Table 4.
Studies on AI-assisted interventions in mental health
| Ref. | Subject description | Mental health condition | Aim | AI-based method | Intervention description | Control group intervention | Intervention duration | Outcome measures | Results and accuracy | Conclusions |
|---|---|---|---|---|---|---|---|---|---|---|
| Karkosz et al. (2024) | People reported at least mild depressive or anxiety symptoms (n = 81) | Depression, anxiety, positive and negative affect, global life satisfaction and loneliness | To assess Fido’s efficacy, a therapy chatbot targets depressive and anxiety symptoms through CBT techniques | AI chatbot | Fido focuses on dialogue to recognize and modify cognitive biases using Socratic questioning. It identifies suicidal ideation, guiding users to emergency hotlines. Fido utilizes the ABC technique from CBT, provides psychoeducation on mental health, and offers gratitude practice exercises | Received a book containing psychoeducation and self-help exercises, similar in content to that provided to the intervention group | Two weeks |
|
Depressive and anxiety symptoms decreased after the intervention and remained stable at the 1-month follow-up. Although loneliness was not significantly different between groups post-intervention, frequent Fido users showed a decline in loneliness | Fido provided sufficient help to reduce anxiety and depressive symptoms and decreased perceived loneliness among high-frequency users |
| Kleinau et al. (2024) | Health workers from public and private healthcare facilities (n = 1584) | Depression, anxiety, loneliness and burnout | To assess the effectiveness of the interactive chatbot, Vitalk, in improving mental well-being and resilience outcomes among health workers | AI chatbot | Vitalk utilizes an automated chatbot named Viki to provide mental health support through conversations based on CBT and positive psychology principles. Users engage in reflective discussions, access mood tracking tools, and participate in themed conversations to manage stress, mood, and anxiety. The platform also offers feedback, guidance, and emergency support information when necessary | Received access to a webpage with links to mental health resources (also wait list) | Eight weeks |
|
Although there were statistically significant differences in the average scores for mental health (depression, anxiety, loneliness, and burnout) between the control and treatment groups, these differences were very small and both groups fell within the same risk category. The Difference-in-Differences estimates suggested a significant positive effect of Vitalk in reducing anxiety and depression. Depression showed the largest difference in effect size between the control and treatment groups |
Vitalk’s positive impact on mental well-being and resilience makes it a promising tool against work-related stress and burnout |
| Schillings et al. (2023) | People without diagnosed mental disorders reported a moderate to high perceived stress level (PSS–10 score ≥ 14) (n = 118) | Stress and subjective well-being | To evaluate the effects of a chatbot intervention led by ELME on reducing stress and improving health-related parameters in individuals with medium to high stress levels | AI chatbot | ELME, a rule-based chatbot accessible as a web-based mobile application, provided psychoeducation, real-time dialogues, audio exercises, and personalized feedback to participants. It conducted two daily interactive intervention sessions (10–20 minutes each), focusing on stress, mindfulness, and interoception. The sessions were flexible, allowing participants to postpone exercises and receive SMS reminders |
Usual care | Three weeks |
|
There were no significant changes in perceived stress levels over time (from T1 to T3) and no significant effects of group or interactions between time and group on momentary perceived stress in the two models Subjective well-being showed an average improvement over time in both groups. No significant differences were found between the groups and there were no significant changes observed over time and between groups |
Further research is needed to optimize the effectiveness of chatbot interventions for stress reduction through considerations such as intervention duration, target populations, and tailored approaches |
| Suharwardy et al. (2023) | Women within 72 hours postpartum (n = 192) | Depression and anxiety | To evaluate the acceptability and preliminary efficacy of a mental health chatbot for mood management in a general postpartum population | AI chatbot | Peripartum-specific content and psychotherapeutic techniques from CBT and IPT for postpartum mood were integrated into the AI chatbot to assist postpartum mothers in coping with mood and anxiety | Usual care | Six weeks |
|
There was a statistically significant difference in mean change scores from baseline to 6 weeks for PHQ–9 between the two groups. However, at the 6-week mark, there were no statistically or clinically significant variances between the groups in terms of EPDS scores, and there were also no differences observed in anxiety levels between the two groups | Given that the sample did not screen positive for depression at baseline, the potential of the chatbot to reduce depressive symptoms in this general obstetric population was limited |
| Yamada et al. (2024) | Patients with hematological malignancies who have undergone hematopoietic stem cell transplantation in a protective isolation unit (n = 21) | Stress and depressive symptoms | To examine if using a robotic puppy, aibo, could benefit the mental health of patients with hematological malignancies undergoing stem cell transplantation | AI robotic puppy with deep learning | Each patient was allowed to pet an aibo during the treatment period and was free to name the aibo Aibo uses sensing technology to recognize users and their surroundings, enabling it to assess situations and make decisions using artificial intelligence and deep learning. Its personality is shaped by interactions, experiences, and environmental mapping, allowing it to embody a unique animal-like character |
Usual care | Entire stay in isolation unit |
|
At discharge, the intervention group exhibited a significant decrease in CgA, a significant increase in oxytocin, and a significantly more pronounced decrease in cortisol compared to the control group. Throughout the hospitalization period, the intervention group demonstrated a significant decrease in CgA levels and a significant increase in oxytocin levels, while the control group showed no significant change in CgA and a significant decrease in oxytocin No significant difference in the overall QIDS-J score, but there was a significant improvement in the psychomotor activity subscale in the intervention group |
The AI robotic puppy intervention during a stay in an isolation unit can improve the mental health of patients with hematological malignancies |
| Chen et al. (2023) | Patients aged 20–65 with schizophrenia who lived at a psychiatric daycare center (n = 105) | Schizophrenia | To explore the effectiveness of intervention with the MedAdhere app on medication adherence and accuracy in patients with schizophrenia | AI-based app | Nighttime medication involves the MedAdhere app, a tool for medication management including scheduling, reminders, tracking, adherence assessments, and facial and antipsychotic recognition | Nighttime medication is self-administered by the patient without intervention | 12 weeks |
|
|
The app effectively and significantly improved medication adherence and the psychiatric symptoms of patients with schizophrenia |
| Sabour et al. (2023) | Healthy adults (n = 301) | Depression, anxiety, positive and negative affect | To evaluate Emohaa’s effectiveness in reducing mental distress symptoms through CBT-Bot exercises and guided conversations | AI chatbot | The CBT chatbot, rooted in CBT principles, uses interactive exercises like automatic thoughts training and guided expressive writing to address irrational thoughts and enhance mental well-being. Users engage in exercises via conversational choices, focusing on diary entries and hypothetical scenarios to gain new perspectives. Post-exercise, users report mood and emotions. The emotion support (ES) chatbot platform employs a strategy-driven dialogue model and a safety feature to detect suicidal signs, providing immediate help. It allows free-flowing conversations. Group 1: only CBT chatbot Group 2: CBT and ES chatbot |
Wait list | Three weeks |
|
Compared to the control group, participants who used two types of Emohaa experienced significantly more improvement in symptoms of mental distress, including depression and negative affect | Emohaa is a practical and effective tool for reducing mental distress |
| Sadeh-Sharvit et al., 2023 | Participants diagnosed with depressive or anxiety disorders who require outpatient individual CBT (n = 47) | Depression and anxiety | To evaluate the feasibility, acceptability, and initial effectiveness of an AI platform designed to assist therapists in delivering mental health services | AI platform assists therapists in mental health services | The Eleos Health Platform is a secure tool for behavioral health professionals, supporting clinical decisions, automating tasks, analyzing therapist-patient dialogues, offering feedback on evidence-based practices, and facilitating measurement-based care and progress note generation. Therapists receive training to independently deliver interventions without prescribed practices |
Usual care | Two months |
|
The intervention group attended 67% more sessions than the control group, with reductions of 34% in depression and 29% in anxiety, compared to 20% and 8% in the control group, respectively, showing significant benefits of therapy supported by the AI platform | Providing therapy in behavioral health settings with the support of an AI platform was more effective than usual care |
| Danieli et al. (2022) | Active workers over 55 with stress symptoms and mild-to-moderate anxiety (n = 60) | Stress and anxiety | To evaluate the contribution of TEO, a mobile personal health care agent with conversational AI | AI chatbot | TEO Mobile personal health care agent to recognize users’ emotional states, beliefs, and personal events, followed by implementing strategies designed by professionals |
|
Eight weeks |
|
In the mixed treatment (tradition in-person and TEO) group, statistically significant difference within group was observed in perceived stress, obsessiveness and compulsiveness, interpersonal sensitivity, depression, hostility, paranoid ideation, psychoticism, task-orientation, mental health, and physical health In the TEO group, statistically significant difference within group was observed in the interpersonal sensitivity and paranoid ideation In the tradition in-person group, statistically significant difference within group was observed in the paranoid ideation and logic In the no treatment group, no significant difference was observed |
Mixed treatment with in-person and TEO components is the most effective in reducing stress and anxiety |
| Ogawa et al. (2022) | Patients aged 20 to 80 from the outpatient clinic with a diagnosis of clinically established or probable PD (n = 20) | Mood, mainly focusing on depression | To assess the feasibility and efficacy of using an AI-based chatbot to improve smile and speech in participants with PD, and to explore the potential predictive value of objective face and speech parameters for motor symptoms, cognition, and mood | AI chatbot | Daily chatbot (including multi-turn conversation to simulate a typical teleconsultation, with a report generated for each session) and weekly video-conferencing sessions with a neurologist | Weekly video-conferencing sessions with a neurologist. | Five months |
|
A significant interaction effect was found on the smile index and speech features, but no significant interaction effects were observed for depression. The explorative analysis using statistical and machine learning models revealed that the smile indices and several speech features were associated with depression | An AI-based chatbot improves smile and speech in patients with PD, which indirectly capture the small improvement in depression that cannot be detected by conventional scales |
| Dimeff et al. (2021) | Individuals aged 18 or above who were suicidal and reached out to the Emergency department based psychiatric crisis services (n = 31) | Suicide | To evaluate the feasibility, acceptability, and effectiveness of a tablet-based app, Jaspr Health, among suicidal adults in Emergency Departments | AI tablet- based app with AI chatbot | Jaspr Health, a tablet-based app to conduct a comprehensive suicide risk assessment, a crisis stability plan, lethal means counseling, and education on behavioral skills to improve individual capacity to tolerate future crises, by identifying and treating patient- articulated drivers of suicide | Care as usual | Two hours |
|
Significant decreases in distress and agitation, along with significant increases in learning to cope more effectively with current and future suicidal thoughts, were observed among participants using Jaspr Health compared with those receiving care as usual | Jaspr Health is feasible, acceptable, and clinically effective for use by patients at the Emergency Department who are suicidal and reach out to the Emergency Department based psychiatric crisis services |
| Klos et al. (2021) | University students aged 18 to 33 (n = 181) | Anxiety and depression symptoms | To evaluate the viability, acceptability, and potential impact of using Tess, an AI-based chatbot that delivers brief text conversations as comprehensive support for mental health | AI chatbot | A chatbot developed to send reminders, psycho- educational content, and emotional support responses based on what the users express | An electronic psychoeducation book on depression. | Eight weeks |
|
The GAD–7 score of the participants in the AI chatbot group was significantly reduced compared to the control group No significant difference in the PHQ–9 score was observed |
Tess is effective in addressing anxiety but not depressive symptoms |
| Fulmer et al. (2018) | Students aged 18 and older (n = 75) | Anxiety and depression symptom | To assess the feasibility and efficacy of using an integrative psychological AI, Tess, to reduce self-identified symptoms of depression and anxiety in college students | AI chatbot | A psychological AI chatbot designed to deliver brief conversations in the form of integrative mental health support, psychoeducation, and reminders | An electronic link to the National Institute of Mental Health’s eBook on depression among college students | Two or four weeks |
|
A significant difference was found on the PHQ–9 and the PANAS between participants in the AI chatbot and the control group after a two-week use of AI chatbot. The score of GAD–7 of the participants with two-week or four-week use of AI chatbot was significantly reduced compared to the control group |
Psychological AI has the potential to reduce symptoms of depression and anxiety by delivering CBT-based interventions in the form of conversations |
Abbreviations: AI: Artificial Intelligence; BARS: Brief Agitation Rating Scale; BDI-II: Beck Depression Inventory-II; CBT: Cognitive Behavioral Therapy; CESD-R: Center for Epidemiologic Studies Depression Scale Revised; CgA: Chromogranin A; CSDD: Cornell Scale for Symptoms of Depression in Dementia; EPDS: Edinburgh Postnatal Depression Scale; GAD-7: General Anxiety Disorders-7 scales; OLBI: Oldenburg Burnout Inventory; OSI: Occupational Stress Indicator; PANAS: Positive and Negative Affect Scale; PANSS: Positive and Negative Syndrome Scale; PD: Parkinson’s Disease; PHQ-8: Patient Health Questionnaire-8; PHQ-9: Patient Health Questionnaire-9; PSS: Perceived Stress Scale; PSWQ: Penn State Worry Questionnaire; QIDS: Quick Inventory of Depressive Symptomatology Self-Report; SCL-90-R: Symptom Checklist-90-Revised; SIDQ: Safety and Imminent Distress Questionnaire; SRCS: Suicide-Related Coping Scale; STAI: State–Trait Anxiety Inventory; SWLS: Satisfaction With Life Scale; TEO: Therapy Empowerment Opportunity; WHO-5: 5-item WHO Well-Being Index.