns using biomedical principles; and (3) translating practitioner insights into design implications. We propose three design directions: (1) clinical assessment hubs using AI to analyze multimodal patient data for transparent decision-making; (2) NLP-AI tools for richer understanding of patients’ experiences; and (3) LLM-powered personalized, adaptive mental health support systems that maintain human-AI partnership and practitioners’ central role. OBJECTIVE To explore mental health practitioners' perspectives on AI technologies across the main stages of therapeutic practice (assessment, diagnosis, and treatment) and identify stage-specific ethical challenges mapped to biomedical ethics principles. METHODS We conducted semi-structured interviews with 18 mental health practitioners (16 female, 2 male; mean experience = 9 years) in the UK. Participants were accredited members of the National Association for Counselling and Psychotherapy. Each interview began with a 10-minute presentation on AI concepts and a scenario involving an AI system for depression detection. Interviews lasted 48-65 minutes and were conducted via Microsoft Teams. Data were analyzed using hybrid coding (deductive and inductive) with ATLAS.ti software, applying biomedical ethics principles (non-maleficence, beneficence, justice, autonomy) and therapeutic stages as main codes. Over 1,130 codes were generated through iterative analysis. RESULTS Practitioners assigned different levels of value to AI across therapeutic stages: treatment (55%), clinical assessment (38%), and diagnosis (6%). They emphasized that AI should validate and support practitioners' decisions rather than operate autonomously, particularly for diagnosis, the most ethically sensitive stage. Practitioners saw value in AI for gathering and analyzing multimodal patient data, recommending personalized interventions, monitoring treatment effectiveness, and supporting patient engagement. Ethical challenges were organized under four principles: non-maleficence (concerns about limited data validity and potential harm), beneficence (need for comprehensive assessment tools and therapeutic rapport), justice (inclusivity, accessibility, and diversity in AI systems), and autonomy (privacy, transparency, and explainability of AI decisions). Depression-specific challenges included gender-based diagnostic bias, age-related symptom variations, and ethnic disparities in diagnosis. CONCLUSIONS Mental health practitioners view AI as valuable when integrated as a supportive tool under human oversight within a human-AI partnership model. We propose four design directions: clinical assessment hubs integrating multimodal data, NLP-AI tools for deeper understanding of patient experiences, co-designed trust models positioning AI as a validator of diagnostic judgment, and LLM-powered personalized support systems maintaining practitioner oversight.