Mental health practitioners’ perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study

Mental health practitioners’ perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Mental health practitioners’ perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study Julia Cecil, Anne-Kathrin Kleine, Eva Lermer, Susanne Gaube This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4692251/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: As mental health disorders continue to surge, exceeding the capacity of available therapeutic resources, the emergence of technologies enabled by artificial intelligence (AI) offers promising solutions for supporting and delivering patient care. However, there is limited research on mental health practitioners’ understanding, familiarity, and adoption intentions regarding these AI technologies. We, therefore, examined to what extent practitioners’ characteristics are associated with their learning and use intentions of AI technologies in four application domains (diagnostics, treatment, feedback, and practice management). These characteristics include medical AI readiness with its subdimensions, AI anxiety with its subdimensions, technology self-efficacy, affinity for technology interaction, and professional identification. Methods: Mixed-methods data from N = 392 German and US practitioners, encompassing psychotherapists (in training), psychiatrists, and clinical psychologists, was analyzed. A deductive thematic approach was employed to evaluate mental health practitioners’ understanding and familiarity with AI technologies. Additionally, structural equation modeling (SEM) was used to examine the relationship between practitioners’ characteristics and their adoption intentions for different technologies. Results : Qualitative analysis unveiled a substantial gap in familiarity with AI applications in mental healthcare among practitioners. While some practitioner characteristics were only associated with specific AI application areas (e.g., cognitive readiness with learning intentions for feedback tools), we found that learning intention, ethical knowledge, and affinity for technology interaction were relevant across all four application areas, making them key drivers for the adoption of AI technologies in mental healthcare. Conclusion : In conclusion, this pre-registered study underscores the importance of recognizing the interplay between diverse factors for training opportunities and consequently, a streamlined implementation of AI-enabled technologies in mental healthcare. mental healthcare artificial intelligence technology implementation use intention learning intention Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction One in eight people worldwide is affected by a mental disorder, and the trend is rising (1). Frequently, the demand for therapeutic support exceeds available resources, especially since the number of mental health practitioners is not increasing quickly enough (2). Simultaneously, technologies enabled by artificial intelligence (AI) are advancing and gaining relevance in the support and delivery of patient care, owing to their potential for improving patient outcomes through an early detection of mental disorders and personalized treatment (3), and facilitating the work of practitioners (4). Given the proposed benefits, AI-enabled technologies provide an opportunity to bridge the gap between mental healthcare needs and available therapeutic resources. Applications of AI-enabled technologies in mental healthcare AI-enabled technologies refer to systems or applications characterized by humanlike capabilities, including decision-making through problem solving and continuous learning (3). To execute their tasks effectively, these technologies rely on large amounts of data. Common data sources for AI-enabled technologies in mental healthcare include behavioral data (e.g., video and audio recordings), followed by biological (e.g., blood samples) and neuroimaging data (e.g., electroencephalogram) (5). Within mental healthcare, AI-enabled technologies utilized by clinicians that leverage these datasets can be broadly categorized into four application areas: diagnostic support, treatment support, feedback, and practice management. The first two application areas, diagnostic and treatment support, refer to patient-centered technologies. Diagnostic applications leverage AI to enhance the accuracy and efficiency of mental health assessments by evaluating a range of patient data, such as genetic information, language, voice, and facial expressions (6–8). For example, certain tools can distinguish between diagnoses that share similar symptoms but require different treatment approaches, such as various types of dementia or bipolar and unipolar depression (9). The second area of technologies provides treatment support, making mental health treatments more personalized and precise (10). These technologies are predominantly working with genetic, neuroimaging, clinical and demographical datasets (11). For instance, AI-enabled technologies can be utilized at the beginning of therapy to estimate a patient’s potential response to different medications, such as antidepressants, or to predict remission rates (11). Besides these patient-centered technologies, an increasing number of practitioner-centered applications are emerging, with the third area comprising feedback tools for mental health professionals: These types of applications aim to provide practitioners with feedback on the quality of their patient interactions by evaluating session data, for instance, through speech signals and the language patterns of the interaction (12–15). Feedback reports usually include an assessment of the session’s strengths and potential areas for improvement, such as increasing the times for reflections or including more open-ended questions (16). Finally, the fourth application area of AI-enabled technologies for mental health is practice management. They are supposed to automate clinical and administrative workflows and thereby reduce the administrative burden for mental healthcare professionals (16). For example, by automatically transcribing therapy sessions using speech data and integrating the transcripts into medical records (16), patient data entry can become more efficient and structured (17). Adoption of AI-enabled tools in mental healthcare and its antecedents The proposed benefits of using AI tools such as an early detection of mental disorders, increasing patient access, and personalized treatment will only be realized if practitioners use them as intended (7). However, studies show widespread skepticism regarding the use of AI-enabled technologies in healthcare (18,19,9,20,21). A lack of understanding or knowledge of the mechanisms and processes underlying the technology may explain some of the suspicion that impacts the uptake of technologies (22,23). Therefore, gaining deeper insights into the current state of mental health practitioners’ understanding of and experiences with AI-enabled tools is the first step to recognize barriers to the adoption and determine starting point for measures aimed at promoting safe technology practices. However, to the best of our knowledge, no study has investigated practitioners’ understanding of AI-enabled tools for mental healthcare (RQ1), their familiarity with these technologies (RQ2), in what context they learned about them (RQ3), and whether they have used any of these tools in their clinical practice (RQ4). Besides knowledge and exposure, technology acceptance and effective use is influenced by numerous individual variables. The role of learning in the adoption of AI-enabled technologies Studies have highlighted the pivotal role of learning opportunities and training in the implementation process by equipping healthcare professionals with the requisite skills to effectively use AI-enabled technologies in their practice (24–26). Conversely, healthcare professionals ranked the lack of instruction and training on technology use as the primary technology-related cause of medical errors (27). Training is believed to reduce the perceived risk associated with using such tools and, further, minimize the workload arising from the implementation of AI technologies (28). It has been shown that the willingness to receive training about an AI technology is positively associated with clinicians’ use of it (28). We, therefore, hypothesized that learning intention is positively associated with use intention for AI-enabled technologies in mental healthcare (H1). Figure 1 depicts the proposed model with the related hypotheses and research questions. However, learning intentions and use intentions represent different levels of engagement with technologies. The willingness to learn and receive training is a rather theoretical interaction with a technology centered around updating knowledge (29). Yet, use intention implies the willingness to make the necessary effort to use the technology in practice (30,31). Hence, it is important to study both the learning and use intention and their respective antecedents independently. Individual-level factors in the adoption of AI-enabled technologies Most studies have focused on AI adoption in general healthcare settings (see (32) for a review) or different medical specialties such as dermatology (33). However, less is known about individual-level factors associated with practitioners’ intentions to learn about and use AI-enabled technologies in mental healthcare. User characteristics represent one of the key determinants for the adoption of healthcare technologies (34). Research showed that common demographic and individual differences such as gender (35), age (36), personality (32,33,37), and country of residence (38,39) influence technology uptake. Further, practitioners’ intention to use AI-enabled technologies in mental health is greatly influenced by their individual beliefs, attitudes, and perceptions (19). Hence, this study seeks to extend existing literature by systematically investigating individual factors that contribute to a holistic understanding of the determinants affecting the learning and use intention of AI-enabled technology in mental healthcare. The Capability-Opportunity-Motivation Behavior (COM-B) model developed by Michie et al.(40), a well-validated behavior change theory, has been used successfully in synthesizing and understanding healthcare-related technology adoption (for instance, see (41,42)). The COM-B model indicates that individuals’ capabilities, motivation, and opportunities determine their behavior (40). Capability is defined as an individual's psychological and physical ability required for a particular behavior, including the essential knowledge and skills. Motivation encompasses reflective or automatic cognitive processes that direct behavior, extending beyond conscious decision-making to habitual patterns, emotional responses, and analytical reasoning. Opportunity relates to external factors lying outside an individual's immediate control that influence behavior, including social and physical opportunity (40). Upon reviewing the empirical literature, we identified the most important individual-level factors relevant to technology adoption and ultimately integrated them into the COM-B framework. As opportunity includes factors outside the individual, we focused on the domains of capabilities and motivations. First, individuals’ capability is important for engaging in a respective behavior (40). Different aspects of capability, including AI knowledge, have been found to be relevant for AI adoption. A positive relation between AI knowledge and the intention to use AI technology was found among prospective physicians (43) and among prospective therapists for feedback providing AI tools (21). Similarly, a lack of technology-related skills and knowledge among therapists was identified as a barrier in the use of technology in forensic psychiatry (44). However, one study found no significant association between AI knowledge and medical students’ intention to learn about AI (45). As AI knowledge referred to different aspects in each study, and the mixed findings consequently might have resulted from methodological differences, we are adopting a broader construct called readiness for medical AI . Readiness for medical AI can be divided into different subdimensions (46): Cognitive readiness encompasses peoples’ cognitive abilities such as knowledge of and critical thinking about AI technologies. Vision readiness involves the ability to envision and anticipate the potential impact, benefits, and challenges associated with AI technologies. Ethical readiness refers to an individual’s awareness, knowledge and adherence to ethical standards or guidelines for the use of AI technologies. The relationship between the subdimensions of medical AI readiness and the learning and use intentions of AI-enabled technologies in mental healthcare has not been examined in-depth. Only one study found a positive association between cognitive readiness and the intention to use a feedback tool in mental healthcare (21). We expected that cognitive readiness (H2a, H3a), vision readiness (H2b, H3b), and ethical readiness (H2c, H3c) are all positively associated with the learning and use intentions of AI tools for mental health (see Figure 1 for all hypotheses). Second, automatic motivational processes influence a particular behavior (40). In the context of technology adoption, automatic processes like emotions, as a sub-component of motivation, have been shown to have an influence (40). Usually, negative valanced variables, such as AI anxiety, have been investigated (47). AI anxiety refers to the apprehension, concern, or fear experienced in response to the implementation, use, or potential consequences of AI technologies (48). The construct encompasses three subdimensions: learning anxiety , sociotechnical blindness , and job replacement anxiety (47). Learning anxiety refers to the anxiety regarding acquiring knowledge and skills related to AI technologies. S ociotechnical blindness relates to anxiety arising from a lack of understanding that AI systems currently do not operate independently without human oversight. Job replacement anxiety refers to a person’s fear that their occupation will be replaced or disrupted by AI technologies (37,49). Y.-M. Wang et al., showed that AI learning anxiety negatively affected intrinsic and extrinsic learning motivation (47). They also found that job replacement anxiety positively influenced extrinsic but not intrinsic learning motivation, indicating that some people might only gain AI-relevant skills and knowledge to avoid unemployment. Regarding use intentions, technology anxiety emerged as one important barrier of technology use in healthcare (50). AI anxiety correlated negatively with the use intention of AI-based technology in healthcare among nurses (51) and the intention to use AI-based treatment and feedback tools among prospective psychotherapists (21). While there is consistent evidence, that AI anxiety hinders AI adoption, none of these studies explored associations between all three subdimensions and learning and use intentions for AI-enabled technologies simultaneously. Therefore, we incorporated all three subdimension separately into our research model. We hypothesized that AI learning anxiety (H2d, H3d) and sociotechnical blindness (H2e, H3e) are negatively associated with both the learning and use intentions of AI tools. Job replacement anxiety is thought to be positively associated with the AI learning intentions (H2f) and but negative with use intentions (H3f). Third, in addition to automatic motivational processes, reflective processes, are also crucial, with self-efficacy being an important factor influencing behavior uptake (40). The subcategory tailored to technology is technology-self efficacy which refers to a person’s belief in their capacity to effectively accomplish a technologically advanced task (52). It is well established that technology self-efficacy is an important predictor of technology adoption in healthcare (53). Higher technology self-efficacy has been positively associated with medical students’ intention to learn technologies (45), healthcare professionals’ readiness to adopt technologies (54) as well as their intention to use nursing apps and AI technology (51,55,56). In accordance with this large body of research, it is hypothesized, that technology self-efficacy is positively associated with AI learning and use intentions among mental health practitioners (H2g, H3g). Fourth, affinity for technology interaction represents another motivational process. It serves as a fundamental resource for technology adoption as it is characterized as the tendency to proactively partake in extensive technological interaction (57). Higher affinity for technology was positively related to using a wider range of learning strategies for different healthcare systems among physician trainees (58). Among clinicians, a positive association between affinity for technology and attitude towards technology use has been found and higher technology affinity was linked to a preference for more advanced technologies (59,60). To the best of our knowledge, the relationship between affinity for technology interaction and the intention to learn or use AI technologies in mental healthcare has not been investigated. Based on previous evidence from the medical context, we hypothesized that affinity for technology interaction is positively associated with AI learning and use intentions (H2h, H3h). Finally, the relevance of people’s perception of their social and professional role and identity as a motivational factor has also been highlighted in the context of technology adoption, often through professional identification. Professional identification refers to the degree to which an individual feels a deep connection and unity with their chosen occupation (61). Professional identification plays an important role in the adoption of novel work behavior (61), particularly important with the integration of AI-enabled technologies that affects practitioners’ daily tasks (62). However, changes in the workplace are likely to be resisted if they are perceived as a threat to professional identity (63). It has been shown that threats to professional identity directly impacted healthcare practitioners’ technology use (64). Moreover, aligned professional beliefs with the designated roles of technology are fundamental for technology adoption (65) as one’s professional identification influences technology integration (63). Given these insights, the following research questions are proposed as we could not derive a clear direction of the effects from the literature: Is professional identification associated with AI learning intention (RQ5) and AI use intention (RQ6)? Prior research has shown that there are differences in use intentions and its predictors across AI tools for different application areas (21). As AI-enabled technologies in mental healthcare differ vastly in their purpose, they might also be perceived differently by mental health practitioners. Therefore, we believe it is important to look at the learning and use intentions and their antecedents individually for each application area. Providing such a nuanced understanding enables technology developers and healthcare organizations who purchase these technologies to consider the factors relevant to the tool in question, thereby facilitating a more efficient and safe design and implementation process. As a consistent methodology that allows comparisons across the different application areas on the same level is fundamental for this, we applied the same research design and sample across all four application areas of AI-enabled technologies in mental healthcare. This allows us to systematically identify potential differences, ultimately resulting in a comprehensive overview of different application areas and their antecedents. The present study The main goal of this mixed method study was twofold. First, we want to investigate mental health practitioners’ general understanding, familiarity, and experience with AI technologies (RQ1 – RQ4) and their attitudes towards different application areas of AI-enabled tools using qualitative content and descriptive analysis. In this line, we also examined differences in attitudes toward technology across different professions, gender, and countries. Second, this work aims to provide a differentiated insight into factors associated with learning and use intentions of AI-enabled technologies for mental health, separated by application areas (H1, H2a – H2h, H3a – 3h, and RQ5 and RQ6). Gaining a deeper understanding of the relative importance of individual factors might help for deriving training and intervention strategies tailored specifically towards practitioners' needs for different technology application areas. Methods Participants Data for the pre-registered (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497) cross-sectional, mixed-methods survey study was collected between July and October 2023. Participants included psychotherapists in training, psychotherapists, psychiatrists, and clinical psychologists. Participants were recruited via emails distributed among universities and psychotherapy training institutes in Germany and the US, social media postings, and Prolific. The online survey was available in German and English language. For the German version of the survey, all items were translated using back-and-forth translation. Both survey versions can be found in the online supplements (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497). In total, 670 mental health practitioners agreed to participate, of which 227 did not finish the survey and 51 failed at least one attention check item, resulting in N = 392 participants included in the data analysis. This number exceeded the minimum sample size determined by the a priori power analysis for structural equation modeling (SEM), which required at least 50 practitioners per country (Germany and US). Demographic information of the included participants can be found in Table 1. The study was approved by the Ethics Committee of [blinded for review]. Procedure First, demographic and occupation-related information was assessed in the survey. Second, participants’ understanding of, familiarity and experiences with, and use of AI-enabled tools were assessed. Third, participants were then introduced to the four different application areas of AI-enabled technologies in mental health. For each area, participants received a short description and an example (see Table 2). We measured learning and use intentions as dependent variables for each application area, the individual level factors as predictor variables, several control and occupation-related variables (occupation, therapeutic approach, workplace, working experience in years) as described below. Measurements Understanding : Participants were asked to describe what they understand by AI-enabled technologies in the field of psychotherapy/psychiatry and how they could be used in their daily work in their own words, using an open text box. Familiarity : Next, they were asked to choose one of three options regarding their familiarity with AI-enabled technologies (a: “I have never heard of AI-enabled technologies in psychotherapy/psychiatry”; b: “I have heard of AI-enabled technologies in psychotherapy/psychiatry”; c: “I have actively looked into AI-enabled technologies in psychotherapy/psychiatry”). Participants who had stated to have heard of AI-enabled technologies were asked in which context they did so (open question). Participants who had stated that they actively looked into AI technology, were given three context options: “I have informed myself independently (e.g., online, …)”, “I attended voluntary information sessions on AI-enabled technologies in psychotherapy/psychiatry”, and “I have participated in trainings on this topic (e.g., to get training points).” Use : To determine previous use, participants were asked to state whether they had used AI-enabled technologies in their clinical practice (yes/no). Dependent variables Two dependent variables, learning intention and use intention , were assessed for each of the four described application areas for AI tools in mental health. Learning intention was measured with “I intend to learn about AI technologies in [application area]” on a 5-point Likert scale from 1 ( strongly disagree ) to 5 ( strongly agree ) based on Venkatesh et al. (31). Similarly, use intention was assessed with the item “I intend to use AI technologies in [application area] in my work” with the same response format (31). Predictor variables Medical AI Readiness : Cognitive, vision and ethical readiness for medical AI was based on the Medical Artificial Intelligence Readiness Scale (MAIRS) from Karaca et al. (66). For each of the subscales we omitted items for two reasons. First, items measuring the actual use of technology were removed, as we assumed that most practitioners are not currently using AI-enabled tools and therefore these questions could not be answered properly. Second, items with low factor loadings were removed to keep the survey reasonable short. Consequently, we included 11 items, rated on a 5-point Likert scale from 1 ( strongly disagree ) to 5 ( strongly agree ). The scale showed acceptable (𝛼 Vision = 0.79, 𝛼 Ethics = 0.73) to good internal consistency (𝛼 Cognition = 0.81). Anxiety : AI learning anxiety, job replacement anxiety and sociotechnical blindness were assessed using the 18-item Artificial Intelligence Anxiety Scale (AIAS) by Wang & Wang (49) on a 7-point Likert scale from 1 ( strongly disagree ) to 7 ( strongly agree ). The internal consistency of the sociotechnical blindness subscale was acceptable (𝛼 Sociotechnical = 0.78), that of the job replacement anxiety subscale good (𝛼 Jobreplacement = 0.87) and that of the AI learning anxiety subscale was excellent (𝛼 Learning = 0.93). Affinity for technology interaction was measured with the Affinity for Technology Interaction Scale (ATI-S (67)). The four items were rated on a 7-point Likert scale from 1 ( completely disagree ) to 7 ( completely agree ). The scale showed good internal consistency (𝛼 Affinity for technology = 0.81). Technology self-efficacy was assessed using the five-item scale of McDonald and Siegall (52) on a 7-point Likert scale from 1 ( strongly disagree ) to 7 ( strongly agree ). The internal consistency of the scale was acceptable (𝛼 Technology self-efficacy = 0.71). Professional identification was measured using the five items from Hekman et al. (61) on a 5-point Likert scale from 1 ( strongly disagree ) to 5 ( strongly agree ). The scale showed acceptable internal consistency (𝛼 Professional identification = 0.77). Control variables Age, gender, and personality were included as control variables based on research showing that all three variables have an impact on technology adoption (32,33,35–37). Participants‘ personality traits were assessed using the Big Five Inventory (68), on a 5-point Likert scale from 1 ( strongly disagree ) to 5 ( strongly agree ), including the main dimensions openness, conscientiousness, extraversion, agreeableness, and neuroticism. Data analysis Data was analyzed using R (Version 4.3.2, R Core Team, 2023). Answers to the open questions were coded using Excel. Qualitative and descriptive analysis First, we conducted a qualitative content analysis to get in-depth insights into mental healthcare practitioners’ understanding of AI-technology for their field of work (RQ1), and allowing for participants’ viewpoints to emerge (69). To gain these insights, we used a deductive thematic analysis (70) to identify how many types of AI applications were mentioned by practitioners. Participants’ responses were clustered into the four predefined application areas and then analyzed for their frequency, to gain insights about the most known and common areas. Further, the precision of their description of AI-enabled technologies in mental healthcare was assessed. We examined whether practitioners could not give a description if the descriptions solely included the technology’s potential area of application or if also the tool’s underlying functions or operational mechanism were explained properly. For answers to the open question regarding the context in which they have heard about the AI technologies (RQ3), an inductive approach (70) was employed to identify recurrent categories within the data. Participants’ responses were coded based on similarities and organized subsequently into themes representing higher-level concepts. All responses were independently coded by two researchers to review and validate the identified themes with subsequent discussion in cases with coding discrepancies. The code book can be found in the online material on OSF (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497). SEM Next, to look at the learning and use intentions, we specified one SEM model for each application area using the ‘lavaan’ package (71). Confirmatory factor analyses (CFA) were calculated for each model. For the model fit, root-mean-square error of approximation (RMSEA) values smaller than 0.05 are considered good and smaller than 0.08 acceptable (72). Standardized root-mean-square residual (SRMR) values up to 0.08 are considered satisfactory (73). Models showing comparative fit index (CFI) and Tucker Lewis index (TLI) values near to or surpassing 0.90 possess a reasonable level of fit (73). For each application area, we analyzed models to predict learning and use intention from the predictor variables and the control variables age, gender, and personality. Further, we calculated three more parsimonious theoretical models to avoid overfitting and ensure the distinctness of the variables. For the first parsimonious model, we combined the subscales of readiness for medical AI. In the second parsimonious model, the subscales of AI anxiety were merged, and in the third parsimonious model, affinity for technology interaction and technology self-efficacy were combined. All in all, SEMs were calculated for one research model per application area with and without control variables, as well as the three more parsimonious models, totaling eleven models. Explorative analysis of demographic and tool differences Finally, for the analysis of potential group differences, we assessed the mean values, standards deviations, and correlations between the variables used in the SEM. Group differences across the four application areas and practitioners’ subgroups (profession, gender, country) were assessed using t-tests or one-way ANOVAs with post-hoc Tukey-HSD. The data was found to be normally distributed following testing for assumptions, with only minor violations observed for learning and use intentions. However, simulation studies demonstrated that, particularly in studies with larger samples, such violations have a negligible impact on the results (74). Additionally, familiarity and use experiences with AI-enabled technologies among mental health practitioners and their context (RQ2-4) were analyzed descriptively. Results Practitioners’ understanding and familiarity with different application areas When participants were asked to explain their understanding of AI-enabled technologies in mental healthcare and how they could be used in their daily work in their own words, 10.5% could not provide a description. Over half of those that provided a description (53.7%) mentioned only one application area, while a further 37.6% stated two categories (RQ1). Merely 8.1% of participants named three areas, whilst only 0.6% of participants (n = 2) listed all four. AI-enabled tools for supporting treatment decisions emerged as the most frequently mentioned area (69.8%), followed by diagnostic (43.4%) and practice management tools (41.1%). Only six participants mentioned feedback tools (1.7%). Participants exhibited varying levels of precision in the description of these technologies, however mostly demonstrating a basic understanding through their explanations. While a majority provided less detailed statements, such as indicating AI’s role as "diagnostic assistance" (Clinical psychologist, 45), a minority offered more elaborate descriptions, exemplified by one professional’s description that “AI could help to make diagnosis […] more efficient and precise by pooling larger data sources together (e.g., interview data, EHR data, patient-reported outcomes, biomarker data)” (Clinical psychologist, 47). For treatment tools, most participants also solely addressed their general purpose, such as “tools that have been programmed to respond to folks in crisis” (Psychotherapist in training, 32). A smaller subset displayed a deeper understanding by mentioning the underlying working mechanism: “By considering an individual's unique history, symptoms, and responses to therapy, AI can recommend specific interventions and strategies tailored to their needs” (Psychiatrist, 69). Professionals mostly described feedback tools briefly as tools that “give input into your performance as a therapist” (Clinical psychologist, 26). Only two participants provided additional information by stating that “there are programs that listen to and transcribe therapy sessions and from this identify themes, relational patterns, and can even rate the therapist on various qualities and suggest interventions” (Clinical psychologist, 35). Likewise, a disparity in the precision level of participants’ responses emerged about practice management tools, ranging from succinct descriptions, such as “documentation of visit” (Psychiatrist, 46) and “can be used to write notes” (Psychotherapist, 34) to more elaborate insights: “I think predictive text could be used for things like notes and that AI software can be used for recording and transcribing sessions, and then generating notes” (Clinical psychologist, 33). Experiences of mental health practitioners with AI-enabled technologies Nearly half of the practitioners ( n = 178, 45.4%) stated that they have never heard of AI-enabled technologies in the field of psychotherapy/psychiatry, while 44.9% ( n = 176) did (RQ2). Figure 2 displays their sources of information. Overall, only 9.7% ( n = 38) actively looked into this topic, whose majority obtained information independently through online research ( n = 29, 76.3%). A further 10.5% ( n = 4) stated that they attended voluntary information sessions and only 13.2% ( n = 5) participated in formal trainings (RQ3). The vast majority of participating practitioners (n = 366, 93.37%) have not used AI-enabled technologies in their clinical practice (RQ4). Learning and use intentions across application areas The data were normally distributed, with mild violations for learning and use intentions. However, simulation studies showed that especially for larger samples as in our study, mild violations have little to no effect on the results The overall learning intention was significantly higher than the overall use intention, t (781) = 8.17, p < 0.001, d = 0.584; M Learning = 3.65, SD Learning = 0.88; M Use = 3.14, SD Use = 0.88). Further, both differed across the four application areas. Practitioners’ intention to learn was significantly higher for AI-enabled management tools ( M = 3.91, SD = 1.01) compared to diagnostic ( M = 3.53, SD = 1.12), treatment ( M = 3.65, SD = 1.09), and feedback tools ( M = 3.53, SD = 1.19; F (3, 1564) = 10.38, p < .001, η p 2 = 0.02; see Figure 3a). Practitioners’ use intentions were significantly higher for AI-enabled tools for feedback ( M = 3.13, SD = 1.22) than diagnosis ( M = 2.78, SD = 1.15) and again, for management tools ( M = 3.70, SD = 1.10) compared to diagnosis, treatment ( M = 2.96, SD = 1.16), and feedback ( F (3, 1564) = 46.2, p < 0.001, η p 2 = 0.08; see Figure 3b). The results indicate that mental health practitioners are more hesitant to learn about and use AI-enabled tools that are more patient-centered compared to more therapist-centered tools that have a less direct influence on decisions that affect patients. Learning and use intentions across different occupational and demographic groups Learning and use intentions differed across occupations, with psychiatrists reporting significantly higher intentions to learn ( F (4, 387) = 4.87, p = 0.002, η p 2 = 0.04) and use AI-enabled technologies compared to psychotherapists in training, psychotherapists, and clinical psychologists ( F (4, 387) = 4.52, p = 0.001, η p 2 = 0.04; see Table A1 in the online appendix). All other differences were non-significant ( p > 0.05). Male practitioners showed higher learning intentions ( t (153.39) = 2.95, p = 0.004, d = 4.17) and use intentions compared to female practitioners ( t (134.73 = 3.02, p = 0.003, d = 3.45; see Table A1 in the online appendix). German practitioners reported significantly lower learning intentions compared to their US counterparts, t (363.55) = -4.03, p 0.05). SEM For all variables used in the SEM models, means, standard deviations, and correlations can be found in Table A2 in the online appendix. Across all four application areas, the complete models showed better fit indices than the parsimonious models, indicating that the model variables were sufficiently distinct (see Table A3 in the online appendix). In all models, one item from the technology self-efficacy scale had standardized factor loadings below 0.40 and was therefore excluded (75). The measurement model of the initially proposed model showed only a partially acceptable fit. Therefore, a second version was calculated, which included the correlated error terms for the two reversed-worded items of the ATI scale. Correlating the measurement errors did not significantly alter the parameter estimates of the underlying measurement model. Table 6 shows the fit indices for each of the final models. The model fit indices for RMSEA (≤ 0.056) and SRMR (≤ 0.063) are acceptable to good. The CFI and TFI close to 0.9 are considered marginal levels (76). As the cutoff-levels for the goodness-of-fit indices depend on model characteristics, such as the sample size and number of variables (77), the complexity of the model and rather small sample size might be the reasons for the CFI and TLI just below the threshold (78). Controlling for age, gender, and personality did not substantially affect the models for treatment and feedback tools. For the diagnostic tool, the association between professional identification and learning intention, and for practice management tools, the association between cognitive readiness and learning intention were no longer significant (see Table A4 – A7 in the online appendix). The results of the final SEM models are presented in Tables 6 - 9. All significant paths are highlighted in Figure 5. Across the four models, predictor variables accounted for 46.7 – 61.0% of the variance in learning intention and 8.1 – 17.0% in use intentions. Across all application areas, the intention to learn about AI-enabled technologies was positively associated with the intention to use these technologies, supporting H1 for each model. Some paths for the subconstructs of medical artificial intelligence readiness, AI anxiety, beliefs about technological capabilities and professional identity were also relevant across all application areas, however, others differed for each application area (see Table 6-9). Regarding AI knowledge, cognitive readiness (H2a) was positively associated with the learning intention of the feedback tool, vision readiness (H2b) with the learning intention of the feedback tool, and ethical readiness (H3c) with the use intention across each application area. For the automatic motivational factor AI anxiety, sociotechnical blindness (H2e) demonstrated a positive relationship with the learning intentions of the treatment and practice management tool. For reflective motivational factors, technology self-efficacy (H3g) was negatively related to the use intentions for the diagnostic, treatment, and practice management tool. Further, practitioners’ affinity for technology interaction showed a consistent positive link with the use intentions for all application areas, supporting H3h for each model. Lastly, professional identification (RQ5) was positively associated with the learning intention for the diagnostic, treatment, and feedback tool. Discussion Amidst the increasing integration of AI-enabled technologies in healthcare, the present study investigated mental health practitioners’ understanding and familiarity across different application areas for AI-enabled support tools in mental healthcare. Additionally, we examined factors influencing the intention to learn and use AI-enabled technologies across the different areas. Current familiarity gaps among mental healthcare professionals Our study reveals a limited understanding of AI-enabled technologies and significant gap in mental health practitioners’ familiarity with AI-enabled tools for mental health, with nearly half of the surveyed practitioners unaware of these technologies. This low familiarity indicates that many professionals are not informed about the development and potential clinical applications of AI in mental healthcare. Additionally, practitioners primarily gained information through mainstream media such as social media or newspaper articles and less than one-tenth of practitioners who had heard about AI technologies received formal education on the topic, a trend consistent with prior research ( 79 ). Furthermore, the present findings align with an international survey of psychiatrists, which found that less than a quarter had received formal technology training ( 80 ). Adding to the literature, the fact that the majority of our participants were psychotherapists currently enrolled in training suggests that current training programs may not adequately cover AI-related topics, thereby limiting practitioners' exposure and understanding. As a lack of training and instructions on technology use in healthcare further contributes to an unsafe work environment and medical errors ( 27 ), the results underline the need of adjusting the training to emerging technologies. Professionals’ varying adoption intentions and application-specific hesitation The surveyed practitioners were more inclined towards learning rather than actively using AI-enabled technologies in their clinical practice. This supports existing literature indicating that learning and use intentions represent different levels of engagement with technology ( 29 , 30 ). For the more practical level of intending to use technologies, practitioners’ main concerns regarding AI technologies, including the lack of transparency of model predictions, data privacy, cyber security, and patient safety ( 45 ), might have contributed to their greater use hesitation. Besides, awareness of the need to inform patients about the use of AI technologies in psychotherapeutic decisions and obtain their consent ( 81 , 82 ), along with understanding how these issues affect their work and patients, might contribute to lower usage intentions. Moreover, participants demonstrated different levels of willingness to engage with AI-enabled technologies across the application areas. Notably, they were less hesitant towards clinician-centered feedback or practice management tools compared to patient-centered tools, aligning with previous findings ( 21 , 83 ). This may be attributed to the higher stakes associated with using technology to inform diagnosis or treatment decisions compared to receiving feedback or administrative support as diagnostic or treatment errors can have severe negative consequences, potentially resulting in wrong or delayed treatment and a worse prognosis ( 84 , 85 ). Additionally, our results revealed profession-specific differences, with psychiatrists demonstrating higher learning and use intentions compared to psychotherapists and clinical psychologists. This difference might stem from the specific characteristics of education and work in each occupation. Psychiatrists undergo medical training that already integrates AI-enabled technologies into the curricula, albeit with a focus on other specialties ( 86 ). However, their greater exposure to clinical technologies and closer connection to the broader medical field, where AI use is more prevalent than in psychology, might contribute to their higher adoption intentions. Additionally, since medical prescription are part of psychiatrists’ daily tasks and this area holds widespread potential for AI utilization (for instance see ( 87 )), it might be more natural for them to envision using AI into their practice. The practices of psychotherapists and clinical psychologists in turn are centered more around interpersonal treatment and the patient-therapist relationship ( 88 ). In this context, technology is often perceived not as a substitute for human care ( 83 ), hence, it may be challenging for psychotherapists to envision the integration of AI technology into their professional practice, possibly leading to their greater hesitation. Individual-level predictors of AI adoption intentions We found a robust association between the intention to learn and use AI-enabled technologies across all application areas. This aligns with results showing that the willingness to engage in training enhances professionals’ intention to use AI technologies ( 28 ). Consequently, willingness to learn is a pivotal initial step in engaging with AI technologies and understanding the predictors for both learning and use intention is important. First, regarding AI knowledge , the domain ethical readiness emerged as a significant predictor for use intentions across all application areas, making it a driving force for the intention to use AI-enabled technologies in healthcare. This is in line with research showing that AI ethics awareness was positively correlated with the use intention of AI-based technology in nursing care ( 51 ). The consistent link across all application areas may be explained by the high value of ethics in mental health. Besides general medical ethics, it encompasses elements such as the emotional therapist-patient relationship and handling highly sensitive information, requiring strict adherence to ethical standards ( 89 ). However, learning intentions were influenced differently depending on the application area. On the one hand, the ability to anticipate the technology’s potential impact, involving a deeper understanding of the technologies’ strengths and weaknesses (vision readiness), was positively associated with the intention to learn about treatment support tools. As practitioner were most familiar with treatment tools, it is not surprising that practitioners with a more nuanced understanding are more likely to deepen their knowledge in tools they are already familiar with, likely aiming to refine their knowledge. On the other hand, the basic understanding about AI technologies (cognitive readiness) was positively associated with the intention to learn about feedback tools which practitioners were least familiar with. Practitioners with a basic understanding are therefore eager to explore less familiar tools, potentially driven by curiosity and a desire to broaden their knowledge. Hence, the findings suggest that learning intentions vary based on different facets of practitioners’ AI knowledge, with a basic knowledge leading to a higher intention to learn about new tools and advanced knowledge driving deeper exploration of known tools. These study findings on AI knowledge might help to understand the mixed results found in prior literature which showed a positive association with general AI knowledge in some cases ( 21 , 43 ), but not in others ( 45 ); while the present study shows that different facets of AI knowledge have varying influences on the adoption intentions for different tools. Second, none of the subdimensions of AI anxiety was associated with use intentions for any application area, contrary to prior findings indicating that AI anxiety impedes AI adoption ( 21 , 50 , 51 ). However, previous research concentrated on general AI anxiety, without specifically addressing its nuanced facets ( 50 , 51 , 21 ). For instance, looking at the subdimension of job replacement anxiety, the only moderate levels reported by our participants (see Table A2) might have contributed to this result, indicating that they do not view AI as a threat to their profession. This finding aligns with research indicating that only 4% of psychiatrists believe that future technology will make their jobs obsolete ( 18 , 83 ). However, anxiety arising from the belief that AI systems operate without human supervision (sociotechnical blindness) was positively associated with the intention to learn about two AI-enabled application areas: treatment and practice management tools. Contrary to high levels of anxiety, moderate anxiety, as in our study, can have a positive effect on the learning motivation ( 90 ) and this might explain the effect in the opposite direction. The effect might have emerged particularly for these two areas, as they are the ones practitioners are most eager to learn about and, in the case of practice management tools, intend to use. Given the pivotal role of human oversight in successfully implementing AI technology, which requires a certain level of tool understanding to monitor its actions and decisions ( 91 – 93 ), practitioners may be more inclined to learn about AI technologies they see themselves engaging with, aiming to equip themselves for ensuring proper oversight if needed. Third, reflective motivational processes played a pivotal role in both learning and use intentions. Across three application areas (diagnostic, treatment, and practice management), professionals’ technology self-efficacy was negatively associated with the intention to use diagnostic, treatment, and practice management tools. However, we found a significant positive correlation between technology self-efficacy and the overall use intention (see Table A2). This discrepancy suggests a suppression effect within the models. This effect occurs when there are multiple predictors in the model, and the overall predictive power of the model is improved by the inclusion of additional predictors that uncover different associations compared to when solely considering technology self-efficacy ( 94 ). Consequently, the association between technology self-efficacy and the use intention is hard to interpret. However, the suppression effect indicates that while technology self-efficacy is negatively associated with the use intention for some application areas, its overall positive correlation with the intention to use suggests that practitioners with higher beliefs in their ability to effectively perform technologically advanced tasks are more inclined to use AI-enabled technologies, which aligns with existing literature ( 51 , 53 – 56 ). Fourth, affinity for technology interaction , characterized by the enjoyment and comfort in interacting with technology, showed a positive relationship with the use intention for each tool category. This result was expected based on research from broader hospital settings and other medical domains demonstrating this positive association ( 59 , 60 ). From a behavioral perspective, cross-situational consistency may explain this finding as people often maintain behavior across similar contexts ( 95 ). One’s overall positive perception in interacting with technologies might therefore be also transferable to their engagement with technologies at work. Finally, a strong professional identity exhibited a positive association with intentions to learn about three application areas (diagnostic, treatment, and feedback). The non-significant association with the use of learning intention for the practice management tools may relate to the fact that practitioners do not see administrative tasks as closely related to their identity as mental healthcare professionals. The positive association contributes to existing literature by extending prior insights from general healthcare contexts into mental healthcare ( 63 , 65 , 96 ). Professional identity is a dynamic concept shaped by various factors, including technology implementation ( 97 , 98 ), and prompting (professionals like) mental healthcare worker to continually assess alignment with evolving work contexts ( 99 , 100 ). Despite limited awareness of these technologies, strong identification with their mental health role might motivate them to learn about technologies, facilitating adaption to workplace changes and alignment with their professional identity. Limitations and future research Several limitations should be considered when interpreting the findings of this study. First, the brevity of responses to the open-ended questions may stem from a lack of motivation or time constraints. It is plausible that practitioners possess a more extensive understanding than was conveyed within their response. Future studies could encourage participants to elaborate, for instance by follow-up interviews designed to gather more information on their understanding or by using more objective measures. Second, the inclusion of control variables resulted in the non-significance of vision readiness and professional identification on learning intentions in two models. This, together with the suppression effect on self-efficacy, underscores the complexity of the predictors’ associations and highlights the need for further exploration to understand the nuanced interplay of variables influencing the learning intentions of AI-enabled technologies. Third, no causal relationships could be observed and tested as the present study was cross-sectional. In the future, longitudinal and experimental designs should be employed. Lastly, participants only got concise descriptions of the different AI application areas the AI tools without the opportunity for direct practical interaction with the technologies. This might have restricted participants’ depth of understanding and influenced their responses. Future research should explore using detailed, comprehensive, and interactive representations of AI decision-making processes and technologies ( 101 , 102 ). Practical implications The fact that half of the practitioners have not heard about AI-enabled technology in mental healthcare demonstrates the need for formal education on this topic. The integration of modules on AI-enabled technologies into curricula and professional training programs holds the potential to redirect professional educational frameworks towards future-oriented challenges like technology interaction. Better training regarding the use of technology might prevent medical errors, as research has shown that healthcare practitioners view a lack of technology training as a major cause of errors ( 27 ). Taking it a step further, our study results can also contribute to the development of successful educational frameworks. For instance, ethical knowledge seemed highly relevant for use intentions, hence, education on ethical standards required for technology use is one starting point to ensure their safe and responsible use. As highlighted by Katznelson and Gercke ( 103 ), incorporating AI ethics into healthcare training programs is crucial to prepare healthcare professionals for the ethical complexities accompanying AI implementation. Additionally, since affinity for technology interaction was consistently associated with use intentions, the comfort of interacting with technology should also be fostered via practical experiences and on-the-job training. Moreover, addressing hesitations early on or helping users overcome them could involve considering predictors not only in the design of training programs but also the technology itself. One potential solution could involve ensuring more actively that the technology utilizes health data in accordance with legal and ethical norms. Although regulations such as the MDR (Medical Device Regulation) and AIA (Artificial Intelligence Act) are already in place ( 104 ), transparently displaying the underlying norms to end users can simultaneously advance their ethical knowledge and ensure adherence to ethical principles. With this, developers can better serve practitioners’ needs and facilitate their adoption of AI technologies in mental healthcare. Conclusion Our study reveals a substantial gap in mental healthcare professionals’ familiarity of AI-enabled technologies in their field. It further underscores the nuanced perception of the different application areas, emphasizing the necessity to consider not only the specific AI application area but also the characteristics of different mental health professionals during the implementation process. Recognizing the pivotal role of learning in initiating engagement, our study suggests that cultivating such engagement via tailored training programs considering robust factors like individuals’ ethical knowledge and affinity for technology interaction could subsequently enhance professionals’ inclination towards utilizing these novel technologies. Moving forward, addressing important factors for each application area will be crucial for the safe integration of AI technologies into mental healthcare practices. Doing so will help bridge the gap between the increasing demand for mental healthcare and limited available therapeutic resources, ultimately improving the accessibility and effectiveness of mental health services. Abbreviations AI (Artificial Intelligence) AIA (Artificial Intelligence Act) CFA (Confirmatory factor analyses) ANOVA (Analysis of Variance) CFI (Comparative Fit Index) H (Hypothesis) MDR (Medical Device Regulation) RMSEA (Root-Mean-Square Error of Approximation) RQ (Research Question) SEM (Structural Equation Modeling) SRMR (Standardized Root-Mean-Square Residual) TLI (Tucker Lewis Index) Declarations Ethics approval and consent to participate The protocol for this study was approved by the Ethics Committee of the University of Regensburg (23-3365-101). Consent for publication Not applicable. Availability of data and materials Additional supporting information can be found in the online appendix and on OSF (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497). Declaration of Conflicting Interests The authors declare that they have no conflict of interest. Funding The research was funded by a grant from the Volkswagen Foundation (Grant #: 98525). Authors’ Contributions: J.C.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project Administration, Visualization, Writing – Original Draft A.-K. K.: Conceptualization, Methodology, Supervision, Writing – Review & Editing E.L.: Funding acquisition, Writing – Review & Editing S.G.: Conceptualization, Methodology, Funding acquisition, Supervision, Writing – Review & Editing Acknowledgements We thank Anna Sigl for her help in the qualitative data analysis References World Health Organization. World Mental Health Report.: Transforming mental health for all [Internet]. www.who.int. World Health Organization, 2022. Available from: https://www.who.int/publications/i/item/9789240049338 Minerva F, Giubilini A. Is AI the Future of Mental Healthcare? Topoi [Internet]. 2023 May 31 [cited 2023 Jun 15]; Available from: https://doi.org/10.1007/s11245-023-09932-3 Kellogg KC, Sadeh-Sharvit S. Pragmatic AI-augmentation in mental healthcare: key technologies, potential benefits, and real-world challenges and solutions for frontline clinicians. Front Psychiatry. 2022;13:990370. Blease C, Locher C, Leon-Carlyle M, Doraiswamy M. Artificial intelligence and the future of psychiatry: qualitative findings from a global physician survey. Digit Health. 2020;6:2055207620968355. Kleine AK, Lermer E, Cecil J, Heinrich A, Gaube S. Advancing mental health care with AI-enabled precision psychiatry tools: A patent review. Comput Hum Behav Rep. 2023 Dec;12:100322. Allesøe RL, Thompson WK, Bybjerg-Grauholm J, Hougaard DM, Nordentoft M, Werge T, et al. Deep Learning for Cross-Diagnostic Prediction of Mental Disorder Diagnosis and Prognosis Using Danish Nationwide Register and Genetic Data. JAMA Psychiatry. 2023 Feb 1;80(2):146–55. Kellogg KC, Sadeh-Sharvit S. Pragmatic AI-augmentation in mental healthcare: Key technologies, potential benefits, and real-world challenges and solutions for frontline clinicians. Front Psychiatry. 2022 Sep 6, 13. Yadav U, Sharma AK. A novel automated depression detection technique using text transcript. Int J Imaging Syst Technol. 2023;33(1):108–22. Lee EE, Torous J, De Choudhury M, Depp CA, Graham SA, Kim HC, et al. Artificial Intelligence for Mental Health Care: Clinical Applications, Barriers, Facilitators, and Artificial Wisdom. Biol Psychiatry Cogn Neurosci Neuroimaging. 2021 Sep;6(9):856–64. Aafjes-van Doorn K, Kamsteeg C, Bate J, Aafjes M. A scoping review of machine learning in psychotherapy research. Psychother Res. 2021;31(1):92–116. Lin E, Lin CH, Lane HY. Precision psychiatry applications with pharmacogenomics: artificial intelligence and machine learning approaches. Int J Mol Sci. 2020;21(3):969. Cummins R, Ewbank MP, Martin A, Tablan V, Catarino A, Blackwell AD. TIM: a tool for gaining insights into psychotherapy. In: The World Wide Web Conference. 2019. p. 3503–6. Flemotomos N, Martinez VR, Chen Z, Singla K, Ardulov V, Peri R, et al. Automated evaluation of psychotherapy skills using speech and language technologies. Behav Res Methods. 2022;54(2):690–711. Hirsch T, Soma C, Merced K, Kuo P, Dembe A, Caperton DD, et al. “ It’s hard to argue with a computer” Investigating Psychotherapists’ Attitudes towards Automated Evaluation. In: Proceedings of the 2018 Designing Interactive Systems Conference. 2018. p. 559–71. Imel ZE, Pace BT, Soma CS, Tanana M, Hirsch T, Gibson J, et al. Design feasibility of an automated, machine-learning based feedback system for motivational interviewing. Psychotherapy. 2019;56(2):318. Sadeh-Sharvit S, Camp TD, Horton SE, Hefner JD, Berry JM, Grossman E, et al. Effects of an Artificial Intelligence Platform for Behavioral Interventions on Depression and Anxiety Symptoms: Randomized Clinical Trial. J Med Internet Res. 2023 Jul 10;25(1):e46781. Dawoodbhoy FM, Delaney J, Cecula P, Yu J, Peacock I, Tan J, et al. AI in patient flow: applications of artificial intelligence to improve patient flow in NHS acute mental health inpatient units. Heliyon. 2021 May;7(5):e06993. Blease C, Locher C, Leon-Carlyle M, Doraiswamy M. Artificial intelligence and the future of psychiatry: Qualitative findings from a global physician survey. Digit Health. 2020 Jan 1;6:2055207620968355. Chen ZS, Prathamesh, Kulkarni, Galatzer-Levy IR, Bigio B, Nasca C, et al. Modern Views of Machine Learning for Precision Psychiatry . Patterns. 22 Nov 11;3(11):100602. Rajpurkar P, Chen E, Banerjee O, Topol EJ. AI in health and medicine. Nat Med. 2022 Jan 20;28(1):31–8. Kleine AK, Kokje E, Lermer E, Gaube S. Attitudes Toward the Adoption of 2 Artificial Intelligence–Enabled Mental Health Tools Among Prospective Psychotherapists: Cross-sectional Study. JMIR Hum Factors. 2023 Jul 12;10:e46859. Aafjes-van Doorn K, Kamsteeg C, Bate J, Aafjes M. A scoping review of machine learning in psychotherapy research. Psychother Res. 2021 Jan 2;31(1):92–116. Chekroud AM, Bondar J, Delgadillo J, Doherty G, Wasil A, Fokkema M, et al. The promise of machine learning in predicting treatment outcomes in psychiatry. World Psychiatry. 2021;20(2):154–70. Hummelsberger P, Koch TK, Rauh S, Dorn J, Lermer E, Raue M, et al. Insights on the Current State and Future Outlook of AI in Health Care: Expert Interview Study. JMIR AI. 2023 Oct 31;2:e47353. Pang TY, Lee TK, Murshed M. Towards a New Paradigm for Digital Health Training and Education in Australia: Exploring the Implication of the Fifth Industrial Revolution. Appl Sci. 2023 Jan;13(11):6854. Sharma A, Lin IW, Miner AS, Atkins DC, Althoff T. Human-AI Collaboration Enables More Empathic Conversations in Text-based Peer-to-Peer Mental Health Support. Nat. Mach. Intell. 2023 Jan 23;5(1):46-57. Gaube S, Cecil J, Wagner S, Schicho A. The relationship between health IT characteristics and organizational variables among German healthcare workers. Sci Rep. 2021 Sep 7;11(1):17752. Choudhury A, Asan O. Impact of accountability, training, and human factors on the use of artificial intelligence in healthcare: Exploring the perceptions of healthcare practitioners in the US. Hum Factors Healthc. 2022 Dec 1;2:100021. Lo Presti A, De Rosa A, Viceconte E. I want to learn more! Integrating technology acceptance and task–technology fit models for predicting behavioural and future learning intentions. J Workplace Learn. 2021 Jun 4; 33(8): 591-605. Holden RJ, Karsh BT. The Technology Acceptance Model: Its past and its future in health care. J Biomed Inform. 2010 Feb;43(1):159–72. Venkatesh, Morris, Davis, Davis. User Acceptance of Information Technology: Toward a Unified View. MIS Q. 2003;27(3):425. Knop M, Weber S, Mueller M, Niehaves B. Human Factors and Technological Characteristics Influencing the Interaction of Medical Professionals With Artificial Intelligence–Enabled Clinical Decision Support Systems: Literature Review. JMIR Hum Factors. 2022 Mar 24;9(1):e28639. Felmingham CM, Adler NR, Ge Z, Morton RL, Janda M, Mar VJ. The Importance of Incorporating Human Factors in the Design and Implementation of Artificial Intelligence for Skin Cancer Diagnosis in the Real World. Am J Clin Dermatol. 2021 Mar 1;22(2):233–42. Greenhalgh T, Wherton J, Papoutsi C, Lynch J, Hughes G, A’Court C, et al. Beyond Adoption: A New Framework for Theorizing and Evaluating Nonadoption, Abandonment, and Challenges to the Scale-Up, Spread, and Sustainability of Health and Care Technologies. J Med Internet Res. 2017 Nov 1;19(11):e367. Zhang X, Guo X, Lai K hung, Guo F, Li C. Understanding Gender Differences in m-Health Adoption: A Modified Theory of Reasoned Action Model. Telemed J E-Health 2013 Oct 26;20(1):39-46. Liang Y, Lee SA. Fear of Autonomous Robots and Artificial Intelligence: Evidence from National Representative Data with Probability Sampling. Int J Soc Robot. 2017 Jun 1;9(3):379–84. Kaya F, Aydin F, Schepman A, Rodway P, Yetişensoy O, Demir Kaya M. The Roles of Personality Traits, AI Anxiety, and Demographic Factors in Attitudes toward Artificial Intelligence. Int J Human–Computer Interact. 2022 Dec 7;1–18. Eitle V, Buxmann P. Cultural Differences in Machine Learning Adoption: An International Comparison between Germany and the United States. In:European Conference on Information Systems (ECIS). 2020. Available from: https://aisel.aisnet.org/ecis2020_rp/138/ Sindermann C, Sha P, Zhou M, Wernicke J, Schmitt HS, Li M, et al. Assessing the Attitude Towards Artificial Intelligence: Introduction of a Short Measure in German, Chinese, and English Language. KI - Künstl Intell. 2021 Mar 1;35(1):109–18. Michie S, van Stralen MM, West R. The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implement Sci. 2011 Apr 23;6(1):42. Nickbakht M, Meyer C, Scarinci N, Beswick R. Exploring factors influencing the use of an eHealth intervention for families of children with hearing loss: An application of the COM-B model. Disabil Health J. 2020 Oct;13(4):100921. Park J, Woo SE. Who Likes Artificial Intelligence? Personality Predictors of Attitudes toward Artificial Intelligence. J Psychol. 2022 Jan 2;156(1):68–94. Wagner G, Raymond L, Paré G. Understanding Prospective Physicians’ Intention to Use Artificial Intelligence in Their Future Medical Practice: Configurational Analysis. JMIR Med Educ. 2023 Mar 22;9:e45631. Ter Harmsel JF, Smulders LM, Noordzij ML, Swinkels LTA, Goudriaan AE, Popma A, et al. Forensic Psychiatric Outpatients’ and Therapists’ Perspectives on a Wearable Biocueing App (Sense-IT) as an Addition to Aggression Regulation Therapy: Qualitative Focus Group and Interview Study. JMIR Form Res. 2023 Feb 1;7:e40237. Li X, Jiang MY chao, Jong MS yung, Zhang X, Chai C sing. Understanding Medical Students’ Perceptions of and Behavioral Intentions toward Learning Artificial Intelligence: A Survey Study. Int J Environ Res Public Health. 2022 Jul 18;19(14):8733. Karaca O, Çalışkan SA, Demir K. Medical artificial intelligence readiness scale for medical students (MAIRS-MS) – development, validity and reliability study. BMC Med Educ. 2021 Feb 18;21(1):112. Wang YM, Wei CL, Lin HH, Wang SC, Wang YS. What drives students’ AI learning behavior: a perspective of AI anxiety. Interact Learn Environ. 2022 Dec 11;1–17. Johnson DG, Verdicchio M. AI Anxiety. J Assoc Inf Sci Technol. 2017 Sep;68(9):2267–70. Wang YY, Wang YS. Development and validation of an artificial intelligence anxiety scale: an initial application in predicting motivated learning behavior. Interact Learn Environ. 2022 Apr 3;30(4):619–34. Almarzouqi A, Aburayya A, Salloum SA. Determinants predicting the electronic medical record adoption in healthcare: A SEM-Artificial Neural Network approach. PLoS ONE. 2022 Aug 16;17(8):e0272735. Kwak Y, Seo YH, Ahn JW. Nursing students’ intent to use AI-based healthcare technology: Path analysis using the unified theory of acceptance and use of technology. Nurse Educ Today. 2022 Dec 1;119:105541. McDonald T, Siegall M. The Effects of Technological Self-Efficacy and Job Focus on Job Performance, Attitudes, and Withdrawal Behaviors. J Psychol. 1992 Sep;126(5):465–75. AlQudah AA, Al-Emran M, Shaalan K. Technology Acceptance in Healthcare: A Systematic Review. Appl Sci. 2021 Jan;11(22):10537. Ngusie HS, Kassie SY, Chereka AA, Enyew EB. Healthcare providers’ readiness for electronic health record adoption: a cross-sectional study during pre-implementation phase. BMC Health Serv Res. 2022 Mar 2;22(1):282. Pan M, Gao W. Determinants of the behavioral intention to use a mobile nursing application by nurses in China. BMC Health Serv Res. 2021 Mar 12;21(1):228. Zha H, Liu K, Tang T, Yin YH, Dou B, Jiang L, et al. Acceptance of clinical decision support system to prevent venous thromboembolism among nurses: an extension of the UTAUT model. BMC Med Inform Decis Mak. 2022 Aug 19;22(1):221. Franke T, Attig C, Wessel D. A personal resource for technology interaction: Development and validation of the Affinity for Technology Interaction (ATI) scale. 2019. Wensing M, Paech B, Roth C, Schwill S. Learning, understanding and the use of information technology: a survey study among primary care physician trainees. BMC Health Serv Res. 2019 Oct 22;19(1):728. Labinsky H, Ukalovic D, Hartmann F, Runft V, Wichmann A, Jakubcik J, et al. An AI-Powered Clinical Decision Support System to Predict Flares in Rheumatoid Arthritis: A Pilot Study. Diagnostics. 2023 Jan;13(1):148. Maassen O, Fritsch S, Gantner J, Deffge S, Kunze J, Marx G, et al. Future Mobile Device Usage, Requirements, and Expectations of Physicians in German University Hospitals: Web-Based Survey. J Med Internet Res. 2020 Dec 21;22(12):e23955. Hekman DR, Steensma HK, Bigley GA, Hereford JF. Effects of organizational and professional identification on the relationship between administrators’ social influence and professional employees’ adoption of new work behavior. J Appl Psychol. 2009;94(5):1325–35. Gillan C, Milne E, Harnett N, Purdie TG, Jaffray DA, Hodges B. Professional implications of introducing artificial intelligence in healthcare: an evaluation using radiation medicine as a testing ground. J Radiother Pract. 2019 Mar;18(1):5–9. Liu Q, Geertshuis S. Professional identity and teachers’ learning technology adoption: a review of adopter-related antecedents. [Internet]. Available from: https://2016conference.ascilite.org/wp-content/uploads/ascilite2016_liu_full.pdf Jussupow E, Spohrer K, Heinzl A. Identity Threats as a Reason for Resistance to Artificial Intelligence: Survey Study With Medical Students and Professionals. JMIR Form Res. 2022 Mar 23;6(3):e28750. Greenhalgh T, Abimbola S. The NASSS Framework – A Synthesis of Multiple Theories of Technology Implementation. Stud Health Technol Inform. 2019 Jul 30;263:193-204. Karaca O, Çalışkan SA, Demir K. Medical artificial intelligence readiness scale for medical students (MAIRS-MS) – development, validity and reliability study. BMC Med Educ. 2021 Dec;21(1):112. Wessel D, Attig C, Franke T. ATI-S - An Ultra-Short Scale for Assessing Affinity for Technology Interaction in User Studies. In: Proceedings of Mensch und Computer 2019 [Internet]. Hamburg Germany: ACM; 2019 [cited 2023 May 19]. p. 147–54. Available from: https://dl.acm.org/doi/10.1145/3340764.3340766 Rammstedt B, Kemper CJ, Klein MC, Beierlein C, Kovaleva A. Eine kurze Skala zur Messung der fünf Dimensionen der Persönlichkeit: big-five-inventory-10 (BFI-10) [A short scale to measure the five dimensions of personality: big-five-inventory-10 (BFI-10)]. Methoden, Daten, Analysen (mda). 2013;7(2) Sandelowski M. What’s in a name? Qualitative description revisited. Res Nurs Health. 2010 Feb;33(1):77–84. Fereday J, Muir-Cochrane E. Demonstrating rigor using thematic analysis: A hybrid approach of inductive and deductive coding and theme development. Int J Qual Methods. 2006 Mar 1;5(1):80–92. Rosseel Y. lavaan: An R Package for Structural Equation Modeling. J. Stat. Soft. [Internet]. 2012 May24 [cited 2024 Jul 3]. (J. Stat. Soft.). Available from: https://www.jstatsoft.org/index.php/jss/article/view/v048i02 Browne MW, Cudeck R. Alternative Ways of Assessing Model Fit. Sociol Methods Res. 1992;21(2):230–58. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equ Model Multidiscip J. 1999 Jan;6(1):1–55. Lumley T, Diehr P, Emerson S, Chen L. The Importance of the Normality Assumption in Large Public Health Data Sets. Annu Rev Public Health. 2002 May;23(1):151–69. Cheung GW, Cooper-Thomas HD, Lau RS, Wang LC. Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations. Asia Pac J Manag. 2023 Jan 30;1(1). Hair JF, Anderson R, Tatham R, Black W. Multivariate Data Analysis. 5 th Edition, Prentice Hall; 1998. Hair JF, Black WC, Babin BJ. Multivariate Data Analysis: A Global Perspective. 7 th Edition, Pearson Education; 2010. Kim H, Ku B, Kim JY, Park YJ, Park YB. Confirmatory and Exploratory Factor Analysis for Validating the Phlegm Pattern Questionnaire for Healthy Subjects. Evid-Based Complement Altern Med ECAM. 2016;2016:2696019. Pucchio A, Rathagirishnan R, Caton N, Gariscsak PJ, Del Papa J, Nabhen JJ, et al. Exploration of exposure to artificial intelligence in undergraduate medical education: a Canadian cross-sectional mixed-methods study. BMC Med Educ. 2022 Nov 28;22(1):815. Bauer R, Glenn T, Monteith S, Whybrow PC, Bauer M. Survey of psychiatrist use of digital technology in clinical practice. Int J Bipolar Disord. 2020 Oct 3;8(1):29. Fiske A, Henningsen P, Buyx A. Your Robot Therapist Will See You Now: Ethical Implications of Embodied Artificial Intelligence in Psychiatry, Psychology, and Psychotherapy. J Med Internet Res. 2019 May 9;21(5):e13216. Gerger H, Nascimento AF, Locher C, Gaab J, Trachsel M. What are the Key Characteristics of a ‘Good’ Psychotherapy? Calling for Ethical Patient Involvement. Front Psychiatry. 2020 Jun 2;11:406. Doraiswamy PM, Blease C, Bodner K. Artificial intelligence and the future of psychiatry: Insights from a global physician survey. Artif Intell Med. 2020 Jan 1;102:101753. Kisely S, Scott A, Denney J, Simon G. Duration of untreated symptoms in common mental disorders: association with outcomes: International study. Br J Psychiatry. 2006 Jul;189(1):79–80. Marshall M, Lewis S, Lockwood A, Drake R, Jones P, Croudace T. Association Between Duration of Untreated Psychosis and Outcome in Cohorts of First-Episode Patients: A Systematic Review. Arch Gen Psychiatry. 2005 Oct 1;62:975–83. Sun L, Yin C, Xu Q, Zhao W. Artificial intelligence for healthcare and medical education: a systematic review. Am J Transl Res. 2023 Jul 15;15(7):4820–8. Lin E, Lin CH, Lane HY. Precision Psychiatry Applications with Pharmacogenomics: Artificial Intelligence and Machine Learning Approaches. Int J Mol Sci. 2020 Feb 1;21(3):969. Robiner WN. The mental health professions: Workforce supply and demand, issues, and challenges. Clin Psychol Rev. 2006 Sep;26(5):600–25. Sidhu N, Srinivasraghavan J. Ethics and Medical Practice: Why Psychiatry is Unique. Indian J Psychiatry. 2016 Dec;58(Suppl 2):S199–202. Ambrosi-Randić N, Ružić H. Motivation and learning strategies in university courses in italian language. Metod Obz Horiz. 2010 Nov 15;5(2):41–50. Alowais SA, Alghamdi SS, Alsuhebany N, Alqahtani T, Alshaya AI, Almohareb SN, et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ. 2023 Sep 22;23(1):689. Young AT, Amara D, Bhattacharya A, Wei ML. Patient and general public attitudes towards clinical artificial intelligence: a mixed methods systematic review. Lancet Digit Health. 2021 Sep;3(9):e599–611. Vo V, Chen G, Aquino YSJ, Carter SM, Do QN, Woode ME. Multi-stakeholder preferences for the use of artificial intelligence in healthcare: A systematic review and thematic analysis. Soc Sci Med. 2023 Dec 1;338:116357. Pandey S, Elliott W. Suppressor Variables in Social Work Research: Ways to Identify in Multiple Regression Models. J Soc Soc Work Res. 2010 Jan;1(1):28–40. Sherman RA, Nave CS, Funder DC. Situational similarity and personality predict behavioral consistency. J Pers Soc Psychol. 2010 Aug;99(2):330–43. Cornelissen L, Egher C, Van Beek V, Williamson L, Hommes D. The Drivers of Acceptance of Artificial Intelligence–Powered Care Pathways Among Medical Professionals: Web-Based Survey Study. JMIR Form Res. 2022 Jun 21;6(6):e33368. Fraser-Arnott MA. Evolving practices and professional identity: How the new ways we work can reshape us as professionals and a profession. IFLA J. 2019 Jun 1;45(2):114–26. Johnson M, Cowin L s., Wilson I, Young H. Professional identity and nursing: contemporary theoretical developments and future research challenges. Int Nurs Rev. 2012;59(4):562–9. Kira M, Balkin DB. Interactions between work and identities: Thriving, withering, or redefining the self? Hum Resour Manag Rev. 2014 Jun 1;24(2):131–43. Schubert S, Buus N, Monrouxe LV, Hunt C. The development of professional identity in clinical psychologists: A scoping review. Med Educ. 2023;57(7):612–26. Koutsouleris N, Hauser TU, Skvortsova V, De Choudhury M. From promise to practice: towards the realisation of AI-informed mental health care. Lancet Digit Health. 2022 Nov;4(11):e829–40. Monteith S, Glenn T, Geddes J, Whybrow PC, Achtyes E, Bauer M. Expectations for Artificial Intelligence (AI) in Psychiatry. Curr Psychiatry Rep. 2022 Nov;24(11):709–21. Katznelson G, Gerke S. The need for health AI ethics in medical school education. Adv Health Sci Educ. 2021 Oct 1;26(4):1447–58. Bretthauer M, Gerke S, Hassan C, Ahmad OF, Mori Y. The New European Medical Device Regulation: Balancing Innovation and Patient Safety. Ann Intern Med. 2023 Jun 20;176(6):844–8. Tables Tables 1–2 and 6–9 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files OnlineappendixMentalhealthpractitioners.docx Tables.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4692251","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":324675395,"identity":"f7c4e82c-dcff-4def-a3ee-c40f77a2c181","order_by":0,"name":"Julia Cecil","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9ElEQVRIie3RsQrCMBAG4AuBTtFuIlT0FSKOgs9iKejiUHARFI0U4iK4VvAh9A1SAk6+gYvuXcTFoYhJpeBidBTMP+TuIB8XCICNzQ8GMcxE3mE8P+la1gcxElQQFFFdnU9Eo6Lh1a8Ijv25uEGn7i4QH4eZBKdyoJCODDtinyVLCFpVifhxzRXxBhRtDmYiCAh/ixUpsSfBJf5hSwZipsmQZAW5m4lUW7pUEUycgjADWZ6ZrNGgGUsUeWveJ47XC5PN/j1pLgJ5SUedhruSyTXM2nXXC3andGIg+Qvoc8AAUf4j4j0AaLwOikxNl21sbGz+NA9dHEyTbiKHnAAAAABJRU5ErkJggg==","orcid":"","institution":"LMU Munich","correspondingAuthor":true,"prefix":"","firstName":"Julia","middleName":"","lastName":"Cecil","suffix":""},{"id":324675396,"identity":"b72e521a-6c0a-4d5c-a4a3-b51279dec2cf","order_by":1,"name":"Anne-Kathrin Kleine","email":"","orcid":"","institution":"LMU Munich","correspondingAuthor":false,"prefix":"","firstName":"Anne-Kathrin","middleName":"","lastName":"Kleine","suffix":""},{"id":324675397,"identity":"17e810df-3d7e-4dbe-8a9a-cfe5eb969bf9","order_by":2,"name":"Eva Lermer","email":"","orcid":"","institution":"LMU Munich","correspondingAuthor":false,"prefix":"","firstName":"Eva","middleName":"","lastName":"Lermer","suffix":""},{"id":324675398,"identity":"7ade7585-7302-420a-a088-8d46d8572655","order_by":3,"name":"Susanne Gaube","email":"","orcid":"","institution":"University College London, UCL East – Marshgate","correspondingAuthor":false,"prefix":"","firstName":"Susanne","middleName":"","lastName":"Gaube","suffix":""}],"badges":[],"createdAt":"2024-07-05 12:26:31","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4692251/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4692251/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":60172779,"identity":"899e6c67-d866-419e-8bbc-a90f2d2ef5d4","added_by":"auto","created_at":"2024-07-12 15:18:29","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":56969,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eProposed research model for each of the following application areas: diagnostics, treatment, feedback, and practice management\u003c/em\u003e. Components of the COM-B model (40) are abbreviated as followed: C = Capability, M = Motivation\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/699075b74fc7e12ce5419bff.png"},{"id":60172782,"identity":"20b098be-3d4a-4271-bd90-009cb85c5919","added_by":"auto","created_at":"2024-07-12 15:18:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":43654,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eDistribution of sources of information regarding AI-enabled technology for mental health. \u003c/em\u003eResponses from participants who heard of AI in mental health (n = 176). Mainstream media included media coverage, news, internet, social media, podcasts, and newspaper articles.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/26cb4d64f345df82523caab5.png"},{"id":60172781,"identity":"155f57b9-89bb-4bc2-9aeb-56507fdc14a6","added_by":"auto","created_at":"2024-07-12 15:18:29","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":27325,"visible":true,"origin":"","legend":"\u003cp\u003ea) Learning intentions and b) use intentions across the different application areas.\u003c/p\u003e\n\u003cp\u003e* p ≤ 0.05; ** p ≤ 0.01, *** ≤ 0.001\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/77efaebe5a1220eee2bb82d3.png"},{"id":60172783,"identity":"fb74a041-90b7-470f-be17-6bd3ea25f602","added_by":"auto","created_at":"2024-07-12 15:18:29","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":218124,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 5\u003c/strong\u003e. \u003cem\u003eFinal structural equation models for a) diagnostic, b) treatment, c) feedback, and d) practice management tools\u003c/em\u003e. Only nonzero paths are displayed. Components of the COM-B model (40) are abbreviated as followed: C = Capability, M = Motivation.\u003c/p\u003e\n\u003cp\u003e* p ≤ 0.05; ** p ≤ 0.01, *** ≤ 0.001\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/8aa2a9267d54e79c606d23be.png"},{"id":60174648,"identity":"23de9968-dc48-4f55-9ce7-6f8aff20a700","added_by":"auto","created_at":"2024-07-12 15:34:30","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1120849,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/cfb22236-8b6c-4bfb-b99d-d28f02e4a75b.pdf"},{"id":60174045,"identity":"298edca6-ae2d-4ea9-817a-f59191eaf4f2","added_by":"auto","created_at":"2024-07-12 15:26:29","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":50284,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineappendixMentalhealthpractitioners.docx","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/789b29d20f0c9b2d96c9575c.docx"},{"id":60174046,"identity":"3bc47718-e4b1-48ce-b222-5061eacd4d4c","added_by":"auto","created_at":"2024-07-12 15:26:29","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":4013640,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.docx","url":"https://assets-eu.researchsquare.com/files/rs-4692251/v1/26f83dfdcb37253fc587d38a.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Mental health practitioners’ perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eOne in eight people worldwide is affected by a mental disorder, and the trend is rising\u0026nbsp;(1). Frequently, the demand for therapeutic support exceeds available resources, especially since the number of mental health practitioners is not increasing quickly enough\u0026nbsp;(2). Simultaneously, technologies enabled by artificial intelligence (AI) are advancing and gaining relevance in the support and delivery of patient care, owing to their potential for improving patient outcomes through an early detection of mental disorders and personalized treatment\u0026nbsp;(3), and facilitating the work of practitioners\u0026nbsp;(4). Given the proposed benefits, AI-enabled technologies provide an opportunity to bridge the gap between mental healthcare needs and available therapeutic resources.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eApplications of AI-enabled technologies in mental healthcare\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAI-enabled technologies refer to systems or applications characterized by humanlike capabilities, including decision-making through problem solving and continuous learning\u0026nbsp;(3). To execute their tasks effectively, these technologies rely on large amounts of data. Common data sources for AI-enabled technologies in mental healthcare include behavioral data (e.g., video and audio recordings), followed by biological (e.g., blood samples) and neuroimaging data (e.g., electroencephalogram)\u0026nbsp;(5). Within mental healthcare, AI-enabled technologies utilized by clinicians that leverage these datasets can be broadly categorized into four application areas: diagnostic support, treatment support, feedback, and practice management.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe first two application areas, diagnostic and treatment support, refer to patient-centered technologies. \u003cem\u003eDiagnostic\u0026nbsp;\u003c/em\u003eapplications leverage AI to enhance the accuracy and efficiency of mental health assessments by evaluating a range of patient data, such as genetic information, language, voice, and facial expressions\u0026nbsp;(6\u0026ndash;8). For example, certain tools can distinguish between diagnoses that share similar symptoms but require different treatment approaches, such as various types of dementia or bipolar and unipolar depression\u0026nbsp;(9).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe second area of technologies provides \u003cem\u003etreatment\u003c/em\u003e support, making mental health treatments more personalized and precise\u0026nbsp;(10). These technologies are predominantly working with genetic, neuroimaging, clinical and demographical datasets\u0026nbsp;(11). For instance, AI-enabled technologies can be utilized at the beginning of therapy to estimate a patient\u0026rsquo;s potential response to different medications, such as antidepressants, or to predict remission rates\u0026nbsp;(11).\u003c/p\u003e\n\u003cp\u003eBesides these patient-centered technologies, an increasing number of practitioner-centered applications are emerging, with the third area comprising \u003cem\u003efeedback\u0026nbsp;\u003c/em\u003etools for mental health professionals: These types of applications aim to provide practitioners with feedback on the quality of their patient interactions by evaluating session data, for instance, through speech signals and the language patterns of the interaction\u0026nbsp;(12\u0026ndash;15). Feedback reports usually include an assessment of the session\u0026rsquo;s strengths and potential areas for improvement, such as increasing the times for reflections or including more open-ended questions\u0026nbsp;(16).\u003c/p\u003e\n\u003cp\u003eFinally, the fourth application area of AI-enabled technologies for mental health is \u003cem\u003epractice management.\u0026nbsp;\u003c/em\u003eThey\u003cem\u003e\u0026nbsp;\u003c/em\u003eare supposed to automate clinical and administrative workflows and thereby reduce the administrative burden for mental healthcare professionals\u0026nbsp;(16). For example, by automatically transcribing therapy sessions using speech data and integrating the transcripts into medical records\u0026nbsp;(16), patient data entry can become more efficient and structured\u0026nbsp;(17).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdoption of AI-enabled tools in mental healthcare and its antecedents\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe proposed benefits of using AI tools such as an early detection of mental disorders, increasing patient access, and personalized treatment will only be realized if practitioners use them as intended\u0026nbsp;(7). However, studies show widespread skepticism regarding the use of AI-enabled technologies in healthcare\u0026nbsp;(18,19,9,20,21). A lack of understanding or knowledge of the mechanisms and processes underlying the technology may explain some of the suspicion that impacts the uptake of technologies\u0026nbsp;(22,23). Therefore, gaining deeper insights into the current state of mental health practitioners\u0026rsquo; understanding of and experiences with AI-enabled tools is the first step to recognize barriers to the adoption and determine starting point for measures aimed at promoting safe technology practices. However, to the best of our knowledge, no study has investigated practitioners\u0026rsquo; understanding of AI-enabled tools for mental healthcare (RQ1), their familiarity with these technologies (RQ2), in what context they learned about them (RQ3), and whether they have used any of these tools in their clinical practice (RQ4). Besides knowledge and exposure, technology acceptance and effective use is influenced by numerous individual variables.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe role of learning in the adoption of AI-enabled technologies\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStudies have highlighted the pivotal role of learning opportunities and training in the implementation process by equipping healthcare professionals with the requisite skills to effectively use AI-enabled technologies in their practice\u0026nbsp;(24\u0026ndash;26). Conversely, healthcare professionals ranked the lack of instruction and training on technology use as the primary technology-related cause of medical errors\u0026nbsp;(27). Training is believed to reduce the perceived risk associated with using such tools and, further, minimize the workload arising from the implementation of AI technologies\u0026nbsp;(28). It has been shown that the willingness to receive training about an AI technology is positively associated with clinicians\u0026rsquo; use of it\u0026nbsp;(28). We, therefore, hypothesized that \u003cem\u003elearning intention\u003c/em\u003e is positively associated with \u003cem\u003euse intention\u003c/em\u003e for AI-enabled technologies in mental healthcare (H1). Figure 1 depicts the proposed model with the related hypotheses and research questions. However, learning intentions and use intentions represent different levels of engagement with technologies. The willingness to learn and receive training is a rather theoretical interaction with a technology centered around updating knowledge\u0026nbsp;(29). Yet, use intention implies the willingness to make the necessary effort to use the technology in practice\u0026nbsp;(30,31). Hence, it is important to study both the learning and use intention and their respective antecedents independently.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIndividual-level factors in the adoption of AI-enabled technologies\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMost studies have focused on AI adoption in general healthcare settings (see\u0026nbsp;(32)\u0026nbsp;for a review) or different medical specialties such as dermatology\u0026nbsp;(33). However, less is known about individual-level factors associated with practitioners\u0026rsquo; intentions to learn about and use AI-enabled technologies in mental healthcare.\u0026nbsp;User characteristics represent one of the key determinants for the adoption of healthcare technologies\u0026nbsp;(34). Research showed that common demographic and individual differences such as gender\u0026nbsp;(35), age\u0026nbsp;(36), personality\u0026nbsp;(32,33,37), and country of residence\u0026nbsp;(38,39)\u0026nbsp;influence technology uptake. Further, practitioners\u0026rsquo; intention to use AI-enabled technologies in mental health is greatly influenced by their individual beliefs, attitudes, and perceptions\u0026nbsp;(19). Hence, this study seeks to extend existing literature by systematically investigating individual factors that contribute to a holistic understanding of the determinants affecting the learning and use intention of AI-enabled technology in mental healthcare. The Capability-Opportunity-Motivation Behavior (COM-B) model developed by Michie et al.(40), a well-validated behavior change theory, has been used successfully in synthesizing and understanding healthcare-related technology adoption (for instance, see\u0026nbsp;(41,42)). The COM-B model indicates that individuals\u0026rsquo; capabilities, motivation, and opportunities determine their behavior\u0026nbsp;(40). Capability is defined as an individual\u0026apos;s psychological and physical ability required for a particular behavior, including the essential knowledge and skills. Motivation encompasses reflective or automatic cognitive processes that direct behavior, extending beyond conscious decision-making to habitual patterns, emotional responses, and analytical reasoning. Opportunity relates to external factors lying outside an individual\u0026apos;s immediate control that influence behavior, including social and physical opportunity\u0026nbsp;(40). Upon reviewing the empirical literature, we identified the most important individual-level factors relevant to technology adoption and ultimately integrated them into the COM-B framework. As opportunity includes factors outside the individual, we focused on the domains of capabilities and motivations.\u003c/p\u003e\n\u003cp\u003eFirst, individuals\u0026rsquo; capability is important for engaging in a respective behavior\u0026nbsp;(40). Different aspects of capability, including AI knowledge, have been found to be relevant for AI adoption. A positive relation between AI knowledge and the intention to use AI technology was found among prospective physicians\u0026nbsp;(43)\u0026nbsp;and among prospective therapists for feedback providing AI tools\u0026nbsp;(21). Similarly, a lack of technology-related skills and knowledge among therapists was identified as a barrier in the use of technology in forensic psychiatry\u0026nbsp;(44). However, one study found no significant association between AI knowledge and medical students\u0026rsquo; intention to learn about AI\u0026nbsp;(45). As AI knowledge referred to different aspects in each study, and the mixed findings consequently might have resulted from methodological differences, we are adopting a broader construct called \u003cem\u003ereadiness for medical AI\u003c/em\u003e.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eReadiness for medical AI can be divided into different subdimensions\u0026nbsp;(46): \u003cem\u003eCognitive readiness\u003c/em\u003e encompasses peoples\u0026rsquo; cognitive abilities such as knowledge of and critical thinking about AI technologies. \u003cem\u003eVision readiness\u003c/em\u003e involves the ability to envision and anticipate the potential impact, benefits, and challenges associated with AI technologies. \u003cem\u003eEthical readiness\u003c/em\u003e refers to an individual\u0026rsquo;s awareness, knowledge and adherence to ethical standards or guidelines for the use of AI technologies. The relationship between the subdimensions of medical AI readiness and the learning and use intentions of AI-enabled technologies in mental healthcare has not been examined in-depth. Only one study found a positive association between cognitive readiness and the intention to use a feedback tool in mental healthcare\u0026nbsp;(21). We expected that cognitive readiness (H2a, H3a), vision readiness (H2b, H3b), and ethical readiness (H2c, H3c) are all positively associated with the learning and use intentions of AI tools for mental health (see Figure 1 for all hypotheses).\u003c/p\u003e\n\u003cp\u003eSecond, automatic motivational processes influence a particular behavior\u0026nbsp;(40). In the context of technology adoption, automatic processes like emotions, as a sub-component of motivation, have been shown to have an influence\u0026nbsp;(40). Usually, negative valanced variables, such as \u003cem\u003eAI\u003c/em\u003e \u003cem\u003eanxiety,\u0026nbsp;\u003c/em\u003ehave been investigated\u0026nbsp;(47). \u003cem\u003eAI anxiety\u003c/em\u003e refers to the apprehension, concern, or fear experienced in response to the implementation, use, or potential consequences of AI technologies\u0026nbsp;(48). The construct encompasses three subdimensions: \u003cem\u003elearning anxiety\u003c/em\u003e, \u003cem\u003esociotechnical blindness\u003c/em\u003e, and \u003cem\u003ejob replacement anxiety\u003c/em\u003e (47). \u003cem\u003eLearning anxiety\u003c/em\u003e refers to the anxiety regarding acquiring knowledge and skills related to AI technologies. S\u003cem\u003eociotechnical blindness\u003c/em\u003e relates to anxiety arising from a lack of understanding that AI systems currently do not operate independently without human oversight. \u003cem\u003eJob replacement anxiety\u003c/em\u003e refers to a person\u0026rsquo;s fear that their occupation will be replaced or disrupted by AI technologies\u0026nbsp;(37,49). Y.-M. Wang et al., showed that AI learning anxiety negatively affected intrinsic and extrinsic learning motivation\u0026nbsp;(47). They also found that job replacement anxiety positively influenced extrinsic but not intrinsic learning motivation, indicating that some people might only gain AI-relevant skills and knowledge to avoid unemployment. Regarding use intentions, technology anxiety emerged as one important barrier of technology use in healthcare\u0026nbsp;(50). AI anxiety correlated negatively with the use intention of AI-based technology in healthcare among nurses\u0026nbsp;(51)\u0026nbsp;and the intention to use AI-based treatment and feedback tools among prospective psychotherapists\u0026nbsp;(21). While there is consistent evidence, that AI anxiety hinders AI adoption, none of these studies explored associations between all three subdimensions and learning and use intentions for AI-enabled technologies simultaneously. Therefore, we incorporated all three subdimension separately into our research model. We hypothesized that AI learning anxiety (H2d, H3d) and sociotechnical blindness (H2e, H3e) are negatively associated with both the learning and use intentions of AI tools. Job replacement anxiety is thought to be positively associated with the AI learning intentions (H2f) and but negative with use intentions (H3f).\u003c/p\u003e\n\u003cp\u003eThird, in addition to automatic motivational processes, reflective processes, are also crucial, with self-efficacy being an important factor influencing behavior uptake\u0026nbsp;(40). The subcategory tailored to technology is \u003cem\u003etechnology-self efficacy\u003c/em\u003e which refers to a person\u0026rsquo;s belief in their capacity to effectively accomplish a technologically advanced task\u0026nbsp;(52). It is well established that \u003cem\u003etechnology self-efficacy\u003c/em\u003e is an important predictor of technology adoption in healthcare\u0026nbsp;(53). Higher technology self-efficacy has been positively associated with medical students\u0026rsquo; intention to learn technologies\u0026nbsp;(45), healthcare professionals\u0026rsquo; readiness to adopt technologies\u0026nbsp;(54)\u0026nbsp;as well as their intention to use nursing apps and AI technology\u0026nbsp;(51,55,56). In accordance with this large body of research, it is hypothesized, that technology self-efficacy is positively associated with AI learning and use intentions among mental health practitioners (H2g, H3g).\u003c/p\u003e\n\u003cp\u003eFourth, \u003cem\u003eaffinity for technology interaction\u003c/em\u003e represents another motivational process. It serves as a fundamental resource for technology adoption as it is characterized as the tendency to proactively partake in extensive technological interaction\u0026nbsp;(57). Higher affinity for technology was positively related to using a wider range of learning strategies for different healthcare systems among physician trainees\u0026nbsp;(58). Among clinicians, a positive association between affinity for technology and attitude towards technology use has been found and higher technology affinity was linked to a preference for more advanced technologies\u0026nbsp;(59,60).\u0026nbsp;To the best of our knowledge, the relationship between affinity for technology interaction and the intention to learn or use AI technologies in mental healthcare has not been investigated. Based on previous evidence from the medical context, we hypothesized that affinity for technology interaction is positively associated with AI learning and use intentions (H2h, H3h).\u003c/p\u003e\n\u003cp\u003eFinally, the relevance of people\u0026rsquo;s perception of their social and professional role and identity as a motivational factor has also been highlighted in the context of technology adoption, often through \u003cem\u003eprofessional identification.\u003c/em\u003e Professional identification refers to the degree to which an individual feels a deep connection and unity with their chosen occupation\u0026nbsp;(61). Professional identification plays an important role in the adoption of novel work behavior\u0026nbsp;(61), particularly important with the integration of AI-enabled technologies that affects practitioners\u0026rsquo; daily tasks\u0026nbsp;(62). However, changes in the workplace are likely to be resisted if they are perceived as a threat to professional identity\u0026nbsp;(63). It has been shown that threats to professional identity directly impacted healthcare practitioners\u0026rsquo; technology use\u0026nbsp;(64).\u0026nbsp;Moreover, aligned professional beliefs with the designated roles of technology are fundamental for technology adoption\u0026nbsp;(65)\u0026nbsp;as one\u0026rsquo;s professional identification influences technology integration\u0026nbsp;(63). Given these insights, the following research questions are proposed as we could not derive a clear direction of the effects from the literature: Is professional identification associated with AI learning intention (RQ5) and AI use intention (RQ6)?\u003c/p\u003e\n\u003cp\u003ePrior research has shown that there are differences in use intentions and its predictors across AI tools for different application areas\u0026nbsp;(21). As AI-enabled technologies in mental healthcare differ vastly in their purpose, they might also be perceived differently by mental health practitioners. Therefore, we believe it is important to look at the learning and use intentions and their antecedents individually for each application area. Providing such a nuanced understanding enables technology developers and healthcare organizations who purchase these technologies to consider the factors relevant to the tool in question, thereby facilitating a more efficient and safe design and implementation process. As a consistent methodology that allows comparisons across the different application areas on the same level is fundamental for this, we applied the same research design and sample across all four application areas of AI-enabled technologies in mental healthcare. This allows us to systematically identify potential differences, ultimately resulting in a comprehensive overview of different application areas and their antecedents.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe present study\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe main goal of this mixed method study was twofold. First, we want to investigate mental health practitioners\u0026rsquo; general understanding, familiarity, and experience with AI technologies (RQ1 \u0026ndash; RQ4) and their attitudes towards different application areas of AI-enabled tools using qualitative content and descriptive analysis. In this line, we also examined differences in attitudes toward technology across different professions, gender, and countries. Second, this work aims to provide a differentiated insight into factors associated with learning and use intentions of AI-enabled technologies for mental health, separated by application areas (H1, H2a \u0026ndash; H2h, H3a \u0026ndash; 3h, and RQ5 and RQ6). Gaining a deeper understanding of the relative importance of individual factors might help for deriving training and intervention strategies tailored specifically towards practitioners\u0026apos; needs for different technology application areas.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eParticipants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData for the pre-registered (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497) cross-sectional, mixed-methods survey study was collected between July and October 2023. Participants included psychotherapists in training, psychotherapists, psychiatrists, and clinical psychologists. Participants were recruited via emails distributed among universities and psychotherapy training institutes in Germany and the US, social media postings, and Prolific. The online survey was available in German and English language. For the German version of the survey, all items were translated using back-and-forth translation. Both survey versions can be found in the online supplements (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497). In total, 670 mental health practitioners agreed to participate, of which 227 did not finish the survey and 51 failed at least one attention check item, resulting in \u003cem\u003eN\u003c/em\u003e = 392 participants included in the data analysis. This number exceeded the minimum sample size determined by the a priori power analysis for structural equation modeling (SEM), which required at least 50 practitioners per country (Germany and US). Demographic information of the included participants can be found in Table 1. The study was approved by the Ethics Committee of [blinded for review].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProcedure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFirst, demographic and occupation-related information was assessed in the survey. Second, participants\u0026rsquo; understanding of, familiarity and experiences with, and use of AI-enabled tools were assessed. Third, participants were then introduced to the four different application areas of AI-enabled technologies in mental health. For each area, participants received a short description and an example (see Table 2). We measured learning and use intentions as dependent variables for each application area, the individual level factors as predictor variables, several control and occupation-related variables (occupation, therapeutic approach, workplace, working experience in years) as described below.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeasurements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eUnderstanding\u003c/em\u003e: Participants were asked to describe what they understand by AI-enabled technologies in the field of psychotherapy/psychiatry and how they could be used in their daily work in their own words, using an open text box.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eFamiliarity\u003c/em\u003e:\u0026nbsp;Next, they were asked to choose one of three options regarding their familiarity with AI-enabled technologies (a: \u0026ldquo;I have never heard of AI-enabled technologies in psychotherapy/psychiatry\u0026rdquo;; b: \u0026ldquo;I have heard of AI-enabled technologies in psychotherapy/psychiatry\u0026rdquo;; c: \u0026ldquo;I have actively looked into AI-enabled technologies in psychotherapy/psychiatry\u0026rdquo;). Participants who had stated to have heard of AI-enabled technologies were asked in which context they did so (open question). Participants who had stated that they actively looked into AI technology, were given three context options: \u0026ldquo;I have informed myself independently (e.g., online, \u0026hellip;)\u0026rdquo;, \u0026ldquo;I attended voluntary information sessions on AI-enabled technologies in psychotherapy/psychiatry\u0026rdquo;, and \u0026ldquo;I have participated in trainings on this topic (e.g., to get training points).\u0026rdquo;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eUse\u003c/em\u003e: To determine previous use, participants were asked to state whether they had used AI-enabled technologies in their clinical practice (yes/no).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDependent variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTwo dependent variables, \u003cem\u003elearning intention\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;use intention\u003c/em\u003e, were assessed for each of the four described application areas for AI tools in mental health. Learning intention was measured with \u0026ldquo;I intend to learn about AI technologies in [application area]\u0026rdquo; on a 5-point Likert scale from 1 (\u003cem\u003estrongly disagree\u003c/em\u003e) to 5 (\u003cem\u003estrongly agree\u003c/em\u003e) based on Venkatesh et al. (31). Similarly, use intention was assessed with the item \u0026ldquo;I intend to use AI technologies in [application area] in my work\u0026rdquo; with the same response format\u0026nbsp;(31).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePredictor variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMedical AI Readiness\u003c/em\u003e: \u003cem\u003eCognitive, vision\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;ethical readiness\u003c/em\u003e \u003cem\u003efor medical AI\u003c/em\u003e was based on the Medical Artificial Intelligence Readiness Scale (MAIRS) from Karaca et al.\u0026nbsp;(66). For each of the subscales we omitted items for two reasons. First, items measuring the actual use of technology were removed, as we assumed that most practitioners are not currently using AI-enabled tools and therefore these questions could not be answered properly. Second, items with low factor loadings were removed to keep the survey reasonable short. Consequently, we included 11 items, rated on a 5-point Likert scale from 1 (\u003cem\u003estrongly disagree\u003c/em\u003e) to 5 (\u003cem\u003estrongly agree\u003c/em\u003e). The scale showed acceptable (𝛼\u003csub\u003eVision\u003c/sub\u003e = 0.79,\u0026nbsp;𝛼\u003csub\u003eEthics\u003c/sub\u003e = 0.73) to good internal consistency (𝛼\u003csub\u003eCognition\u003c/sub\u003e = 0.81).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAnxiety\u003c/em\u003e: \u003cem\u003eAI learning anxiety, job replacement anxiety\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;sociotechnical blindness\u003c/em\u003e were assessed using the 18-item Artificial Intelligence Anxiety Scale (AIAS) by Wang \u0026amp; Wang\u0026nbsp;(49)\u0026nbsp;on a 7-point Likert scale from 1 (\u003cem\u003estrongly disagree\u003c/em\u003e) to 7 (\u003cem\u003estrongly agree\u003c/em\u003e). The internal consistency of the sociotechnical blindness subscale was acceptable (𝛼\u003csub\u003eSociotechnical\u003c/sub\u003e = 0.78), that of the job replacement anxiety subscale good (𝛼\u003csub\u003eJobreplacement\u003c/sub\u003e = 0.87) and that of the AI learning anxiety subscale was excellent (𝛼\u003csub\u003eLearning\u003c/sub\u003e = 0.93).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAffinity for technology\u003c/em\u003e \u003cem\u003einteraction\u0026nbsp;\u003c/em\u003ewas measured with the Affinity for Technology Interaction Scale (ATI-S\u0026nbsp;(67)). The four items were rated on a 7-point Likert scale from 1 (\u003cem\u003ecompletely disagree\u003c/em\u003e) to 7 (\u003cem\u003ecompletely agree\u003c/em\u003e). The scale showed good internal consistency (𝛼\u003csub\u003eAffinity for technology\u003c/sub\u003e = 0.81).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTechnology self-efficacy\u003c/em\u003e was assessed using the five-item scale of McDonald and \u0026nbsp;Siegall\u0026nbsp;(52)\u0026nbsp;on a 7-point Likert scale from 1 (\u003cem\u003estrongly disagree\u003c/em\u003e) to 7 (\u003cem\u003estrongly agree\u003c/em\u003e). The internal consistency of the scale was acceptable (𝛼\u003csub\u003eTechnology self-efficacy\u003c/sub\u003e = 0.71).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eProfessional identification\u003c/em\u003e was measured using the five items from Hekman et al.\u0026nbsp;(61)\u0026nbsp;on a 5-point Likert scale from 1 (\u003cem\u003estrongly disagree\u003c/em\u003e) to 5 (\u003cem\u003estrongly agree\u003c/em\u003e). The scale showed acceptable internal consistency (𝛼\u003csub\u003eProfessional identification\u003c/sub\u003e = 0.77).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eControl variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAge, gender, and personality were included as control variables based on research showing that all three variables have an impact on technology adoption\u0026nbsp;(32,33,35\u0026ndash;37). Participants\u0026lsquo; personality traits were assessed using the Big Five Inventory\u0026nbsp;(68), on a 5-point Likert scale from 1 (\u003cem\u003estrongly disagree\u003c/em\u003e) to 5 (\u003cem\u003estrongly agree\u003c/em\u003e), including the main dimensions openness, conscientiousness, extraversion, agreeableness, and neuroticism.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData was analyzed using R (Version 4.3.2, R Core Team, 2023). Answers to the open questions were coded using Excel. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQualitative and descriptive analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFirst, we conducted a qualitative content analysis to get in-depth insights into mental healthcare practitioners\u0026rsquo; understanding of AI-technology for their field of work (RQ1), and allowing for participants\u0026rsquo; viewpoints to emerge\u0026nbsp;(69). To gain these insights, we used a deductive thematic analysis\u0026nbsp;(70)\u0026nbsp;to identify how many types of AI applications were mentioned by practitioners. Participants\u0026rsquo; responses were clustered into the four predefined application areas and then analyzed for their frequency, to gain insights about the most known and common areas. Further, the precision of their description of AI-enabled technologies in mental healthcare was assessed. We examined whether practitioners could not give a description if the descriptions solely included the technology\u0026rsquo;s potential area of application or if also the tool\u0026rsquo;s underlying functions or operational mechanism were explained properly. For answers to the open question regarding the context in which they have heard about the AI technologies (RQ3), an inductive approach\u0026nbsp;(70)\u0026nbsp;was employed to identify recurrent categories within the data. Participants\u0026rsquo; responses were coded based on similarities and organized subsequently into themes representing higher-level concepts. All responses were independently coded by two researchers to review and validate the identified themes with subsequent discussion in cases with coding discrepancies. The code book can be found in the online material on OSF (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSEM\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNext, to look at the learning and use intentions, we specified one SEM model for each application area using the \u0026lsquo;lavaan\u0026rsquo; package\u0026nbsp;(71). Confirmatory factor analyses (CFA) were calculated for each model. For the model fit, root-mean-square error of approximation (RMSEA) values smaller than 0.05 are considered good and smaller than 0.08 acceptable\u0026nbsp;(72). Standardized root-mean-square residual (SRMR) values up to 0.08 are considered satisfactory\u0026nbsp;(73). Models showing comparative fit index (CFI) and Tucker Lewis index (TLI) values near to or surpassing 0.90 possess a reasonable level of fit\u0026nbsp;(73). For each application area, we analyzed models to predict learning and use intention from the predictor variables and the control variables age, gender, and personality. Further, we calculated three more parsimonious theoretical models to avoid overfitting and ensure the distinctness of the variables.\u0026nbsp;For the first parsimonious model, we combined the subscales of readiness for medical AI. In the second parsimonious model, the subscales of AI anxiety were merged, and in the third parsimonious model, affinity for technology interaction and technology self-efficacy were combined. All in all, SEMs were calculated for one research model per application area with and without control variables, as well as the three more parsimonious models, totaling eleven models.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExplorative analysis of demographic and tool differences\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFinally, for the analysis of potential group differences, we assessed the mean values, standards deviations, and correlations between the variables used in the SEM. Group differences across the four application areas and practitioners\u0026rsquo; subgroups (profession, gender, country) were assessed using t-tests or one-way ANOVAs with post-hoc Tukey-HSD. The data was found to be normally distributed following testing for assumptions, with only minor violations observed for learning and use intentions. However, simulation studies demonstrated that, particularly in studies with larger samples, such violations have a negligible impact on the results (74). Additionally, familiarity and use experiences with AI-enabled technologies among mental health practitioners and their context (RQ2-4) were analyzed descriptively.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003ePractitioners\u0026rsquo; understanding and familiarity with different application areas\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhen participants were asked to explain their understanding of AI-enabled technologies in mental healthcare and how they could be used in their daily work in their own words, 10.5% could not provide a description. Over half of those that provided a description (53.7%) mentioned only one application area, while a further 37.6% stated two categories (RQ1). Merely 8.1% of participants named three areas, whilst only 0.6% of participants (n = 2) listed all four. AI-enabled tools for supporting treatment decisions emerged as the most frequently mentioned area (69.8%), followed by diagnostic (43.4%) and practice management tools (41.1%). Only six participants mentioned feedback tools (1.7%). Participants exhibited varying levels of precision in the description of these technologies, however mostly demonstrating a basic understanding through their explanations. While a majority provided less detailed statements, such as indicating AI\u0026rsquo;s role as \u0026quot;diagnostic assistance\u0026quot; (Clinical psychologist, 45), a minority offered more elaborate descriptions, exemplified by one professional\u0026rsquo;s description that \u0026ldquo;AI could help to make diagnosis [\u0026hellip;] more efficient and precise by pooling larger data sources together (e.g., interview data, EHR data, patient-reported outcomes, biomarker data)\u0026rdquo; (Clinical psychologist, 47). For treatment tools, most participants also solely addressed their general purpose, such as \u0026ldquo;tools that have been programmed to respond to folks in crisis\u0026rdquo; (Psychotherapist in training, 32). A smaller subset displayed a deeper understanding by mentioning the underlying working mechanism: \u0026ldquo;By considering an individual\u0026apos;s unique history, symptoms, and responses to therapy, AI can recommend specific interventions and strategies tailored to their needs\u0026rdquo; (Psychiatrist, 69). Professionals mostly described feedback tools briefly as tools that \u0026ldquo;give input into your performance as a therapist\u0026rdquo; (Clinical psychologist, 26). Only two participants provided additional information by stating that \u0026ldquo;there are programs that listen to and transcribe therapy sessions and from this identify themes, relational patterns, and can even rate the therapist on various qualities and suggest interventions\u0026rdquo; (Clinical psychologist, 35). Likewise, a disparity in the precision level of participants\u0026rsquo; responses emerged about practice management tools, ranging from succinct descriptions, such as \u0026ldquo;documentation of visit\u0026rdquo; (Psychiatrist, 46) \u0026nbsp;and \u0026ldquo;can be used to write notes\u0026rdquo; (Psychotherapist, 34) to more elaborate insights: \u0026ldquo;I think predictive text could be used for things like notes and that AI software can be used for recording and transcribing sessions, and then generating notes\u0026rdquo; (Clinical psychologist, 33).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExperiences of mental health practitioners with AI-enabled technologies\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNearly half of the practitioners (\u003cem\u003en\u003c/em\u003e = 178, 45.4%) stated that they have never heard of AI-enabled technologies in the field of psychotherapy/psychiatry, while 44.9% (\u003cem\u003en\u003c/em\u003e = 176) did (RQ2). Figure 2 displays their sources of information. Overall, only 9.7% (\u003cem\u003en\u003c/em\u003e = 38) actively looked into this topic, whose majority obtained information independently through online research (\u003cem\u003en\u003c/em\u003e = 29, 76.3%). A further 10.5% (\u003cem\u003en\u003c/em\u003e = 4) stated that they attended voluntary information sessions and only 13.2% (\u003cem\u003en\u003c/em\u003e = 5) participated in formal trainings (RQ3). The vast majority of participating practitioners (n = 366, 93.37%) have not used AI-enabled technologies in their clinical practice (RQ4).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLearning and use intentions across application areas\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data were normally distributed, with mild violations for learning and use intentions. However, simulation studies showed that especially for larger samples as in our study, mild violations have little to no effect on the results The overall learning intention was significantly higher than the overall use intention, \u003cem\u003et\u003c/em\u003e(781) = 8.17, \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001, d = 0.584; \u003cem\u003eM\u003c/em\u003e\u003csub\u003eLearning\u003c/sub\u003e = 3.65, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eLearning\u003c/sub\u003e = 0.88; \u003cem\u003eM\u003c/em\u003e\u003csub\u003eUse\u003c/sub\u003e = 3.14, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eUse\u003c/sub\u003e = 0.88). Further, both differed across the four application areas. Practitioners\u0026rsquo; intention to learn was significantly higher for AI-enabled management tools (\u003cem\u003eM\u003c/em\u003e = 3.91, \u003cem\u003eSD\u003c/em\u003e = 1.01) compared to diagnostic (\u003cem\u003eM\u0026nbsp;\u003c/em\u003e= 3.53, \u003cem\u003eSD\u003c/em\u003e = 1.12), treatment (\u003cem\u003eM\u003c/em\u003e = 3.65, \u003cem\u003eSD\u003c/em\u003e = 1.09), and feedback tools (\u003cem\u003eM\u003c/em\u003e = 3.53, \u003cem\u003eSD\u003c/em\u003e = 1.19; \u003cem\u003eF\u003c/em\u003e(3, 1564) = 10.38, \u003cem\u003ep\u0026nbsp;\u003c/em\u003e\u0026lt; .001, \u0026eta;\u003csub\u003ep\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e = 0.02; see Figure 3a). Practitioners\u0026rsquo; use intentions were significantly higher for AI-enabled tools for feedback (\u003cem\u003eM\u003c/em\u003e = 3.13, \u003cem\u003eSD\u003c/em\u003e = 1.22) than diagnosis (\u003cem\u003eM\u003c/em\u003e = 2.78, \u003cem\u003eSD\u003c/em\u003e = 1.15) and again, for management tools (\u003cem\u003eM\u003c/em\u003e = 3.70, \u003cem\u003eSD\u003c/em\u003e = 1.10) compared to diagnosis, treatment (\u003cem\u003eM\u003c/em\u003e = 2.96, \u003cem\u003eSD\u003c/em\u003e = 1.16), and feedback (\u003cem\u003eF\u003c/em\u003e(3, 1564) = 46.2, \u003cem\u003ep\u0026nbsp;\u003c/em\u003e\u0026lt; 0.001, \u0026eta;\u003csub\u003ep\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e = 0.08; see Figure 3b). The results indicate that mental health practitioners are more hesitant to learn about and use AI-enabled tools that are more patient-centered compared to more therapist-centered tools that have a less direct influence on decisions that affect patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLearning and use intentions across different occupational and demographic groups\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eLearning and use intentions differed across occupations, with psychiatrists reporting significantly higher intentions to learn (\u003cem\u003eF\u003c/em\u003e(4, 387) = 4.87, \u003cem\u003ep\u0026nbsp;\u003c/em\u003e= 0.002, \u0026eta;\u003csub\u003ep\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e = 0.04) and use AI-enabled technologies compared to psychotherapists in training, psychotherapists, and clinical psychologists (\u003cem\u003eF\u003c/em\u003e(4, 387) = 4.52, \u003cem\u003ep\u0026nbsp;\u003c/em\u003e= 0.001, \u0026eta;\u003csub\u003ep\u003c/sub\u003e\u003csup\u003e2\u003c/sup\u003e = 0.04; see Table A1 in the online appendix). All other differences were non-significant (\u003cem\u003ep\u0026nbsp;\u003c/em\u003e\u0026gt; 0.05). Male practitioners showed higher learning intentions (\u003cem\u003et\u003c/em\u003e(153.39) = 2.95, \u003cem\u003ep\u003c/em\u003e = 0.004, d = 4.17) and use intentions compared to female practitioners (\u003cem\u003et\u003c/em\u003e(134.73 = 3.02, \u003cem\u003ep\u003c/em\u003e = 0.003, d = 3.45; see Table A1 in the online appendix). German practitioners reported significantly lower learning intentions compared to their US counterparts, \u003cem\u003et\u003c/em\u003e(363.55) = -4.03, \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001, d = 4.57), however, surprisingly, their use intentions did not differ significantly (\u003cem\u003ep\u003c/em\u003e \u0026gt; 0.05).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSEM\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor all variables used in the SEM models, means, standard deviations, and correlations can be found in Table A2 in the online appendix. Across all four application areas, the complete models showed better fit indices than the parsimonious models, indicating that the model variables were sufficiently distinct (see Table A3 in the online appendix). In all models, one item from the technology self-efficacy scale had standardized factor loadings below 0.40 and was therefore excluded (75). The measurement model of the initially proposed model showed only a partially acceptable fit. Therefore, a second version was calculated, which included the correlated error terms for the two reversed-worded items of the ATI scale. Correlating the measurement errors did not significantly alter the parameter estimates of the underlying measurement model. Table 6 shows the fit indices for each of the final models. The model fit indices for RMSEA (\u0026le; 0.056) and SRMR (\u0026le; 0.063) are acceptable to good. The CFI and TFI close to 0.9 are considered marginal levels (76). As the cutoff-levels for the goodness-of-fit indices depend on model characteristics, such as the sample size and number of variables (77), the complexity of the model and rather small sample size might be the reasons for the CFI and TLI just below the threshold (78).\u003c/p\u003e\n\u003cp\u003eControlling for age, gender, and personality did not substantially affect the models for treatment and feedback tools. For the diagnostic tool, the association between professional identification and learning intention, and for practice management tools, the association between cognitive readiness and learning intention were no longer significant (see Table A4 \u0026ndash; A7 in the online appendix).\u003c/p\u003e\n\u003cp\u003eThe results of the final SEM models are presented in Tables 6 - 9. All significant paths are highlighted in Figure 5. Across the four models, predictor variables accounted for 46.7 \u0026ndash; 61.0% of the variance in learning intention and 8.1 \u0026ndash; 17.0% in use intentions. Across all application areas, the intention to learn about AI-enabled technologies was positively associated with the intention to use these technologies, supporting H1 for each model. Some paths for the subconstructs of medical artificial intelligence readiness, AI anxiety, beliefs about technological capabilities and professional identity were also relevant across all application areas, however, others differed for each application area (see Table 6-9).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Regarding AI knowledge, cognitive readiness (H2a) was positively associated with the learning intention of the feedback tool, vision readiness (H2b) with the learning intention of the feedback tool, and ethical readiness (H3c) with the use intention across each application area. For the automatic motivational factor AI anxiety, sociotechnical blindness (H2e) demonstrated a positive relationship with the learning intentions of the treatment and practice management tool. For reflective motivational factors, technology self-efficacy (H3g) was negatively related to the use intentions for the diagnostic, treatment, and practice management tool. Further, practitioners\u0026rsquo; affinity for technology interaction showed a consistent positive link with the use intentions for all application areas, supporting H3h for each model. Lastly, professional identification (RQ5) was positively associated with the learning intention for the diagnostic, treatment, and feedback tool.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eAmidst the increasing integration of AI-enabled technologies in healthcare, the present study investigated mental health practitioners\u0026rsquo; understanding and familiarity across different application areas for AI-enabled support tools in mental healthcare. Additionally, we examined factors influencing the intention to learn and use AI-enabled technologies across the different areas.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003eCurrent familiarity gaps among mental healthcare professionals\u003c/h2\u003e \u003cp\u003eOur study reveals a limited understanding of AI-enabled technologies and significant gap in mental health practitioners\u0026rsquo; familiarity with AI-enabled tools for mental health, with nearly half of the surveyed practitioners unaware of these technologies. This low familiarity indicates that many professionals are not informed about the development and potential clinical applications of AI in mental healthcare. Additionally, practitioners primarily gained information through mainstream media such as social media or newspaper articles and less than one-tenth of practitioners who had heard about AI technologies received formal education on the topic, a trend consistent with prior research (\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e). Furthermore, the present findings align with an international survey of psychiatrists, which found that less than a quarter had received formal technology training (\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e). Adding to the literature, the fact that the majority of our participants were psychotherapists currently enrolled in training suggests that current training programs may not adequately cover AI-related topics, thereby limiting practitioners' exposure and understanding. As a lack of training and instructions on technology use in healthcare further contributes to an unsafe work environment and medical errors (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e), the results underline the need of adjusting the training to emerging technologies.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003eProfessionals\u0026rsquo; varying adoption intentions and application-specific hesitation\u003c/h2\u003e \u003cp\u003eThe surveyed practitioners were more inclined towards learning rather than actively using AI-enabled technologies in their clinical practice. This supports existing literature indicating that learning and use intentions represent different levels of engagement with technology (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). For the more practical level of intending to use technologies, practitioners\u0026rsquo; main concerns regarding AI technologies, including the lack of transparency of model predictions, data privacy, cyber security, and patient safety (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e), might have contributed to their greater use hesitation. Besides, awareness of the need to inform patients about the use of AI technologies in psychotherapeutic decisions and obtain their consent (\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e, \u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e), along with understanding how these issues affect their work and patients, might contribute to lower usage intentions.\u003c/p\u003e \u003cp\u003eMoreover, participants demonstrated different levels of willingness to engage with AI-enabled technologies across the application areas. Notably, they were less hesitant towards clinician-centered feedback or practice management tools compared to patient-centered tools, aligning with previous findings (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e). This may be attributed to the higher stakes associated with using technology to inform diagnosis or treatment decisions compared to receiving feedback or administrative support as diagnostic or treatment errors can have severe negative consequences, potentially resulting in wrong or delayed treatment and a worse prognosis (\u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e84\u003c/span\u003e, \u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAdditionally, our results revealed profession-specific differences, with psychiatrists demonstrating higher learning and use intentions compared to psychotherapists and clinical psychologists. This difference might stem from the specific characteristics of education and work in each occupation. Psychiatrists undergo medical training that already integrates AI-enabled technologies into the curricula, albeit with a focus on other specialties (\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e). However, their greater exposure to clinical technologies and closer connection to the broader medical field, where AI use is more prevalent than in psychology, might contribute to their higher adoption intentions. Additionally, since medical prescription are part of psychiatrists\u0026rsquo; daily tasks and this area holds widespread potential for AI utilization (for instance see (\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e)), it might be more natural for them to envision using AI into their practice. The practices of psychotherapists and clinical psychologists in turn are centered more around interpersonal treatment and the patient-therapist relationship (\u003cspan citationid=\"CR88\" class=\"CitationRef\"\u003e88\u003c/span\u003e). In this context, technology is often perceived not as a substitute for human care (\u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e), hence, it may be challenging for psychotherapists to envision the integration of AI technology into their professional practice, possibly leading to their greater hesitation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003eIndividual-level predictors of AI adoption intentions\u003c/h2\u003e \u003cp\u003eWe found a robust association between the intention to learn and use AI-enabled technologies across all application areas. This aligns with results showing that the willingness to engage in training enhances professionals\u0026rsquo; intention to use AI technologies (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e). Consequently, willingness to learn is a pivotal initial step in engaging with AI technologies and understanding the predictors for both learning and use intention is important.\u003c/p\u003e \u003cp\u003eFirst, regarding \u003cem\u003eAI knowledge\u003c/em\u003e, the domain ethical readiness emerged as a significant predictor for use intentions across all application areas, making it a driving force for the intention to use AI-enabled technologies in healthcare. This is in line with research showing that AI ethics awareness was positively correlated with the use intention of AI-based technology in nursing care (\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e). The consistent link across all application areas may be explained by the high value of ethics in mental health. Besides general medical ethics, it encompasses elements such as the emotional therapist-patient relationship and handling highly sensitive information, requiring strict adherence to ethical standards (\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e89\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, learning intentions were influenced differently depending on the application area. On the one hand, the ability to anticipate the technology\u0026rsquo;s potential impact, involving a deeper understanding of the technologies\u0026rsquo; strengths and weaknesses (vision readiness), was positively associated with the intention to learn about treatment support tools. As practitioner were most familiar with treatment tools, it is not surprising that practitioners with a more nuanced understanding are more likely to deepen their knowledge in tools they are already familiar with, likely aiming to refine their knowledge. On the other hand, the basic understanding about AI technologies (cognitive readiness) was positively associated with the intention to learn about feedback tools which practitioners were least familiar with. Practitioners with a basic understanding are therefore eager to explore less familiar tools, potentially driven by curiosity and a desire to broaden their knowledge. Hence, the findings suggest that learning intentions vary based on different facets of practitioners\u0026rsquo; AI knowledge, with a basic knowledge leading to a higher intention to learn about new tools and advanced knowledge driving deeper exploration of known tools. These study findings on AI knowledge might help to understand the mixed results found in prior literature which showed a positive association with general AI knowledge in some cases (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e), but not in others (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e); while the present study shows that different facets of AI knowledge have varying influences on the adoption intentions for different tools.\u003c/p\u003e \u003cp\u003eSecond, none of the subdimensions of \u003cem\u003eAI anxiety\u003c/em\u003e was associated with use intentions for any application area, contrary to prior findings indicating that AI anxiety impedes AI adoption (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e). However, previous research concentrated on general AI anxiety, without specifically addressing its nuanced facets (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). For instance, looking at the subdimension of job replacement anxiety, the only moderate levels reported by our participants (see Table A2) might have contributed to this result, indicating that they do not view AI as a threat to their profession. This finding aligns with research indicating that only 4% of psychiatrists believe that future technology will make their jobs obsolete (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e). However, anxiety arising from the belief that AI systems operate without human supervision (sociotechnical blindness) was positively associated with the intention to learn about two AI-enabled application areas: treatment and practice management tools. Contrary to high levels of anxiety, moderate anxiety, as in our study, can have a positive effect on the learning motivation (\u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e) and this might explain the effect in the opposite direction. The effect might have emerged particularly for these two areas, as they are the ones practitioners are most eager to learn about and, in the case of practice management tools, intend to use. Given the pivotal role of human oversight in successfully implementing AI technology, which requires a certain level of tool understanding to monitor its actions and decisions (\u003cspan additionalcitationids=\"CR92\" citationid=\"CR91\" class=\"CitationRef\"\u003e91\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR93\" class=\"CitationRef\"\u003e93\u003c/span\u003e), practitioners may be more inclined to learn about AI technologies they see themselves engaging with, aiming to equip themselves for ensuring proper oversight if needed.\u003c/p\u003e \u003cp\u003eThird, reflective motivational processes played a pivotal role in both learning and use intentions. Across three application areas (diagnostic, treatment, and practice management), professionals\u0026rsquo; \u003cem\u003etechnology self-efficacy\u003c/em\u003e was negatively associated with the intention to use diagnostic, treatment, and practice management tools. However, we found a significant positive correlation between technology self-efficacy and the overall use intention (see Table A2). This discrepancy suggests a suppression effect within the models. This effect occurs when there are multiple predictors in the model, and the overall predictive power of the model is improved by the inclusion of additional predictors that uncover different associations compared to when solely considering technology self-efficacy (\u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e94\u003c/span\u003e). Consequently, the association between technology self-efficacy and the use intention is hard to interpret. However, the suppression effect indicates that while technology self-efficacy is negatively associated with the use intention for some application areas, its overall positive correlation with the intention to use suggests that practitioners with higher beliefs in their ability to effectively perform technologically advanced tasks are more inclined to use AI-enabled technologies, which aligns with existing literature (\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e, \u003cspan additionalcitationids=\"CR54 CR55\" citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFourth, \u003cem\u003eaffinity for technology interaction\u003c/em\u003e, characterized by the enjoyment and comfort in interacting with technology, showed a positive relationship with the use intention for each tool category. This result was expected based on research from broader hospital settings and other medical domains demonstrating this positive association (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e). From a behavioral perspective, cross-situational consistency may explain this finding as people often maintain behavior across similar contexts (\u003cspan citationid=\"CR95\" class=\"CitationRef\"\u003e95\u003c/span\u003e). One\u0026rsquo;s overall positive perception in interacting with technologies might therefore be also transferable to their engagement with technologies at work.\u003c/p\u003e \u003cp\u003eFinally, a strong \u003cem\u003eprofessional identity\u003c/em\u003e exhibited a positive association with intentions to learn about three application areas (diagnostic, treatment, and feedback). The non-significant association with the use of learning intention for the practice management tools may relate to the fact that practitioners do not see administrative tasks as closely related to their identity as mental healthcare professionals. The positive association contributes to existing literature by extending prior insights from general healthcare contexts into mental healthcare (\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e, \u003cspan citationid=\"CR96\" class=\"CitationRef\"\u003e96\u003c/span\u003e). Professional identity is a dynamic concept shaped by various factors, including technology implementation (\u003cspan citationid=\"CR97\" class=\"CitationRef\"\u003e97\u003c/span\u003e, \u003cspan citationid=\"CR98\" class=\"CitationRef\"\u003e98\u003c/span\u003e), and prompting (professionals like) mental healthcare worker to continually assess alignment with evolving work contexts (\u003cspan citationid=\"CR99\" class=\"CitationRef\"\u003e99\u003c/span\u003e, \u003cspan citationid=\"CR100\" class=\"CitationRef\"\u003e100\u003c/span\u003e). Despite limited awareness of these technologies, strong identification with their mental health role might motivate them to learn about technologies, facilitating adaption to workplace changes and alignment with their professional identity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eLimitations and future research\u003c/h2\u003e \u003cp\u003eSeveral limitations should be considered when interpreting the findings of this study. First, the brevity of responses to the open-ended questions may stem from a lack of motivation or time constraints. It is plausible that practitioners possess a more extensive understanding than was conveyed within their response. Future studies could encourage participants to elaborate, for instance by follow-up interviews designed to gather more information on their understanding or by using more objective measures. Second, the inclusion of control variables resulted in the non-significance of vision readiness and professional identification on learning intentions in two models. This, together with the suppression effect on self-efficacy, underscores the complexity of the predictors\u0026rsquo; associations and highlights the need for further exploration to understand the nuanced interplay of variables influencing the learning intentions of AI-enabled technologies. Third, no causal relationships could be observed and tested as the present study was cross-sectional. In the future, longitudinal and experimental designs should be employed. Lastly, participants only got concise descriptions of the different AI application areas the AI tools without the opportunity for direct practical interaction with the technologies. This might have restricted participants\u0026rsquo; depth of understanding and influenced their responses. Future research should explore using detailed, comprehensive, and interactive representations of AI decision-making processes and technologies (\u003cspan citationid=\"CR101\" class=\"CitationRef\"\u003e101\u003c/span\u003e, \u003cspan citationid=\"CR102\" class=\"CitationRef\"\u003e102\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003ePractical implications\u003c/h2\u003e \u003cp\u003eThe fact that half of the practitioners have not heard about AI-enabled technology in mental healthcare demonstrates the need for formal education on this topic. The integration of modules on AI-enabled technologies into curricula and professional training programs holds the potential to redirect professional educational frameworks towards future-oriented challenges like technology interaction. Better training regarding the use of technology might prevent medical errors, as research has shown that healthcare practitioners view a lack of technology training as a major cause of errors (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). Taking it a step further, our study results can also contribute to the development of successful educational frameworks. For instance, ethical knowledge seemed highly relevant for use intentions, hence, education on ethical standards required for technology use is one starting point to ensure their safe and responsible use. As highlighted by Katznelson and Gercke (\u003cspan citationid=\"CR103\" class=\"CitationRef\"\u003e103\u003c/span\u003e), incorporating AI ethics into healthcare training programs is crucial to prepare healthcare professionals for the ethical complexities accompanying AI implementation. Additionally, since affinity for technology interaction was consistently associated with use intentions, the comfort of interacting with technology should also be fostered via practical experiences and on-the-job training. Moreover, addressing hesitations early on or helping users overcome them could involve considering predictors not only in the design of training programs but also the technology itself. One potential solution could involve ensuring more actively that the technology utilizes health data in accordance with legal and ethical norms. Although regulations such as the MDR (Medical Device Regulation) and AIA (Artificial Intelligence Act) are already in place (\u003cspan citationid=\"CR104\" class=\"CitationRef\"\u003e104\u003c/span\u003e), transparently displaying the underlying norms to end users can simultaneously advance their ethical knowledge and ensure adherence to ethical principles. With this, developers can better serve practitioners\u0026rsquo; needs and facilitate their adoption of AI technologies in mental healthcare.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur study reveals a substantial gap in mental healthcare professionals\u0026rsquo; familiarity of AI-enabled technologies in their field. It further underscores the nuanced perception of the different application areas, emphasizing the necessity to consider not only the specific AI application area but also the characteristics of different mental health professionals during the implementation process. Recognizing the pivotal role of learning in initiating engagement, our study suggests that cultivating such engagement via tailored training programs considering robust factors like individuals\u0026rsquo; ethical knowledge and affinity for technology interaction could subsequently enhance professionals\u0026rsquo; inclination towards utilizing these novel technologies. Moving forward, addressing important factors for each application area will be crucial for the safe integration of AI technologies into mental healthcare practices. Doing so will help bridge the gap between the increasing demand for mental healthcare and limited available therapeutic resources, ultimately improving the accessibility and effectiveness of mental health services.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAI (Artificial Intelligence)\u003c/p\u003e\n\u003cp\u003eAIA (Artificial Intelligence Act)\u003c/p\u003e\n\u003cp\u003eCFA (Confirmatory factor analyses)\u003c/p\u003e\n\u003cp\u003eANOVA (Analysis of Variance)\u003c/p\u003e\n\u003cp\u003eCFI (Comparative Fit Index)\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eH (Hypothesis)\u003c/p\u003e\n\u003cp\u003eMDR (Medical Device Regulation)\u003c/p\u003e\n\u003cp\u003eRMSEA (Root-Mean-Square Error of Approximation)\u003c/p\u003e\n\u003cp\u003eRQ (Research Question)\u003c/p\u003e\n\u003cp\u003eSEM (Structural Equation Modeling)\u003c/p\u003e\n\u003cp\u003eSRMR (Standardized Root-Mean-Square Residual)\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTLI (Tucker Lewis Index)\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe protocol for this study was approved by the Ethics Committee of the University of Regensburg (23-3365-101).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAdditional supporting information can be found in the online appendix and on OSF (https://osf.io/9jxwy/?view_only=dff933d9f0234235bc51e61a6b439497).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Conflicting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe research was funded by a grant from the Volkswagen Foundation (Grant #: 98525).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’ Contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJ.C.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project Administration, Visualization, Writing – Original Draft\u003c/p\u003e\n\u003cp\u003eA.-K. K.: Conceptualization, Methodology, Supervision, Writing – Review \u0026amp; Editing\u003c/p\u003e\n\u003cp\u003eE.L.: Funding acquisition, Writing – Review \u0026amp; Editing\u003c/p\u003e\n\u003cp\u003eS.G.: Conceptualization, Methodology, Funding acquisition, Supervision, Writing – Review \u0026amp; Editing\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Anna Sigl for her help in the qualitative data analysis\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eWorld Health Organization. World Mental Health Report.: Transforming mental health for all [Internet]. www.who.int. World Health Organization, 2022. Available from: https://www.who.int/publications/i/item/9789240049338\u003c/li\u003e\n \u003cli\u003eMinerva F, Giubilini A. Is AI the Future of Mental Healthcare? Topoi [Internet]. 2023 May 31 [cited 2023 Jun 15]; Available from: https://doi.org/10.1007/s11245-023-09932-3\u003c/li\u003e\n \u003cli\u003eKellogg KC, Sadeh-Sharvit S. Pragmatic AI-augmentation in mental healthcare: key technologies, potential benefits, and real-world challenges and solutions for frontline clinicians. Front Psychiatry. 2022;13:990370.\u003c/li\u003e\n \u003cli\u003eBlease C, Locher C, Leon-Carlyle M, Doraiswamy M. Artificial intelligence and the future of psychiatry: qualitative findings from a global physician survey. Digit Health. 2020;6:2055207620968355.\u003c/li\u003e\n \u003cli\u003eKleine AK, Lermer E, Cecil J, Heinrich A, Gaube S. Advancing mental health care with AI-enabled precision psychiatry tools: A patent review. Comput Hum Behav Rep. 2023 Dec;12:100322.\u003c/li\u003e\n \u003cli\u003eAlles\u0026oslash;e RL, Thompson WK, Bybjerg-Grauholm J, Hougaard DM, Nordentoft M, Werge T, et al. Deep Learning for Cross-Diagnostic Prediction of Mental Disorder Diagnosis and Prognosis Using Danish Nationwide Register and Genetic Data. JAMA Psychiatry. 2023 Feb 1;80(2):146\u0026ndash;55.\u003c/li\u003e\n \u003cli\u003eKellogg KC, Sadeh-Sharvit S. Pragmatic AI-augmentation in mental healthcare: Key technologies, potential benefits, and real-world challenges and solutions for frontline clinicians. Front Psychiatry. 2022 Sep 6, 13.\u003c/li\u003e\n \u003cli\u003eYadav U, Sharma AK. A novel automated depression detection technique using text transcript. Int J Imaging Syst Technol. 2023;33(1):108\u0026ndash;22.\u003c/li\u003e\n \u003cli\u003eLee EE, Torous J, De Choudhury M, Depp CA, Graham SA, Kim HC, et al. Artificial Intelligence for Mental Health Care: Clinical Applications, Barriers, Facilitators, and Artificial Wisdom. Biol Psychiatry Cogn Neurosci Neuroimaging. 2021 Sep;6(9):856\u0026ndash;64.\u003c/li\u003e\n \u003cli\u003eAafjes-van Doorn K, Kamsteeg C, Bate J, Aafjes M. A scoping review of machine learning in psychotherapy research. Psychother Res. 2021;31(1):92\u0026ndash;116.\u003c/li\u003e\n \u003cli\u003eLin E, Lin CH, Lane HY. Precision psychiatry applications with pharmacogenomics: artificial intelligence and machine learning approaches. Int J Mol Sci. 2020;21(3):969.\u003c/li\u003e\n \u003cli\u003eCummins R, Ewbank MP, Martin A, Tablan V, Catarino A, Blackwell AD. TIM: a tool for gaining insights into psychotherapy. In: The World Wide Web Conference. 2019. p. 3503\u0026ndash;6.\u003c/li\u003e\n \u003cli\u003eFlemotomos N, Martinez VR, Chen Z, Singla K, Ardulov V, Peri R, et al. Automated evaluation of psychotherapy skills using speech and language technologies. Behav Res Methods. 2022;54(2):690\u0026ndash;711.\u003c/li\u003e\n \u003cli\u003eHirsch T, Soma C, Merced K, Kuo P, Dembe A, Caperton DD, et al. \u0026ldquo; It\u0026rsquo;s hard to argue with a computer\u0026rdquo; Investigating Psychotherapists\u0026rsquo; Attitudes towards Automated Evaluation. In: Proceedings of the 2018 Designing Interactive Systems Conference. 2018. p. 559\u0026ndash;71.\u003c/li\u003e\n \u003cli\u003eImel ZE, Pace BT, Soma CS, Tanana M, Hirsch T, Gibson J, et al. Design feasibility of an automated, machine-learning based feedback system for motivational interviewing. Psychotherapy. 2019;56(2):318.\u003c/li\u003e\n \u003cli\u003eSadeh-Sharvit S, Camp TD, Horton SE, Hefner JD, Berry JM, Grossman E, et al. Effects of an Artificial Intelligence Platform for Behavioral Interventions on Depression and Anxiety Symptoms: Randomized Clinical Trial. J Med Internet Res. 2023 Jul 10;25(1):e46781.\u003c/li\u003e\n \u003cli\u003eDawoodbhoy FM, Delaney J, Cecula P, Yu J, Peacock I, Tan J, et al. AI in patient flow: applications of artificial intelligence to improve patient flow in NHS acute mental health inpatient units. Heliyon. 2021 May;7(5):e06993.\u003c/li\u003e\n \u003cli\u003eBlease C, Locher C, Leon-Carlyle M, Doraiswamy M. Artificial intelligence and the future of psychiatry: Qualitative findings from a global physician survey. Digit Health. 2020 Jan 1;6:2055207620968355.\u003c/li\u003e\n \u003cli\u003eChen ZS, Prathamesh, Kulkarni, Galatzer-Levy IR, Bigio B, Nasca C, et al. Modern Views of Machine Learning for Precision Psychiatry . Patterns. 22 Nov 11;3(11):100602.\u003c/li\u003e\n \u003cli\u003eRajpurkar P, Chen E, Banerjee O, Topol EJ. AI in health and medicine. Nat Med. 2022 Jan 20;28(1):31\u0026ndash;8.\u003c/li\u003e\n \u003cli\u003eKleine AK, Kokje E, Lermer E, Gaube S. Attitudes Toward the Adoption of 2 Artificial Intelligence\u0026ndash;Enabled Mental Health Tools Among Prospective Psychotherapists: Cross-sectional Study. JMIR Hum Factors. 2023 Jul 12;10:e46859.\u003c/li\u003e\n \u003cli\u003eAafjes-van Doorn K, Kamsteeg C, Bate J, Aafjes M. A scoping review of machine learning in psychotherapy research. Psychother Res. 2021 Jan 2;31(1):92\u0026ndash;116.\u003c/li\u003e\n \u003cli\u003eChekroud AM, Bondar J, Delgadillo J, Doherty G, Wasil A, Fokkema M, et al. The promise of machine learning in predicting treatment outcomes in psychiatry. World Psychiatry. 2021;20(2):154\u0026ndash;70.\u003c/li\u003e\n \u003cli\u003eHummelsberger P, Koch TK, Rauh S, Dorn J, Lermer E, Raue M, et al. Insights on the Current State and Future Outlook of AI in Health Care: Expert Interview Study. JMIR AI. 2023 Oct 31;2:e47353.\u003c/li\u003e\n \u003cli\u003ePang TY, Lee TK, Murshed M. Towards a New Paradigm for Digital Health Training and Education in Australia: Exploring the Implication of the Fifth Industrial Revolution. Appl Sci. 2023 Jan;13(11):6854.\u003c/li\u003e\n \u003cli\u003eSharma A, Lin IW, Miner AS, Atkins DC, Althoff T. Human-AI Collaboration Enables More Empathic Conversations in Text-based Peer-to-Peer Mental Health Support. Nat. Mach. Intell. 2023 Jan 23;5(1):46-57.\u003c/li\u003e\n \u003cli\u003eGaube S, Cecil J, Wagner S, Schicho A. The relationship between health IT characteristics and organizational variables among German healthcare workers. Sci Rep. 2021 Sep 7;11(1):17752.\u003c/li\u003e\n \u003cli\u003eChoudhury A, Asan O. Impact of accountability, training, and human factors on the use of artificial intelligence in healthcare: Exploring the perceptions of healthcare practitioners in the US. Hum Factors Healthc. 2022 Dec 1;2:100021.\u003c/li\u003e\n \u003cli\u003eLo Presti A, De Rosa A, Viceconte E. I want to learn more! Integrating technology acceptance and task\u0026ndash;technology fit models for predicting behavioural and future learning intentions. J Workplace Learn. 2021 Jun 4; 33(8): 591-605.\u003c/li\u003e\n \u003cli\u003eHolden RJ, Karsh BT. The Technology Acceptance Model: Its past and its future in health care. J Biomed Inform. 2010 Feb;43(1):159\u0026ndash;72.\u003c/li\u003e\n \u003cli\u003eVenkatesh, Morris, Davis, Davis. User Acceptance of Information Technology: Toward a Unified View. MIS Q. 2003;27(3):425.\u003c/li\u003e\n \u003cli\u003eKnop M, Weber S, Mueller M, Niehaves B. Human Factors and Technological Characteristics Influencing the Interaction of Medical Professionals With Artificial Intelligence\u0026ndash;Enabled Clinical Decision Support Systems: Literature Review. JMIR Hum Factors. 2022 Mar 24;9(1):e28639.\u003c/li\u003e\n \u003cli\u003eFelmingham CM, Adler NR, Ge Z, Morton RL, Janda M, Mar VJ. The Importance of Incorporating Human Factors in the Design and Implementation of Artificial Intelligence for Skin Cancer Diagnosis in the Real World. Am J Clin Dermatol. 2021 Mar 1;22(2):233\u0026ndash;42.\u003c/li\u003e\n \u003cli\u003eGreenhalgh T, Wherton J, Papoutsi C, Lynch J, Hughes G, A\u0026rsquo;Court C, et al. Beyond Adoption: A New Framework for Theorizing and Evaluating Nonadoption, Abandonment, and Challenges to the Scale-Up, Spread, and Sustainability of Health and Care Technologies. J Med Internet Res. 2017 Nov 1;19(11):e367.\u003c/li\u003e\n \u003cli\u003eZhang X, Guo X, Lai K hung, Guo F, Li C. Understanding Gender Differences in m-Health Adoption: A Modified Theory of Reasoned Action Model. Telemed J E-Health 2013 Oct 26;20(1):39-46.\u003c/li\u003e\n \u003cli\u003eLiang Y, Lee SA. Fear of Autonomous Robots and Artificial Intelligence: Evidence from National Representative Data with Probability Sampling. Int J Soc Robot. 2017 Jun 1;9(3):379\u0026ndash;84.\u003c/li\u003e\n \u003cli\u003eKaya F, Aydin F, Schepman A, Rodway P, Yetişensoy O, Demir Kaya M. The Roles of Personality Traits, AI Anxiety, and Demographic Factors in Attitudes toward Artificial Intelligence. Int J Human\u0026ndash;Computer Interact. 2022 Dec 7;1\u0026ndash;18.\u003c/li\u003e\n \u003cli\u003eEitle V, Buxmann P. Cultural Differences in Machine Learning Adoption: An International Comparison between Germany and the United States. In:European Conference on Information Systems (ECIS). 2020. Available from: https://aisel.aisnet.org/ecis2020_rp/138/\u003c/li\u003e\n \u003cli\u003eSindermann C, Sha P, Zhou M, Wernicke J, Schmitt HS, Li M, et al. Assessing the Attitude Towards Artificial Intelligence: Introduction of a Short Measure in German, Chinese, and English Language. KI - K\u0026uuml;nstl Intell. 2021 Mar 1;35(1):109\u0026ndash;18.\u003c/li\u003e\n \u003cli\u003eMichie S, van Stralen MM, West R. The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implement Sci. 2011 Apr 23;6(1):42.\u003c/li\u003e\n \u003cli\u003eNickbakht M, Meyer C, Scarinci N, Beswick R. Exploring factors influencing the use of an eHealth intervention for families of children with hearing loss: An application of the COM-B model. Disabil Health J. 2020 Oct;13(4):100921.\u003c/li\u003e\n \u003cli\u003ePark J, Woo SE. Who Likes Artificial Intelligence? Personality Predictors of Attitudes toward Artificial Intelligence. J Psychol. 2022 Jan 2;156(1):68\u0026ndash;94.\u003c/li\u003e\n \u003cli\u003eWagner G, Raymond L, Par\u0026eacute; G. Understanding Prospective Physicians\u0026rsquo; Intention to Use Artificial Intelligence in Their Future Medical Practice: Configurational Analysis. JMIR Med Educ. 2023 Mar 22;9:e45631.\u003c/li\u003e\n \u003cli\u003eTer Harmsel JF, Smulders LM, Noordzij ML, Swinkels LTA, Goudriaan AE, Popma A, et al. Forensic Psychiatric Outpatients\u0026rsquo; and Therapists\u0026rsquo; Perspectives on a Wearable Biocueing App (Sense-IT) as an Addition to Aggression Regulation Therapy: Qualitative Focus Group and Interview Study. JMIR Form Res. 2023 Feb 1;7:e40237.\u003c/li\u003e\n \u003cli\u003eLi X, Jiang MY chao, Jong MS yung, Zhang X, Chai C sing. Understanding Medical Students\u0026rsquo; Perceptions of and Behavioral Intentions toward Learning Artificial Intelligence: A Survey Study. Int J Environ Res Public Health. 2022 Jul 18;19(14):8733.\u003c/li\u003e\n \u003cli\u003eKaraca O, \u0026Ccedil;alışkan SA, Demir K. Medical artificial intelligence readiness scale for medical students (MAIRS-MS) \u0026ndash; development, validity and reliability study. BMC Med Educ. 2021 Feb 18;21(1):112.\u003c/li\u003e\n \u003cli\u003eWang YM, Wei CL, Lin HH, Wang SC, Wang YS. What drives students\u0026rsquo; AI learning behavior: a perspective of AI anxiety. Interact Learn Environ. 2022 Dec 11;1\u0026ndash;17.\u003c/li\u003e\n \u003cli\u003eJohnson DG, Verdicchio M. AI Anxiety. J Assoc Inf Sci Technol. 2017 Sep;68(9):2267\u0026ndash;70.\u003c/li\u003e\n \u003cli\u003eWang YY, Wang YS. Development and validation of an artificial intelligence anxiety scale: an initial application in predicting motivated learning behavior. Interact Learn Environ. 2022 Apr 3;30(4):619\u0026ndash;34.\u003c/li\u003e\n \u003cli\u003eAlmarzouqi A, Aburayya A, Salloum SA. Determinants predicting the electronic medical record adoption in healthcare: A SEM-Artificial Neural Network approach. PLoS ONE. 2022 Aug 16;17(8):e0272735.\u003c/li\u003e\n \u003cli\u003eKwak Y, Seo YH, Ahn JW. Nursing students\u0026rsquo; intent to use AI-based healthcare technology: Path analysis using the unified theory of acceptance and use of technology. Nurse Educ Today. 2022 Dec 1;119:105541.\u003c/li\u003e\n \u003cli\u003eMcDonald T, Siegall M. The Effects of Technological Self-Efficacy and Job Focus on Job Performance, Attitudes, and Withdrawal Behaviors. J Psychol. 1992 Sep;126(5):465\u0026ndash;75.\u003c/li\u003e\n \u003cli\u003eAlQudah AA, Al-Emran M, Shaalan K. Technology Acceptance in Healthcare: A Systematic Review. Appl Sci. 2021 Jan;11(22):10537.\u003c/li\u003e\n \u003cli\u003eNgusie HS, Kassie SY, Chereka AA, Enyew EB. Healthcare providers\u0026rsquo; readiness for electronic health record adoption: a cross-sectional study during pre-implementation phase. BMC Health Serv Res. 2022 Mar 2;22(1):282.\u003c/li\u003e\n \u003cli\u003ePan M, Gao W. Determinants of the behavioral intention to use a mobile nursing application by nurses in China. BMC Health Serv Res. 2021 Mar 12;21(1):228.\u003c/li\u003e\n \u003cli\u003eZha H, Liu K, Tang T, Yin YH, Dou B, Jiang L, et al. Acceptance of clinical decision support system to prevent venous thromboembolism among nurses: an extension of the UTAUT model. BMC Med Inform Decis Mak. 2022 Aug 19;22(1):221.\u003c/li\u003e\n \u003cli\u003eFranke T, Attig C, Wessel D. A personal resource for technology interaction: Development and validation of the Affinity for Technology Interaction (ATI) scale. 2019.\u003c/li\u003e\n \u003cli\u003eWensing M, Paech B, Roth C, Schwill S. Learning, understanding and the use of information technology: a survey study among primary care physician trainees. BMC Health Serv Res. 2019 Oct 22;19(1):728.\u003c/li\u003e\n \u003cli\u003eLabinsky H, Ukalovic D, Hartmann F, Runft V, Wichmann A, Jakubcik J, et al. An AI-Powered Clinical Decision Support System to Predict Flares in Rheumatoid Arthritis: A Pilot Study. Diagnostics. 2023 Jan;13(1):148.\u003c/li\u003e\n \u003cli\u003eMaassen O, Fritsch S, Gantner J, Deffge S, Kunze J, Marx G, et al. Future Mobile Device Usage, Requirements, and Expectations of Physicians in German University Hospitals: Web-Based Survey. J Med Internet Res. 2020 Dec 21;22(12):e23955.\u003c/li\u003e\n \u003cli\u003eHekman DR, Steensma HK, Bigley GA, Hereford JF. Effects of organizational and professional identification on the relationship between administrators\u0026rsquo; social influence and professional employees\u0026rsquo; adoption of new work behavior. J Appl Psychol. 2009;94(5):1325\u0026ndash;35.\u003c/li\u003e\n \u003cli\u003eGillan C, Milne E, Harnett N, Purdie TG, Jaffray DA, Hodges B. Professional implications of introducing artificial intelligence in healthcare: an evaluation using radiation medicine as a testing ground. J Radiother Pract. 2019 Mar;18(1):5\u0026ndash;9.\u003c/li\u003e\n \u003cli\u003eLiu Q, Geertshuis S. Professional identity and teachers\u0026rsquo; learning technology adoption: a review of adopter-related antecedents. [Internet]. Available from: https://2016conference.ascilite.org/wp-content/uploads/ascilite2016_liu_full.pdf\u003c/li\u003e\n \u003cli\u003eJussupow E, Spohrer K, Heinzl A. Identity Threats as a Reason for Resistance to Artificial Intelligence: Survey Study With Medical Students and Professionals. JMIR Form Res. 2022 Mar 23;6(3):e28750.\u003c/li\u003e\n \u003cli\u003eGreenhalgh T, Abimbola S. The NASSS Framework \u0026ndash; A Synthesis of Multiple Theories of Technology Implementation. Stud Health Technol Inform. 2019 Jul 30;263:193-204.\u003c/li\u003e\n \u003cli\u003eKaraca O, \u0026Ccedil;alışkan SA, Demir K. Medical artificial intelligence readiness scale for medical students (MAIRS-MS) \u0026ndash; development, validity and reliability study. BMC Med Educ. 2021 Dec;21(1):112.\u003c/li\u003e\n \u003cli\u003eWessel D, Attig C, Franke T. ATI-S - An Ultra-Short Scale for Assessing Affinity for Technology Interaction in User Studies. In: Proceedings of Mensch und Computer 2019 [Internet]. Hamburg Germany: ACM; 2019 [cited 2023 May 19]. p. 147\u0026ndash;54. Available from: https://dl.acm.org/doi/10.1145/3340764.3340766\u003c/li\u003e\n \u003cli\u003eRammstedt B, Kemper CJ, Klein MC, Beierlein C, Kovaleva A. Eine kurze Skala zur Messung der f\u0026uuml;nf Dimensionen der Pers\u0026ouml;nlichkeit: big-five-inventory-10 (BFI-10) [A short scale to measure the five dimensions of personality: big-five-inventory-10 (BFI-10)]. Methoden, Daten, Analysen (mda). 2013;7(2)\u003c/li\u003e\n \u003cli\u003eSandelowski M. What\u0026rsquo;s in a name? Qualitative description revisited. Res Nurs Health. 2010 Feb;33(1):77\u0026ndash;84.\u003c/li\u003e\n \u003cli\u003eFereday J, Muir-Cochrane E. Demonstrating rigor using thematic analysis: A hybrid approach of inductive and deductive coding and theme development. Int J Qual Methods. 2006 Mar 1;5(1):80\u0026ndash;92.\u003c/li\u003e\n \u003cli\u003eRosseel Y. lavaan: An R Package for Structural Equation Modeling. J. Stat. Soft. [Internet]. 2012 May24 [cited 2024 Jul 3]. (J. Stat. Soft.). Available from: https://www.jstatsoft.org/index.php/jss/article/view/v048i02\u003c/li\u003e\n \u003cli\u003eBrowne MW, Cudeck R. Alternative Ways of Assessing Model Fit. Sociol Methods Res. 1992;21(2):230\u0026ndash;58.\u003c/li\u003e\n \u003cli\u003eHu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equ Model Multidiscip J. 1999 Jan;6(1):1\u0026ndash;55.\u003c/li\u003e\n \u003cli\u003eLumley T, Diehr P, Emerson S, Chen L. The Importance of the Normality Assumption in Large Public Health Data Sets. Annu Rev Public Health. 2002 May;23(1):151\u0026ndash;69.\u003c/li\u003e\n \u003cli\u003eCheung GW, Cooper-Thomas HD, Lau RS, Wang LC. Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations. Asia Pac J Manag. 2023 Jan 30;1(1).\u003c/li\u003e\n \u003cli\u003eHair JF, Anderson R, Tatham R, Black W. Multivariate Data Analysis. 5\u003csup\u003eth\u003c/sup\u003e Edition, Prentice Hall; 1998.\u003c/li\u003e\n \u003cli\u003eHair JF, Black WC, Babin BJ. Multivariate Data Analysis: A Global Perspective. 7\u003csup\u003eth\u003c/sup\u003e Edition, Pearson Education; 2010.\u003c/li\u003e\n \u003cli\u003eKim H, Ku B, Kim JY, Park YJ, Park YB. Confirmatory and Exploratory Factor Analysis for Validating the Phlegm Pattern Questionnaire for Healthy Subjects. Evid-Based Complement Altern Med ECAM. 2016;2016:2696019.\u003c/li\u003e\n \u003cli\u003ePucchio A, Rathagirishnan R, Caton N, Gariscsak PJ, Del Papa J, Nabhen JJ, et al. Exploration of exposure to artificial intelligence in undergraduate medical education: a Canadian cross-sectional mixed-methods study. BMC Med Educ. 2022 Nov 28;22(1):815.\u003c/li\u003e\n \u003cli\u003eBauer R, Glenn T, Monteith S, Whybrow PC, Bauer M. Survey of psychiatrist use of digital technology in clinical practice. Int J Bipolar Disord. 2020 Oct 3;8(1):29.\u003c/li\u003e\n \u003cli\u003eFiske A, Henningsen P, Buyx A. Your Robot Therapist Will See You Now: Ethical Implications of Embodied Artificial Intelligence in Psychiatry, Psychology, and Psychotherapy. J Med Internet Res. 2019 May 9;21(5):e13216.\u003c/li\u003e\n \u003cli\u003eGerger H, Nascimento AF, Locher C, Gaab J, Trachsel M. What are the Key Characteristics of a \u0026lsquo;Good\u0026rsquo; Psychotherapy? Calling for Ethical Patient Involvement. Front Psychiatry. 2020 Jun 2;11:406.\u003c/li\u003e\n \u003cli\u003eDoraiswamy PM, Blease C, Bodner K. Artificial intelligence and the future of psychiatry: Insights from a global physician survey. Artif Intell Med. 2020 Jan 1;102:101753.\u003c/li\u003e\n \u003cli\u003eKisely S, Scott A, Denney J, Simon G. Duration of untreated symptoms in common mental disorders: association with outcomes: International study. Br J Psychiatry. 2006 Jul;189(1):79\u0026ndash;80.\u003c/li\u003e\n \u003cli\u003eMarshall M, Lewis S, Lockwood A, Drake R, Jones P, Croudace T. Association Between Duration of Untreated Psychosis and Outcome in Cohorts of First-Episode Patients: A Systematic Review. Arch Gen Psychiatry. 2005 Oct 1;62:975\u0026ndash;83.\u003c/li\u003e\n \u003cli\u003eSun L, Yin C, Xu Q, Zhao W. Artificial intelligence for healthcare and medical education: a systematic review. Am J Transl Res. 2023 Jul 15;15(7):4820\u0026ndash;8.\u003c/li\u003e\n \u003cli\u003eLin E, Lin CH, Lane HY. Precision Psychiatry Applications with Pharmacogenomics: Artificial Intelligence and Machine Learning Approaches. Int J Mol Sci. 2020 Feb 1;21(3):969.\u003c/li\u003e\n \u003cli\u003eRobiner WN. The mental health professions: Workforce supply and demand, issues, and challenges. Clin Psychol Rev. 2006 Sep;26(5):600\u0026ndash;25.\u003c/li\u003e\n \u003cli\u003eSidhu N, Srinivasraghavan J. Ethics and Medical Practice: Why Psychiatry is Unique. Indian J Psychiatry. 2016 Dec;58(Suppl 2):S199\u0026ndash;202.\u003c/li\u003e\n \u003cli\u003eAmbrosi-Randić N, Ružić H. Motivation and learning strategies in university courses in italian language. Metod Obz Horiz. 2010 Nov 15;5(2):41\u0026ndash;50.\u003c/li\u003e\n \u003cli\u003eAlowais SA, Alghamdi SS, Alsuhebany N, Alqahtani T, Alshaya AI, Almohareb SN, et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ. 2023 Sep 22;23(1):689.\u003c/li\u003e\n \u003cli\u003eYoung AT, Amara D, Bhattacharya A, Wei ML. Patient and general public attitudes towards clinical artificial intelligence: a mixed methods systematic review. Lancet Digit Health. 2021 Sep;3(9):e599\u0026ndash;611.\u003c/li\u003e\n \u003cli\u003eVo V, Chen G, Aquino YSJ, Carter SM, Do QN, Woode ME. Multi-stakeholder preferences for the use of artificial intelligence in healthcare: A systematic review and thematic analysis. Soc Sci Med. 2023 Dec 1;338:116357.\u003c/li\u003e\n \u003cli\u003ePandey S, Elliott W. Suppressor Variables in Social Work Research: Ways to Identify in Multiple Regression Models. J Soc Soc Work Res. 2010 Jan;1(1):28\u0026ndash;40.\u003c/li\u003e\n \u003cli\u003eSherman RA, Nave CS, Funder DC. Situational similarity and personality predict behavioral consistency. J Pers Soc Psychol. 2010 Aug;99(2):330\u0026ndash;43.\u003c/li\u003e\n \u003cli\u003eCornelissen L, Egher C, Van Beek V, Williamson L, Hommes D. The Drivers of Acceptance of Artificial Intelligence\u0026ndash;Powered Care Pathways Among Medical Professionals: Web-Based Survey Study. JMIR Form Res. 2022 Jun 21;6(6):e33368.\u003c/li\u003e\n \u003cli\u003eFraser-Arnott MA. Evolving practices and professional identity: How the new ways we work can reshape us as professionals and a profession. IFLA J. 2019 Jun 1;45(2):114\u0026ndash;26.\u003c/li\u003e\n \u003cli\u003eJohnson M, Cowin L s., Wilson I, Young H. Professional identity and nursing: contemporary theoretical developments and future research challenges. Int Nurs Rev. 2012;59(4):562\u0026ndash;9.\u003c/li\u003e\n \u003cli\u003eKira M, Balkin DB. Interactions between work and identities: Thriving, withering, or redefining the self? Hum Resour Manag Rev. 2014 Jun 1;24(2):131\u0026ndash;43.\u003c/li\u003e\n \u003cli\u003eSchubert S, Buus N, Monrouxe LV, Hunt C. The development of professional identity in clinical psychologists: A scoping review. Med Educ. 2023;57(7):612\u0026ndash;26.\u003c/li\u003e\n \u003cli\u003eKoutsouleris N, Hauser TU, Skvortsova V, De Choudhury M. From promise to practice: towards the realisation of AI-informed mental health care. Lancet Digit Health. 2022 Nov;4(11):e829\u0026ndash;40.\u003c/li\u003e\n \u003cli\u003eMonteith S, Glenn T, Geddes J, Whybrow PC, Achtyes E, Bauer M. Expectations for Artificial Intelligence (AI) in Psychiatry. Curr Psychiatry Rep. 2022 Nov;24(11):709\u0026ndash;21.\u003c/li\u003e\n \u003cli\u003eKatznelson G, Gerke S. The need for health AI ethics in medical school education. Adv Health Sci Educ. 2021 Oct 1;26(4):1447\u0026ndash;58.\u003c/li\u003e\n \u003cli\u003eBretthauer M, Gerke S, Hassan C, Ahmad OF, Mori Y. The New European Medical Device Regulation: Balancing Innovation and Patient Safety. Ann Intern Med. 2023 Jun 20;176(6):844\u0026ndash;8.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1–2 and 6–9 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"mental healthcare, artificial intelligence, technology implementation, use intention, learning intention","lastPublishedDoi":"10.21203/rs.3.rs-4692251/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4692251/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e As mental health disorders continue to surge, exceeding the capacity of available therapeutic resources, the emergence of technologies enabled by artificial intelligence (AI) offers promising solutions for supporting and delivering patient care. However, there is limited research on mental health practitioners’ understanding, familiarity, and adoption intentions regarding these AI technologies. We, therefore, examined to what extent practitioners’ characteristics are associated with their learning and use intentions of AI technologies in four application domains (diagnostics, treatment, feedback, and practice management). These characteristics include medical AI readiness with its subdimensions, AI anxiety with its subdimensions, technology self-efficacy, affinity for technology interaction, and professional identification.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e Mixed-methods data from \u003cem\u003eN\u003c/em\u003e = 392 German and US practitioners, encompassing psychotherapists (in training), psychiatrists, and clinical psychologists, was analyzed. A deductive thematic approach was employed to evaluate mental health practitioners’ understanding and familiarity with AI technologies. Additionally, structural equation modeling (SEM) was used to examine the relationship between practitioners’ characteristics and their adoption intentions for different technologies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: Qualitative analysis unveiled a substantial gap in familiarity with AI applications in mental healthcare among practitioners. While some practitioner characteristics were only associated with specific AI application areas (e.g., cognitive readiness with learning intentions for feedback tools), we found that learning intention, ethical knowledge, and affinity for technology interaction were relevant across all four application areas, making them key drivers for the adoption of AI technologies in mental healthcare.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e: In conclusion, this pre-registered study underscores the importance of recognizing the interplay between diverse factors for training opportunities and consequently, a streamlined implementation of AI-enabled technologies in mental healthcare.\u003c/p\u003e","manuscriptTitle":"Mental health practitioners’ perceptions and adoption intentions of AI-enabled technologies: an international mixed-methods study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-12 15:18:24","doi":"10.21203/rs.3.rs-4692251/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0eed6cca-eb61-4575-970c-e6e965d286d5","owner":[],"postedDate":"July 12th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-07-12T15:18:25+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-12 15:18:24","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4692251","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4692251","identity":"rs-4692251","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}