A single-arm, 6-week, pre-post pilot intervention study was conducted. Data were collected from June to October 2023. The study population consisted of children and their teachers. Participants were recruited at 3 primary schools in the Netherlands: a special education school, a school in a low socioeconomic status neighborhood, and a school in a higher socioeconomic status neighborhood. In the Dutch primary school system, 1 teacher typically provides instruction to the same group of children throughout the entire week. Children remain in the same classroom, and the teacher covers most subjects. Grades 5, 6, and 7 generally include children aged 8‐10, 9‐11, and 10‐12 years, respectively. Children were eligible when they (1) attended grades 5, 6, or 7; (2) were able to speak, read, and understand the Dutch language; and (3) when their parents or guardians mastered the Dutch language to understand the information letter and informed consent form. Teachers were eligible to participate when they were teaching the participating children.

The Medical Ethical Committee of the Leiden University Medical Center in the Netherlands approved the protocol of the study (decision reference number WSC-2023‐05). According to the Medical Ethics Committee of Leiden University Medical Center, the study did not fall within the scope of the Dutch Medical Research Involving Human Subjects Act (23‐3026). All procedures were conducted in accordance with institutional guidelines and the Declaration of Helsinki. Informed consent was obtained from all parents of participating children before data collection. Participation was voluntary, and participants were free to withdraw at any time without consequence. Data were anonymized before analysis. No personal identifiers were retained, and access to the data was restricted to the research team members to ensure confidentiality. Participants did not receive individual compensation. As part of the study activities, the participating classes received an educational game as an incentive. No images or supplementary materials included in this paper allow for the identification of individual participants.

Each child individually followed a psychological training developed by Therapieland, a Dutch organization that offers evidence-based digital solutions for mental health care, focusing on one or two of the following themes: (1) self-image, (2) social skills, and (3) thinking, feeling, and doing. A short description of each theme can be found in Textbox 1, and a more elaborate description in Multimedia Appendix 1. The training was offered via a robot. The robot delivered the training by speaking text and asking questions to help children reflect on the module’s content, make connections to their own experiences, and create assignments for the upcoming week, as well as by showing educational videos via a tablet connected to the robot. An overview of the setup can be found in Figure 1. The child responded verbally to the robot or answered the robot’s question via the tablet. The robot provided reinforcing feedback on the process (eg, “Thank you for your answer”).

The theme(s) the child worked on was based on the needs of the child as determined by either the parent(s), teacher, and/or researchers (see Procedure section). Within each theme, the teacher, child, and/or researchers decided which specific sessions were done (see Procedure section). An overview of the themes and an example of underlying sessions can be found in Figure 2. The child completed 1 session with the robot per week, for a total of 6 weeks. At the end of each session, the robot gave the child an exercise to complete at home and in the classroom (eg, compliment others this week and write down how it went). In the following session, the robot inquired whether the child practiced with the exercise and how they felt about it. The robot did not provide content-related feedback on the exercise but delivered process-related feedback, such as “yes it was a difficult exercise” or “let’s practice some more.”

Teachers of grades 5, 6, and 7 of the 3 participating schools received a digital information letter; researchers visited the school for a presentation of the robot and the research and invited teachers to participate. Interested teachers were asked to sign the informed consent form. Teachers were provided with an explanation of the intervention and procedure. Children were recruited through the participating teacher; they received an informative flyer. In addition, parents of the children received an official invitation through a school newsletter, a school app for parents, or on paper. This invitation included a QR code for the information letter, screening, and digital informed consent form. In case of questions, the parents were instructed to contact the teacher or the researchers.

After a child was included, the teacher determined which theme(s) was most suitable for the child based on an existing checklist that is periodically completed by the teacher as part of a student tracking system. This system, already implemented in the schools, is used for tracking student progress, results, and behaviors. The checklist provides scores on the child’s self-confidence, self-image, and social interactions, and these scores were used as a guideline for the selection of themes. The teacher then provided the selected suggestions for the theme(s) to the parent(s), and parents could respond in a face-to-face meeting (which would then be scheduled) or via email. If parents did not respond, the teacher decided on the theme(s). In some cases, teachers did not have time to select the theme(s), and the researcher chose the themes instead. By using this approach, the intervention could be tailored to the individual needs of each child. Next, the teacher completed the baseline KIDSCREEN-10 for the child (see Measures section).

At the first session, the child was asked by the researcher to complete the baseline questionnaire on general mental well-being and self-efficacy (see Measures section). Next, the researcher introduced the robot, and the robot also introduced itself using an automated introduction text. The researcher explained that each theme comprised different sessions, and the teacher and child together decided which sessions the child would complete in that week. There were a couple of exceptions where researchers chose the sessions because the teacher did not have sufficient time to do so.

The child received 1 session per week, each lasting up to 20 minutes, over a period of 6 weeks during school hours. Each session took place in a dedicated room where a robot and a tablet were prearranged, along with any required materials such as forms. To preserve the autonomous character of the intervention, the researcher remained present but out of sight. Upon arrival, the child was welcomed by the researcher and invited to sit in front of the robot. Once seated, the researcher stepped aside, and the robot independently initiated the session. During the session, the robot guided the child through various components, including educational animations or videos, reflective questions, and weekly assignments. The researcher intervened only in case of technical difficulties or if the child encountered confusion regarding the next steps. At the end of each session, an evaluation form was completed by the child to assess their experience. Sessions lasted approximately 15 minutes, consisting of a brief welcome and optional recap (3‐4 min), followed by an educational component (4‐5 min), a reflection and assignment phase (3‐4 min), and a short evaluation (1 min). Most children were assigned to the theme Self-image (n=32), followed by Feeling, thinking, doing (n=20), and Social skills (n=17), and some were assigned to sessions from both Feeling, thinking, doing and Self-image (n=4). At the end of the 6 weeks, the teacher completed the postintervention questionnaire on general mental well-being for the child, and the child completed the postintervention questionnaires assessing acceptability, usability, general mental well-being, and self-efficacy. After all the children in the school had completed the intervention, the corresponding teachers of that school completed the postintervention questionnaires assessing feasibility, acceptability, and additional measures. All participants completed pre- and postmeasurements.

Descriptive analyses (eg, mean [SD], N, and percentages) were used to summarize the results regarding sociodemographic characteristics, feasibility, acceptability, usability, mental well-being (both children’s self-report and teacher-reported), and self-efficacy. To identify potential change over time, the total scores on the Child Outcome Rating Scale, KIDSCREEN-10, and 8-item Self-Efficacy Questionnaire for Children were statistically compared between pre- and postintervention using paired samples (2-tailed) t tests. Data were only analyzed in case a child completed 5 or more sessions. The analyses were performed using SPSS (version 25; IBM Corp) [25]. The free-text answers to the open-ended questions were summarized.

The results regarding mental well-being can be found in Table 3. The teacher-reported QoL of the children was significantly improved at postintervention as compared to baseline with a small to medium effect size. However, the children’s self-reported QoL did not significantly change over time and had a small effect size. In addition, no significant changes in general mental well-being and self-efficacy scores of children were found.

The pilot study primarily investigated the feasibility, acceptability, and usability of a robot-delivered mental health–promoting intervention within a primary school setting. The results showed that the intervention was generally perceived as moderately feasible and acceptable according to both teachers and children, although several challenges were identified with the implementation and integration of the intervention into the school setting. More specifically, teachers were not always able to find the time to select the most suitable themes for the children or to consult with the child and parents about which sessions within these themes he/she would be best suited to attend. Furthermore, addressing children’s questions regarding the content of the intervention sessions, as well as the amount of learning required to determine suitable intervention themes for the children, required a substantial time investment by the teachers.

The identified challenges with implementation and integration underscore the importance of considering practical barriers when introducing new programs into educational environments, such as primary schools. Both on a national and international scale, there is a shortage of teachers, leading to increased workload and corresponding limited resources [26,27]. Teachers’ limited time availability for tasks related to adequate implementation of mental health–promoting interventions highlights the need for streamlined and efficient intervention protocols that minimize the burden on educators. Possibly, the social domain could provide support through training and coaching or coordinating resources, thereby supporting educational institutions and their staff. Another way to address such implementation challenges within school contexts is to actively engage teachers and school personnel in the development and implementation process of such interventions to ensure that the intervention aligns with the school’s curriculum and resources. The significant time investment required from teachers to manage children’s inquiries about the robot and the intervention content furthermore highlights the importance of carefully designing such interventions in co-creation with the target population, so that such programs meet their needs and abilities effectively. Another possibility is to enable the robot to select and conduct tasks with the child without the assistance of a human facilitator. Addressing these challenges is essential to ensure the successful integration and sustainability of mental health interventions within school settings, ultimately maximizing their potential benefits for children’s well-being.

Regarding the preliminary effects of the intervention on children’s mental well-being, the findings suggest that while the intervention may have a positive impact on teacher-perceived aspects of children’s well-being, its effects on self-reported measures and broader mental health outcomes may be limited or require further investigation. The latter has been highlighted as a future research topic as well in recent reviews on robot-delivered interventions in the field of mental health [14,15,18]. Furthermore, it should be noted that there was no information available about which specific sessions the children attended within the intervention themes. It is possible that different sessions could have different impacts; this aspect warrants consideration for future research. In addition to studying the effects of mental health–promoting interventions delivered by robots, a more thorough investigation of specific needs of children and teachers is needed, for example, using a qualitative study design exploring end-user needs in more detail.

A strength of this study was its multischool design. However, the small sample size and potential selection bias due to children being selected for the intervention by their teachers limit the generalizability of the findings. In addition, the fact that 2 of the 3 schools started the intervention after the summer holidays may have caused a bias when teachers completed questionnaires about the child, as their knowledge about the children’s well-being may have been limited at that time and improved over the duration of the pilot.

This study provides valuable insights into the feasibility, acceptability, and usability of a robot-delivered mental health–promoting intervention within a primary school setting. However, further research is needed to fully understand its potential benefits and address existing limitations in terms of the implementation of such interventions in the primary school setting. In addition, the burden on teachers is a crucial factor in the feasibility of the intervention within the primary school setting. Therefore, the robot must be intelligent and personalized, enabling students to use it independently.