Investigating the effects of embodiment on presence and perception in remote physician video consultations: a between-participants study comparing a tablet and a telepresence robot

2.1 Presence in video consultations

The concept of Presence in HCI has been widely explored, particularly in the context of mediated communication technologies. Presence, as explicated by Lombard et al., ³⁵ is defined as the “perceptual illusion of non-mediation”, where individuals experience mediated environments or interactions as if they were unmediated. It was shown that Presence is an essential construct in healthcare, as it may improve communication and engagement. ³⁶ This construct can be further understood through its subcategories Telepresence, Self-Presence, and Social Presence. ³⁷

Telepresence refers to the extent to which a mediated environment is perceived as real or as part of the user’s physical space, as stated by Dourish. ³² In the context of video consultations, Telepresence can be understood as the physician’s or patient’s ability to feel present in each other’s physical or social space, despite the spatial separation. Dourish’s framework also connects embodiment with physical interaction, underscoring how strongly or weakly embodied systems – such as telepresence robots versus tablets – impact this sense of Presence. Harrison et al. ³⁸ further emphasize how technologies shape perceptions of space, embodiment, and interaction through the lens of three HCI paradigms, offering foundational insights for contextualizing Telepresence in healthcare communication.

From a social perspective, Social Telepresence focuses on the mediated simulation of physical co-presence and has been primarily studied about communication technologies such as video calls and telepresence robots. ³¹ This concept highlights the interplay between embodiment and interaction fidelity, particularly in scenarios involving remote actors. While some studies, such as, ³¹ compare the effects of various communication media on Telepresence, our study focuses specifically on video consultation systems, investigating how different levels of embodiment affect Social Telepresence. By doing so, it highlights the physician’s Presence in a remote medical context, addressing an underexplored angle.

Self-Presence, a subcategory of Presence, refers to how one perceives oneself in the virtual environment. ³⁷ As the video consultation in our research takes place in a real environment, this subcategory is not relevant to this study.

Social Presence, as outlined by Short et al., ³⁹ relates to the degree to which a medium facilitates a sense of human connection. Lee ³⁴ refined this concept, distinguishing it from Telepresence by focusing on the interpersonal aspects of mediated communication. In healthcare, this involves evaluating how effectively a medium enables the physician to convey empathy, trust, and attentiveness to the patient. For video consultations, Social Presence is critical because it impacts the quality of the physician-patient relationship. ⁴⁰ However, Social Presence has been interpreted in varying ways depending on the domain, and as such, this study adopts the definition used by Vu et al., ⁴¹ who contextualized it as the ability of technology to connect interaction partners in a way that mimics face-to-face interaction.

In this study, the participant is assumed to be located in a local environment simulating a medical institution, while the physician – the remote sender – is situated elsewhere. In such a setup, the Social Presence of the remote sender may be distorted when using a strongly embodied device, such as a telepresence robot, located in the local environment. This phenomenon can be explained by the term Dual Ecologies, which describes the creation of two ecosystems: one at the physician’s remote location and another at the participant’s local environment where the robot is present. ²⁹ These two ecosystems could distort the Presence of the remote sender, who in this case is the physician, when telepresence robots are used. This is why the Presence of the remote sender is perceived as higher for a weakly-embodied telepresence system like a tablet, than for a strongly-embodied telepresence system like a telepresence robot. ³⁰ This study investigates whether these phenomena, observed in prior work, persist in video consultations and how this affects the physician-patient dynamic.

Based on the aforementioned literature, the constructs Social Telepresence and Social Presence are used to assess Presence in this work.

2.2 Perception of video consultations

The Perception of Video Consultations is shaped by several factors, including the technological medium, interaction quality, and contextual elements of the healthcare environment. Drawing on the conceptual framework by Groom et al., ³³ we identify four key constructs influencing patient Perception of Video Consultations: Technology Acceptance, Satisfaction, Security and Privacy Considerations, and Trust.

Technology Acceptance is critical for understanding the willingness of patients and physicians to adopt teleconsultation systems. Based on Davis’s Technology Acceptance Model TAM, ⁴² perceived ease of use and usefulness are central to fostering acceptance. In the context of video consultations, the degree of embodiment in the communication medium may influence these perceptions, as suggested by prior studies. ³¹

Satisfaction is an indicator of a successful consultation ⁴³ and is a prerequisite for repeated usage of video consultations. Satisfaction in healthcare settings is strongly tied to interaction quality, which may be limited by reduced non-verbal communication and the inability to conduct physical examinations. ¹⁵ Our study investigates whether the embodiment of video systems can mitigate these limitations and enhance satisfaction.

Security and Privacy considerations remain significant barriers to the adoption of video consultations. ⁴⁴ Patients often express concerns about data protection, particularly when sensitive health information is transmitted via digital systems. This construct is not only a technical challenge but also a perceptual one, as trust in the system’s security features affects the overall physician-patient relationship. These concerns regarding Security and Privacy considerations also play a role in the judgment of a telepresence robot, ⁴⁵ highlighting the broader implications of it in digitally mediated healthcare interactions.

Trust describes taking risks in a relationship with the expectation that the other party will perform a certain action. ⁴⁶ It is a foundational construct in mediated communication and is particularly critical in healthcare settings. Groom et al. ³³ highlight the role of trust in fostering a positive perception of teleconsultations, framing it as a dynamic interplay of technological reliability, physician empathy, and patient expectations. Trust is further influenced by the embodiment of the communication medium; strongly embodied systems, such as telepresence robots, may disrupt trust by introducing unfamiliar or uncanny dynamics into the interaction. ³¹ Additionally, nonverbal cues such as eye contact play a key role in forming trust during social interactions. ⁴⁷ A study by Schwaninger et al. ⁴⁸ emphasizes that in healthcare and social robot contexts, trust is shaped by users’ desire for control over the system, clear understandability of its functionality, privacy assurances, and context-specific factors such as location.

By integrating the above constructs, this study aims to contribute novel insights into the representation of the remote physician in video consultations from a HCI perspective. While previous research has compared various communication media, e.g., ^{31 , 33} few studies have specifically examined the effects of embodiment on the Perception of Video Consultations. This study addresses this gap by investigating how varying levels of embodiment influence Perception in the context of teleconsultations from a HCI perspective.

2.3 Embodied information providers in health contexts

Since the reduction of effort and the need for physical Presence can theoretically be significantly reduced by an embodied system if designed and applied in the right way, recent research has explored robots as information providers in healthcare contexts already. ^{49 , 50}

This highlights the necessity for such systems like the one of Stoevesandt et al., ⁵¹ who used the humanoid Pepper robot for patient information in a hospital. However, their study focuses primarily on usability, rather than the interaction itself or the construct of Presence. The comparison of embodied and non-embodied representations of a technical solution in their work does not directly address the Presence or Non-Presence of humans, making their research less relevant to our focus on Presence in telemedicine settings.

In addition, Stoevesandt et al. use a well-known usability questionnaire alongside undisclosed survey questions that do not originate from an established questionnaire, which calls into question the scientific validity of their findings. The constructs and hypotheses to be evaluated using this part of the questionnaire cannot be verified or used to compare directly with our method.

3.1 Hypotheses

To address the research question and explore the effects of remote physician embodiment on Presence and Perception, we designed several hypotheses for our study which are explained in the following.

3.2 Study design

A between-participants study design was used to compare the tablet and the telepresence robot for video consultations. This design has the advantage of preventing participants from becoming accustomed to the scenario. ⁵² Since the study was designed as between-participants, each study participant experienced only one device. Participants who experienced the same device therefore formed a group. This is why we had two groups in the study, namely the telepresence robot group and the tablet group. The purpose of the study was to analyze how participants interacted with the system and their perceptions. It should be emphasized that the participants’ behavior was not the main focus of the analysis.

3.3 Participants

The recruitment of participants was done by approaching them in universities, sports clubs, and among relatives. This approach was chosen because previous experiments indicated that survey-based participant recruitment was not particularly successful. As a result, a total of eighteen participants (15 male and 3 female) took part in the study after providing written informed consent, with nine participants in each group. Despite the relatively small sample size, previous research has demonstrated that even with around 10 participants per group, significant differences can still be detected. ⁵³

The participants’ ages ranged from 23 to 77. Nine of the participants were students, three were pensioners and the rest were employees. Ten participants had previous experience with robots, although this experience tended to involve casual encounters with robots. Only two participants had previous experience with a telepresence robot. It is important to note that the participants are not fully representative of the broader population typically found in a medical context.

Each participant was randomly assigned to either the telepresence robot group or the tablet group using an online random number generator. ⁵⁴ No financial compensation was provided to the participants.

In addition to the study participants, 8 medically experienced people were recruited to act as physicians. These individuals included both medical students and students with experiences in the medical field.

3.4 Procedure

The study participants were first welcomed to the laboratory environment and then the experiment process was explained to them. The participants should imagine themselves as patients undergoing an abbreviated medical check-up in a medical practice. This check-up was done as part of a video consultation. Following the introduction, written informed consent was obtained from the participants. It is pointed out that participants had the option to discontinue the study at any time without facing any adverse effects.

The experiments began with the participants independently proceeding to the designated room, where the video consultation would take place. Once arrived, they entered their details, including first name, surname, and age, into the device (tablet or telepresence robot). For this purpose, a virtual keyboard appeared on the respective device, allowing the user to enter the required information by selecting the appropriate letters. Once the details were submitted, the participants were notified by text and audio to wait for the video consultation while the medically experienced person reviewed the details of the participant.

For participants using the telepresence robots, the robot moved to the position opposite the table (see Figure 1a) for the video consultation after they had entered their details. The tablet remained stationary throughout the experiment. The video consultation on the tablet was conducted via Skype, while for participants using the telepresence robot, the consultation was conducted through the integrated video conferencing software. The medically experienced person, playing the role of the physician, initiated the video consultation once they were ready. Participants then received a notification on their devices and accepted the incoming call to begin the consultation. During the consultation, all participants could see the physician’s face on the screen.

Figure 1:

Illustration of the experimental technical setup. (a) Participant has a video consultation with the physician via telepresence robot. The robot receives the measured heart rates from the activity sensor worn by the participant via Bluetooth Low Energy (BLE) and transmits them to the physician via the Message Queueing Telemetry Transport (MQTT). (b) The physician conducts a video consultation with a single participant and obtains their data via MQTT. (c) The participant has a video consultation with the physician via tablet. The tablet receives the measured heart rates from the activity sensor worn by the participant via Bluetooth Low Energy (BLE) and transmits them to the physician via MQTT.

The video consultation reflected a shortened medical check-up. This check-up was restricted to an interview consisting of general health questions and heart rate monitoring. The inclusion of heart rate measurement was due to the fact that, according to Jang et al., ⁵⁵ it is a function requested by medical users for telepresence robots. This is because the heart rate can provide valuable insights into a person’s state of health, for example by identifying cardiac arrhythmias. ⁵⁶ For the heart rate measurement, the experiment leader served as an assistant of the remote medically experienced person and placed an activity sensor around the wrist of the participant. The involvement of the experiment leader in this supportive role was intentionally designed to simulate a realistic healthcare setting, where an assistant might support the physician during procedures.

The video consultation lasted approximately 20 min, with a variation of 10 min, depending on the communication between the physician and the participant. After the consultation, the evaluation of the experiment began, which included an online questionnaire and an interview.

3.5 Setting

The experiments were conducted on a single floor of a technical university to ensure reproducibility. It is worth noting that conducting the experiments in a university setting, rather than in a medical environment, may have affected participants’ perception of the devices and the overall experience of the interaction. The medically experienced person conducted the experiments from their homes in a quiet room. As a telepresence robot, Temi (version 2) ⁵⁷ was integrated into the local environment. The tablet used for the comparison was the HTC Nexus 9. ⁵⁸ To measure the heart rate, the activity sensor Beurer AS 99 was used. ⁵⁹

The choice of Temi as the robot used, as opposed to the more anthropomorphic Pepper robot, was based on the fact that Pepper is unable to navigate effectively in many circumstances, as demonstrated by both our experiments and previous research. ⁶⁰ Navigation plays an important role in our use case, and Temi’s mobility sets it apart from tablets, which are entirely static and lack physical Presence. In addition, Pepper has been discontinued as a commercial product as of early 2024, meaning it is no longer a viable option for ongoing research. Our further research is to be based on the same type of robot, so we did not consider it reasonable to opt for a technology that is no longer available and cannot be reproduced or extended by other research groups that do not already own a Pepper robot.

Furthermore, humanoid robots like Pepper feature a human-resembling face, but studying the appearance of a remote physician as a virtual agent with its actions and emotions mapped onto the robot is out of scope for this research. Therefore, we decided to represent the physician’s face on the robot’s screen, replacing the robot’s face with that of the physician. As for Pepper, this would only be possible on the chest-mounted tablet, hence losing parts of the embodiment by making the robot look less human-like and not facing the user directly.

During the parts of the interaction when the physician was not telepresent on the robot, but the robot was working autonomously, we used speech output as well as an animated face as “mimics” shown on the screen. This created the perception of an interactive agent, even though this particular robot model lacks motor-driven movable limbs. Previous studies have shown that such non-verbal cues, like animated faces and speech, can effectively simulate Social Presence and agency in robots. ⁶¹

Figure 2 illustrates the experimental technical setup and the interaction of the participants with the devices. The technical description of the setting is shown in Figure 1. The heart rate and user details were sent from the corresponding device to the application of the medically experienced person via the network protocol Message Queueing Telemetry Transport (MQTT). To get the measured heart rate value of the activity sensor, Bluetooth Low Energy (BLE) was used. An Android application was developed for the tablet and another one for the telepresence robot. This application was used by the participants for entering their details, receiving of the heart rate value via BLE, and sending these data via MQTT. Furthermore, a user interface was developed for the medically experienced person to see the information of the participant.

Figure 2:

Comparison of interactions with the telepresence robot and the tablet during video consultations. (a) Telepresence Robot. (b) Tablet.

3.6 Data collection

Immediately after the experiment, each participant completed a questionnaire that included the evaluation of statements regarding the Presence and Perception of the video consultation. The statements were rated online using a 5-point Likert scale, where 1 corresponded to “strongly disagree” and 5 to “strongly agree”. It should be noted that these statements were originally written in English and translated into German. The composition of the questionnaire is explained in more detail below.

3.7 Data analysis

The collected data consisted of quantitative data obtained through questionnaires and qualitative data derived from interviews and video recordings. The quantitative data were analyzed using the SPSS software. Mann–Whitney U tests ⁶⁵ were conducted on the quantitative data to compare the two groups: the telepresence robot group and the tablet group. This test was chosen because the samples were independent and the data was not normally distributed. It is based on assigning a rank to each value of the groups and the number of times one group outranks the others is used to check for a significant difference. The null hypothesis of the test is that the distribution is the same in both groups.

Three different values are used to describe the Mann–Whitney U test: U value, z statistic, and probability p. The U value gives the number of times the rank is exceeded in a group and the smaller of these values is used. The z statistic shows the difference between the observed U value and the expected U value under the null hypothesis. Significance is indicated by the probability p, and in this study, the exact significance with a significance level of 0.05 was used due to the small sample size.

To better assess the evaluation of the null hypothesis, we also used the effect size in the evaluation of the quantitative data. The effect size, according to Fritz et al., ⁶⁶ describes what proportion of the variability in the results can be explained by the independent variable, which in this study is the level of embodiment. It is calculated for the Mann–Whitney U Test by dividing the z-statistic by the square root of the number of observations, and the absolute value of this result is used. According to Cohen ⁶⁷ when the effect size is expressed as a correlation coefficient (r), values of 0.1, 0.3, and 0.5 are considered to represent small, medium, and large effects, respectively.

Moreover, a correlation analysis ⁶⁷ was conducted to check for significant correlation between relevant factors. The interpretation of the resulting correlation coefficient is the same as for the effect size.

Once the quantitative data had been analyzed, the qualitative data was evaluated. For this purpose, interviews with each participant were conducted. For the audio recordings of the interviews we performed a top-down Thematic Analysis by Brown et al. ⁶⁸ to focus on a detailed analysis of the existing data. By doing this, we derived codes from the interview transcriptions and analyzed their relationships.

The analysis of the video recordings had the focus of identifying similarities and anomalies. Although no standardized procedure was used for video analysis, observations were made regarding anomalies in posture, gestures, and communication. These qualitative insights were interpreted to complement the results of the questionnaire.

To ensure the privacy of the participants, we anonymized and stored all data, including interviews, videos, and transcriptions, on a secure university cloud. Participants were informed in advance through accompanying documents about the anonymization process and how their privacy and data protection were ensured. Their data was only considered valid if they had given informed consent. It should be noted that no health-related data from the activity sensor was stored, and this setup is used solely as a showcase without tracking real health data.

4.1 Presence

Contrary to H1.1, the Mann–Whitney U test indicated no significant difference in Social Telepresence between the use of the telepresence robot (Mdn = 4.00, M = 4.44, SD = 0.53) and the tablet (Mdn = 4.00, M = 3.89, SD = 0.78) for video consultations, with U = 24.00, z = −1.59, p = 0.180, r = 0.04. However, the results suggest a tendency for higher ratings of Social Telepresence with the telepresence robot, as reflected in the medium effect size.

The interview responses differed between participants for Social Telepresence, but not between the groups. For example, some participants stated that the “physician seemed to be in the same room”, while others remarked “it’s different when I’m with [the physician] in person” due to factors such as physical proximity. These statements are representative examples and reflect perspectives shared by participants across both groups.

Although the participants’ behavior was not the main focus of the analysis, the video recordings provided supplementary insights into interaction patterns. For instance, most participants in the tablet group were observed keeping their heads lowered and holding the tablet in their hands, whereas participants in the telepresence robot group leaned forward in the chair with their heads in a neutral position. Nevertheless, the hands of these participants remained unused during the video consultation with the physician, with only one participant being observed to make extensive use of them.

In addition, contrary to H1.2, the Mann–Whitney U test indicated no significant difference in Social Presence between the use of the telepresence robot (Mdn = 5.00, M = 4.78, SD = 0.35) and the tablet (Mdn = 4.83, M = 4.74, SD = 0.22) for video consultations with U = 27.50, z = −1.19, p = 0.252, r = 0.28. It is important to note that the analysis of Social Presence was limited to the constructs of Co-Presence and Attentional Allocation, and did not consider other relevant factors, such as Perceived Message Understanding or Affective Interdependence.

The interview responses for Social Presence were often regarding the sub-construct Attentional Allocation in both groups. Three participants in the tablet group mentioned that they were “slightly distracted when the assistant came in; I was mentally distracted from the doctor and just looked at what the assistant was doing”. An additional participant in this group “was easily distracted by other things just because of the [heart rate sensor]”. In contrast, only one participant in the telepresence robot group stated that there was a distraction through the assistant. Attentional Allocation depends also on the physician. Some participants stated that “the physician looks at the heart rate [on the physician’s application] and then they were distracted” while others mentioned that the physician “was fully focused” on them.

4.2 Perception of video consultations

Contrary to H2.1, the Mann–Whitney U test indicated no significant difference in Technology Acceptance between the use of the telepresence robot (Mdn = 4.25, M = 4.21, SD = 0.61) and the tablet (Mdn = 4.17, M = 4.21, SD = 0.43) for video consultations with U = 38.50, z = −0.18, p = 0.880, r = 0.04. The majority of participants stated that they would prefer the video consultation “from home via my PC or my tablet”. This preference was not directly related to the experimental setup but reflects broader user expectations and habits.

However, when evaluating the telepresence robot specifically, only three participants in this group saw an added value of the robot compared to mobile devices. The added value was the usefulness for complex medical procedures such as symptom checking and “that the quality at home is not as good”. Moreover, some participants in the telepresence group experienced difficulty filling in their details form on the robot. Additionally, the recordings revealed that participants using the tablet completed the task of entering their details, on average, 20 s faster. This difference can be attributed to their familiarity with the device.

Contrary to H2.2, the Mann–Whitney U test indicated no significant difference in Satisfaction level between the use of the telepresence robot (Mdn = 4.33, M = 4.56, SD = 0.37) and the tablet (Mdn = 4.67, M = 4.63, SD = 0.42) for video consultations, U = 36.00, z = −0.42, p = 0.775, r = 0.10. It should be noted that both devices were recognized as highly recommended for video consultations. Notably, the majority of participants in both groups recommended video consultations for “examinations that do not require complex medical operation”.

Contrary to H2.3, the Mann–Whitney U test indicated no significant difference in Trust level between the use of the telepresence robot (Mdn = 4.50, M = 4.33, SD = 0.22) and the tablet (Mdn = 4.25, M = 4.17, SD = 0.50) for video consultations with U = 35.00, z = −0.50, p = 0.631, r = 0.11. Contrary to expectations, the video analysis revealed no significant differences in eye contact between the two groups. Participants in both groups primarily focused their gaze on the physician’s face.

Participants in both groups reported that their feeling of trust was mainly impacted by concerns that “data might be stolen” by cybercriminals or they do not know “where the data goes”. There was more frequent talk of data security. Data security could vary depending on the life situation. One participant stated that the publication of “their illness is not important as a pensioner, but it could be a recruitment criterion as a worker”. Some participants suggested using a dedicated application for video consultations.

When the video recordings were analyzed, it was found that no participant was reluctant to discuss illness or lifestyle. Additionally, it is noteworthy that only one participant closed the door of the room where the video consultation was conducted, while the other participants left it open throughout the session.

The results support H2.4, as the Mann–Whitney U test indicated no significant difference in Security and Privacy level between the use of the telepresence robot (Mdn = 4.33, M = 4.30, SD = 0.66) and the tablet (Mdn = 4.33, M = 3.93, SD = 0.75) for video consultations with U = 27.50, z = −1.19, p = 0.234, r = 0.28.

It could be seen that Security and Privacy considerations influenced the Trust level according to the interview responses. To further explore this, a correlation analysis was conducted using the quantitative data. The analysis revealed a significant positive correlation between Security and Privacy and Trust, r = 0.55, p < 0.05, suggesting higher security and Privacy considerations being associated with a higher level of Trust. This indicates a large linear relationship.

In order to further clarify the distribution of the quantitative data for the constructs of Presence and Perception of video consultations, a boxplot was created in Figure 3. This diagram illustrates once again that the distribution for Social Telepresence differs considerably between the telepresence robot and tablet than for the other constructs. It should nevertheless be emphasized that the central tendency expressed by the median is also the same for Social Telepresence, as can be seen in Figure 3.

Figure 3:

Boxplot showing the distribution of the presence constructs (Social Telepresence, Social Presence) and perception constructs (Technology Acceptance, Satisfaction, Trust, Security and Privacy) for the use of telepresence robot and tablet in video consultations, measured on a 5-point Likert scale with 1 = “strongly disagree”, 5 = “strongly agree”.

5.1 Summary and interpretation of results

This study explored the influence of different levels of embodiment influence on the Presence and Perception of video consultations from a HCI perspective. For this purpose, two devices were compared in a video consultation, namely a telepresence robot and a tablet.

To assess Presence, the constructs of Social Telepresence and Social Presence were measured. The results showed that the different levels of embodiment had no significant influence on the Presence of the remote physician. It was therefore not possible to validate the Dual Ecology phenomenon from previous studies ²⁹ in this study. However, this study was able to demonstrate that the strongly embodied telepresence system, the telepresence robot, tended to create a more natural way of talking to the physician during the video consultation. Participants using the telepresence robot did not need to tilt their heads down or hold a device in their hands, which is more similar to face-to-face communication. A tablet on a static stand might also support hands-free operation. However, its level of embodiment, as noted by Tanaka et al., ⁶² is expected to have only a small effect on Social Telepresence, making it less effective in fostering a natural interaction.

To assess the Perception of Video Consultations, the constructs of Technology Acceptance, Satisfaction, Trust, and Security and Privacy considerations were measured. The results showed no significant differences between the use of both devices. It can therefore be deduced that the level of embodiment does not have a significant influence on the Perception of Video Consultations. One reason for this conclusion in this study could be that novelty played a role in the judgment, as the telepresence robot was new to most participants. It was shown by Gonzalez et al. ⁶⁹ that novelty has an influence on the interaction and therefore on the Perception of video consultations. In addition, the Trust assessment was significantly influenced by Security and Privacy considerations.

5.2 Implications

This work helps to assess the use of telepresence robots in telemedicine. It showed that the Presence of the remote physician was not distorted by the strongly embodied telepresence system, the telepresence robot. It can therefore be concluded that, from the perspective of the emotional relationship, it does not matter which device is used for the interaction in the context of a video consultation. This contrasts with findings by Choi et al. ³⁰ Furthermore, no significant differences from a HCI perspective were observed between the devices in terms of Technology Acceptance and Satisfaction.

Moreover, the results suggest that Security and Privacy concerns are not device-dependent. This highlights the need for further efforts to reduce these concerns. One potential approach is the use of informed consent, ensuring participants are well-informed about the data collection. ⁷⁰ This approach was also used in this study, and the recordings showed that participants felt less anxious about discussing their health when informed consent was provided.

As a side observation, it was noted that the telepresence robot tended to support more natural communication compared to the tablet, fostering more effective and positive interaction, even though this was not the primary focus of the study. Future research could explore this phenomenon in greater detail.

5.3 Limitations and future research directions

Our study had limitations that affected its results. A limitation arose from the selection of participants, as they were not fully representative of the broader population. Moreover, all participants were open to undergo video consultation. Therefore, it would be of interest to explore the response of more representative participants towards the use of video consultation from a HCI perspective.

The measurement of Presence in the experiments did not show a significant difference between the telepresence robot and the tablet compared to previous studies. ^{29 , 30} This may be attributed to the use of a different questionnaire that was created prior to the full recognition of the Dual Ecologies phenomenon. Additionally, our questionnaire used a portion of the Networked Minds survey ⁶³ to evaluate Social Presence. Consequently, comparative experiments using a different questionnaire for Presence could explore whether the feeling of Presence differs between the telepresence robot and the tablet in a video consultation.

It is also worth noting that the non-significant differences between the use of the telepresence robot and the tablet might be attributed to their similarities in form and functionality. Future studies could investigate whether emphasizing the unique feature of each device – such as leveraging the robot’s mobility more effectively – could lead to different results.

Additionally, the experimental setting may have influenced the results, as the experiments was conducted in a university building rather than a medical practise. This environment could have impacted participants’ perception and sense of presence regarding the telepresence robot, preventing it form demonstrating its full potential. For instance, its mobility – one of its key features – was limited in the smaller rooms of the experiment compared to larger spaces, such as those typically found in hospitals.

Moreover, an assistant was present during the experiment to provide support, which may have influenced participant’s interactions and the overall perception of the devices. Future research could explore how the absence of such assistance affects these constructs.

Furthermore, it would be valuable to investigate whether higher levels of embodiment could lead to different results. Higher levels of embodiment might be achieved through the inclusion of expressive gestures and body poses. A study by Adalgeirsson et al. ⁷¹ demonstrated that a higher level of embodiment positively impacts engagement and likability, although it does not significantly affect co-presence. Future research could further examine the effects of higher levels of embodiment on Presence and Perception in Video Consultations.