1 Introduction

1.1 Social Acceptance and Navigation Style

In this user study, we will explore the social acceptance of wall-following navigation behaviour for indoor office-style environments with narrow corridors. This study involves testing the robot wall following behaviour for indoor environments with the intent of reducing the perceived complexity of the robots behaviour compared to the main navigation choice of shortest path. This mobile behaviour will embody a wide variety of beneficial social signals that have already been established in previous literature [1,2,3,4] to help specifically explore the navigation style and how this style can make robot behavior more transparent and easier to understand [5]. This user study involves an online video-based experiment in which participants are shown wall following and shortest-path navigation in an office environment in various scenarios that people are likely to encounter in a shared space with a mobile robot. It is intended that this work will provide greater insight into the translation of a common navigation method compared to an alternative that could be perceived as being more predictable, safe, and easy to move around, leading to the potential for higher levels of social acceptance. This work will also investigate the use of different driving sides to explore whether a robot that moves along a wall will have greater acceptance for robots that walk on the opposite side (i.e., against the flow of pedestrian traffic) compared to the same side as people (i.e. with the flow of pedestrian traffic). It was predicted that the opposite driving side will be rated more favourably than the same driving side. The following four research questions were investigated:

  1. 1.

    RQ1: Do participants report differences in being predictable, comfortable, and safe for Wall Following and Shortest Path behaviours?

  2. 2.

    RQ2: Do participants report differences for robot social attributes for Wall Following and Shortest Path behaviours?

  3. 3.

    RQ3: Do participants report preference differences for a robot to follow one side of the wall that matches their local traffic flow (left or right) in the country of residence?

  4. 4.

    RQ4: How is the perceived safety of the person around the robots behaviour predicted by metrics such as being predictable and comfortable?

1.2 Related Works

There is a need for a safe and effective navigation method that is socially acceptable to people. For mobile robots, planning of the path is critical to their function, and mobile robots must find an optimal path according to a specific criterion [6], such as the shortest distance to reach its chosen goal. How a robot moves around the environment is just as critical to consider as its overall social acceptability, and effective navigation methods can sometimes be at the expense of human activity or social norms [7]. For example, robots that use the shortest path method can unintentionally end up passing in front of the person’s natural walking path to reach its own goal based on its specific criterion. This can be counterproductive to creating harmonious pedestrian flow in human-based environments. Human-aware navigation planning, while keeping the shortest distance criterion, has attempted to add other optimisation terms, such as distance from people [1], if the robot is visible to humans [2], disturbance level around people [3] and future predicted motion [4]. Recently, learning-based methods have been proposed for human-aware navigation [8, 9], often with the shortest distance criterion embedded in the reward function. Such learning-based algorithms cannot easily explain why the robot takes a particular action [5], making it difficult for people to understand its intended path. As a result, human-aware navigation is often focused on increasing algorithm complexity at the expense of being explainable to the person, which can impact long-term use if people are uncomfortable or feel unsafe around the robot or its selected paths.

1.3 Socially Acceptable Behaviours

Social acceptance of mobile robots and their navigation behaviour are important to their successful integration. Social acceptance is affected by how people interpret and perceive mobile robot navigation. Understanding how the navigation method works can help increase perceived feelings of comfort and safety around the robot that is moving around within an indoor environment. Investigations of the social acceptability of different mobile robots have helped to identify and implement different parameters that shape how mobile robots should behave around people [10, 11]. Simulation studies [12] suggest that mobile robots should avoid rapid acceleration changes and their orientation should be aligned with the direction of motion. Other studies have found that robots should adhere to spatial distance zones in hallway scenarios [13], robots are more liked when approaching people from the right side [14], and robots that use a single hallway side can reduce the time for the robot to reach its intended goal [15, 16]. For mobile robots that could follow a person through an indoor setting, social acceptance was higher for autonomous navigation methods compared to manual control [17], and the direction follow was found to be more natural than the path follow [18]. Mobile robots that communicate their intended movements to the person were found to increase social acceptance [19, 20]. In a scenario where the mobile robot must pass between two people, people expected robots to exhibit social cues, so that people can better anticipate its intended behaviour and for the robot to be as non-disruptive as possible [21]. The ease with which robot behaviour could be understood increased perceived safety in a path crossing scenario tested using first-person videos [22], and through communication tools to indicate when the robot will yield to a pedestrian in the hallway [23]. Taken together, all of these parameters and preferences can help to inform how robots should act when moving around indoors, and should be included in the selected navigation method to increase the success of the robot operated in a busy space. The effectiveness of navigation and social acceptance must be taken into account for a successful integration into home and office spaces. Instead of adding more optimisation terms, mobile navigation might benefit from implementing a navigation method that is more easily understood and interpreted by people. To help create socially acceptable navigation, robots can be directed to perform a navigation behaviour that could be easier for people to understand, including learning how it will move in both new and common scenarios.

1.4 An Alternative Navigation Style to Shortest Path

In recent years, there has been an increasing exploration of wall-following behaviour for mobile robots as a suggested alternative approach to the traditional shortest path method. Wall-following navigation is inspired by bug algorithms [24], which involve the robot moving along the edge of the wall to maintain a safe distance from people [25]. The wall following navigation has initially been tested for the navigation of mobile robots using classification-based learning [16], random forest and genetic algorithms [26]. This method of navigation has not often been used in social settings with people compared to shortest-path navigation. A wall-following approach will allow the robot to maintain distance from the person and perform simple stop and wait functions if it encounters a person, helping to replicate beneficial parameters identified when exploring social acceptance. The intention of this navigation behaviour is to allow people to continue their walking trajectory as naturally as possible and increase the likelihood that people will understand where the robot is most likely to travel in an office environment. This includes being able to reach similar target goals as the shortest path criterion, albeit in a slightly longer time frame. This type of behaviour could lead to the robot being seen as more predictable, safe to be around, and acceptable to be in close proximity. This could then help to create more socially acceptable behaviours, simply by reducing the complexity of attempting to understand the robot and its next set of movements. As a result, wall-following behaviour may be more understandable to people, and therefore, rate higher as being perceived as more predictable and safer to be around for people who have limited experience with robots. Given that the wall following behaviour requires the robot to closely follow one side of the wall, this proposed method opens up a new research question around which wall side the robot should follow: the same side as people will often walk (i.e. with the flow of pedestrian traffic) or the opposite side (i.e. against the flow of pedestrian traffic). The side where people walk is a common path, but the robot could either follow directly behind them or potentially disrupt their natural walking pace by moving ahead of them. However, robots that move on the opposite side of the road as people will allow the robot to be more easily seen and continue its natural path with a lower likelihood of obstructing others, but may be perceived as less socially acceptable on the opposite side of the wall. Overall, the social acceptance of wall following behaviour for home and office-based environments has not been studied beyond initial capability tests of wall following behaviours as an alternative to shortest path navigation, nor the preferred driving side of a robot if this navigation method was deployed in busy office environments.

2 Methodology

2.1 Design

A between-group, repeated measures (4 \(\times \) 2) design involved showing a video set of Wall Following (WF) and Shortest Path (SP) scenarios in a counter balanced order [27]. Half of these video sets were horizontally mirrored, where the robot would be seen to be driving on either side of the hallway. This design method was chosen to eliminate the possibility of order effects where the wall following and shortest path behaviour may have been rated higher or lower due to being the first behaviour seen [27]. This design will also account for testing different robot driving sides in the analysis (left, right). Participants were assigned to one of four conditions:

  1. 1.

    Condition 1: Wall Following on the left (WF-L, Video 1), Shortest Path on the left (SP-L, Video 2)

  2. 2.

    Condition 2: Shortest Path on the left (SP-L, Video 1), Wall Following on the left (WF-L, Video 2)

  3. 3.

    Condition 3: Wall Following on the right (WF-R, Video 1), Shortest Path on the right (SP-R, Video 2)

  4. 4.

    Condition 4: Shortest path on the right (SP-R, Video 1), Wall following on the right (WF-R, Video 2)

The video set can be viewed here: Wall Following Left Side, Shortest Path Left Side, Wall Following Right Side, Shortest Path Right Side. An online questionnaire design was used to conduct this experiment. Video studies represent a common way to measure social acceptance of mobile robots in home and office environments [21,22,23, 28]. An online design can capture rapid insight into the navigation method from a large number of people around the world, increasing the external relevance of the results. Furthermore, it was important for the research question to include a sample of participants that would come from different countries that had routes and route allocations, such as between the United States (right) and Australia / United Kingdom (left) to see if the country of residence also had an impact on the outcome, and to avoid bias on the driving side [27]. The design choice of a video set also allowed people to evaluate the robots intended pathway from an external point of view with detailed text explanation embedded into the video set to see the robots intended behaviour over a longer period of time. This study represented an initial feasibility step as part of a multistage research process to better understand wall following behaviour from robots, including providing initial pilot data before transitioning to in-person testing.

2.2 Robot Platform and Navigation Scenarios

The Fetch Robot was selected for this study as it was deemed suitable to showcase path planning in indoor settings. Its anthropomorphic dimensions (1.49 m in height, 0.5 m in width) enabled the robot’s motions to be observed at eye level. Several navigation scenarios were chosen based on a selection of common events that were likely to be seen in a shared indoor space between humans and robots (see Sect. 2.4). These scenarios were presented in the video set to ensure that the person understood both the benefits and limitations of each navigation method, as well as how the robot would respond to different scenarios. The robot behaviour was represented using remote control to allow greater control over robot actions and as an initial validity test before fully building the autonomous version for in person testing [27], and the results would then inform the need to build and refine the autonomous wall following the variant of the behaviour. Tele-operation was implemented using remote control whilst out of view of the video recordings. Remote control was used to ensure that the robot’s behaviour was suitable for the intended task and best replicated its intended autonomous driving behaviour. The research team planned, controlled, and reviewed each set of recorded video clips of remote control behaviour. The team discarded any scenarios that did not meet quality control measures, such as the robot was veering off-course or there were irregular driving patterns when the remote control method was used. To create the shortest-path behaviour, the robot must reach the intended goal that was shown in each scenario. This included when the robot was heading in a straight direction or if there was one main obstacle on the path. This could include the corner of the hallway or a human. All goals were in the robot field of view, and these trajectory goals did not extend over multiple hallways or rooms. This allowed the viewer to understand the short-term goal of the robot’s trajectory.

2.3 Shortest-Path and Wall-Following Methods

2.3.1 Shortest-Path

Shortest Path: This method uses the most widely accepted navigation strategy in mobile robotics, which is to follow the shortest route [6]. In this method, the path the robot takes varies depending on the start and the goal positions, which means that it is not possible for bystanders to know the exact path the robot would take. This could include taking directions such as heading towards opposite corners of a hallway to reach the end goal.

2.3.2 Wall-Following

Wall-Following: In designing a method that contrasts to the shortest path method, the wall following behaviour involved the robot moving along the edge of the wall to maintain a safe distance from people [25]. Instead of following the shortest path, the robot would follow the wall as much as possible and the robot would stop and wait when it gets close to people rather than going around them. The intent was to exhibit behaviours that would be easily recognisable by bystanders, even when the robot’s goal and start positions differ. The following scenarios were presented in further detail, and each scenario is included in the video presentation of the navigation method. A cluster of scenarios was tested rather than showing only a single behaviour or event, to allow participants to evaluate the longer-term decisions made by the navigation system and how it would perform in many different scenarios.

Fig. 1
figure 1

Doorway Goal and Actions

Full size image
Fig. 2
figure 2

Corner Junction Goal and Actions

Full size image
Fig. 3
figure 3

Corridor Goal and Actions

Full size image

2.4 Doorway

The start and goal positions are shown in Fig. 1a. Wall Following Behaviour: The doorway scenario featured three versions: 1) approaching a closed door, 2) approaching an open door, and 3) approaching the doorway as a human will pass through the door. In these scenarios, the wall following robot slows to a stop when it approaches the doorway, remains stationary for longer when the door is open, followed by slowly moving past the doorway if no human movement is detected. In a last scenario where the human opens the door (Fig. 1c), the wall following robot remains stationary until the human has left the doorway and the door is closed, before proceeding along its path, which is consistent with the navigation behaviour to continue to follow the wall. Shortest Path Behaviour: The doorway scenario features two versions: 1) passing a closed door and 2) passing a doorway as a human will open and walk through the door. If it detects a human approaching its path by opening the door (Fig. 1b), it will seek to exercise obstacle avoidance and divert away from the doorway as it drives past, returning to its original path when the human is out of range.

2.5 Corner Junction

The start and goal positions are shown in Fig. 2a. Wall Following Behaviour: This scenario features two versions: 1) robot only and 2) the robot encounters a human on its path. In both scenarios, the robot aims to move adjacent to the wall in order to reach the goal position, and follow the path as shown above (Fig. 2c). The robot pauses before entering the intersection and turns left around the corner. As in the doorway scenario, if it detects a human approaching, the robot will stop in place and wait for the human to walk past (Fig. 2d) before continuing along its path of Wall Following until it reaches the goal destination. Shortest Path Behaviour: This scenario features three versions: 1) robot only 2) encountering a human walking past from robot’s left and 3) encountering a human turning the corner from robot’s right. In the first version, the robot takes the shortest path directly to the opposite side of the intersection (Fig. 2b). In the second version (Fig. 2c), the robot tries to exercise collision avoidance to divert its path around the human walking past, before re-aligning itself moving towards the goal when the human has passed. The third version is similar to the second, but with a different start position for the human.

2.6 Hallway

The start and goal positions are shown in Fig. 3a. This scenario features four videos for each robot behaviour: a human walking down a corridor may walk on the same or the opposite side as the robot, and in the same direction or in the opposite direction. Wall Following Behaviour: In all four versions (Fig. 3b, c), the robot’s trajectory is a straight line parallel to the wall from the starting position to the goal position. Shortest Path Behaviour: The robot’s initial path in all four versions is a straight line across the hallway from the starting position to the goal position (Fig. 3d, e).

2.7 Materials

Participants were asked to report their gender, age, country of birth, country of residence and their experience with robots, including a rating of their robot experience (0 = no experience, 10 = highly experienced), if they had interacted with a robot (yes/no), and if they had controlled a robot (yes/no). Preference scoring was collected across three questions to assess general perceptions of each robot behaviour. The responses were on a 0–10 scale (0 = Not at all, 10 = Extremely). These items were custom-made to assess the navigation behaviours being presented:

  1. 1.

    Comfort: How comfortable would you feel to walk to a destination around this robot?

  2. 2.

    Safety: How safe would you feel to navigate around this robot?

  3. 3.

    Predictability: How predictable would you find this robot’s behaviour?

  4. 4.

    Likely: How likely would you be to choose to have this robot around you?

The Robotics Social Attributes Scale (RoSAS) was used to collect social attributions to robot navigation behaviour. RoSAS is an 18-item scale by Carpinella et al. [29] with 3 subscales: Warmth, Competence, and Discomfort.

2.8 Procedure

Ethical approval was granted by the Monash University Human Research Ethics Committee ID: 27103. Participants were recruited to the 10-min online questionnaire using research ads for the general public and a listing on Amazon Mechanical Turk (MTurk) to collect responses from around the world. Only MTurk participants with at least an approval rating 80% and those with more than 50 completed human intelligence tasks (HIT) were allowed to complete the questionnaire. Local distribution to the general public involved social media posts and online notice boards. Prospective participants gave their informed consent via a digital consent form. Informed consent was obtained from all individual participants included in the study. Participants were informed through the consent form that they would be asked to observe various robot behaviours. All participants were shown both behaviours to ensure that observing only one behaviour in isolation would not produce differences between conditions, especially if the participants lacked prior knowledge of robots behaving in various ways. Participants were instructed to watch the video set corresponding to their designated condition when entering the questionnaire and were allowed to proceed only after the video had ended. Subsequently, participants were asked to complete a series of questionnaires about their perceptions of the robot and its behaviour (Video 1, V1). Following this, participants viewed the second video featuring a different navigation method and responded to an identical set of questions (Video 2, V2).

3 Data Analysis

3.1 Preliminary Analysis

Data were analysed using IBM SPSS Statistics 25. A total of 157 responses with no missing data were collected in 23 days of recruitment. Participants were required to pass a CAPTCHA checkpoint to start the questionnaire as a quality control measure of automated survey entries. A total of 23 outliers were identified and removed, leaving 134 cases for statistical analysis. The Shapiro-Wilk test indicated that the data of interest did not follow a normal distribution. The decision was made to proceed with ANOVA since the groups were similarly skewed and ANOVA can be fairly robust to some deviations of normality [30]. The “likely to use” question did not result in any differences in data analysis.

4 Results

The 143 participants gave their consent and met the criteria of being 18 years or older. There were 40 in Condition 1 (28%), 32 in Condition 2 (22.4%), 37 in Condition 3 (25. 9%) and 34 in Condition 4 (23.8%). A total of 86 people identified as female (60.1%), 55 as male (38.5%) and 2 as non-binary (1.4%). There was a wide range of ages from 18 to 68 years of age with a mean age of 36.52 (SD = 11.90). The primary country of residence was between the USA (n = 54, 37.8%), India (n = 52, 36.4%) and Australia (n = 33, 23.1%), with one from Italy, Singapore, the UK and Macedonia.

Fig. 4
figure 4

ROSAS Warmth

Full size image
Fig. 5
figure 5

ROSAS Discomfort

Full size image

4.1 Robotics Experience and Country of Traffic

The country side of traffic was derived based on their reported county of residence: 85 participants (59.4%) reported a left-hand traffic flow, and 58 (40. 6%) reported the right-hand side. Across a rating of 0 to 10 (no experience to highly experienced), most participants reported low levels of experience with robots (M = 3.82, SD = 3.16, Median = 3, Mode = 0). However, almost half of the participants had previously interacted with a robot (n = 80, 55.9%) compared to those who had not (n = 63, 44.1%). In total, 72 people (50.3%) saw the Wall Following behaviour first, and 71 people (49.7%) saw the Shortest-Path behavior first. In other words, 77 people (53.8%) saw the robot on the left-hand side of the hallway and 66 on the right-hand side (46.2%).

4.2 Gender, Age and Robotics Experience

Mixed Model ANOVAs found that there were no significant effects on gender or age on Wall Following (WF) or Shortest-Path (SP) scores (See Figs. 4, 5 and 6). Pearson’s correlation coefficients were performed to assess the linear relationship between robotic experience and dependent variables. There was a positive correlation between robotics experience and the following variables, showing that more experience with robotics lead to stronger opinions about the robot across all sub scales, including both positive and negative emotions:

  1. 1.

    WF RoSAS warmth scores, r(143) = .592, p <.001

  2. 2.

    SP RoSAS warmth scores, r(143) = .588, p <.001

  3. 3.

    WF RoSAS discomfort scores, r(143) = .454, p <.001

  4. 4.

    SP RoSAS discomfort scores, r(143) = .475, p <.001

Fig. 6
figure 6

ROSAS Competence

Full size image

4.3 Main Evaluation of Robot Behaviour

A set of Independent Sample T-Tests [31] were conducted to determine if Wall Following (WF: Left or Right) or Shortest Path (SP: Left or Right) affected scores on the robot being seen as predictable, comfortable, and safe for the first time point exposure to the robots behaviour. Wall Following was rated higher for Timepoint 1, except for the Likely score. The mean scores for comfort, safety and predictability are shown in Fig. 7.

Fig. 7
figure 7

Independent Samples T Test for Timepoint 1 Analysis for Wall-Following versus Shortest Path

Full size image

To follow on from Independent Sample T-Tests [31] to investigate the data across both video exposures, a set of Repeated-Measure Analysis of Variance (ANOVA)s tests [31] were conducted to determine if Wall Following (WF: Left or Right) or Shortest Path (SP: Left or right) affected scores for the robot to be seen as predictable, comfortable, and safe across both timepoint exposures to the robots behaviour.

There was a significant interaction effect on perceived comfort ratings with robot navigation behaviour, F (3,138) = 3.600, p \(=\) 0.015, \(\eta \)p2 = 0.075, See Fig. 9. There was a significant interaction effect on how predictable people considered the robot’s navigation behaviour, F(3,134) = 3.794, p = 0.012, \(\eta \)p2 = 0.073, See Fig. 8. There was a significant interaction effect for robot safety scores, F(3,134) = 2.816, p = 0.42, \(\eta \)p2 = 0.059, See Fig. 10.

Fig. 8
figure 8

Interaction Effect for Robot Predictable Scores

Full size image
Fig. 9
figure 9

Interaction Effect for Robot Comfort Score

Full size image
Fig. 10
figure 10

Interaction Effect for Robot Safety Score

Full size image

4.4 Prediction of Perception of Safety

Safety is critical to acceptance. Therefore, safety is a clear variable of interest for wider-scale deployment [32]. Linear regression was used to test whether scores for being predictable and comfortable could be used to predict the perceived safety of the person for the robot to navigate around the person using the shown robot navigation style. Two regression models were performed, one for the SP behaviour and the other for the WF behavior. For the SP behaviour, the overall regression via the enter method was statistically significant (R = .841, F(2,140) = 168.666, p <.001). The two factors were found to contribute significantly to the prediction: comfort (\(\beta \) = 0.643, p <.001) and predictable (\(\beta \) = 0.243, p < .001). For the WF behavior, the overall regression was statistically significant (R = .819, F(2,140) = 142.314, p < .001). The two factors were found to contribute significantly to the prediction: comfort (\(\beta \) = 0.609, p < .001) and predictable (\(\beta \) = 0.387, p < .001).

4.5 Investigation into Driving Side Preference

Mixed Model ANOVAs were conducted to investigate preference based on participant country of traffic and robot side of traffic. There were no significant differences in the main effect for the country traffic side compared to seeing the robot on the side or on the right side. The preference for the driver side of the robot did not match the country of resi- dence, nor did it have a significant impact on robot ratings.

5 Discussion and Design Implications

This experiment explored social attribution and robot evaluation scores based on two methods of robot navigation in an office scenario: travelling around the wall to navigate to the desired goal (Wall Following), or using the most efficient path to reach the desired goal (Shortest Path). Regardless of the presentation order of robot behaviours, Wall Following was on average rated higher compared to the Shortest Path navigation method for ratings of being more comfortable and predictable. There were no significant differences between the navigation styles on social attribution for the subscales of warmth, competence, and discomfort, showing that neither navigation style was reported more favourably. People did not seem to want the robot to match the side they drive on within their country of residence, providing an important design implication for mobile robots. For instance, it would be sufficient to design a single type of robot behaviour that navigates from left or right side, rather than selecting the side the robot navigates in the deployment country. This could be further investigated for specific scenarios, such as a higher volume of people in the hallway who demonstrate a clear flow of traffic, rather than a single person in the hallway. It was possible to fit a linear regression model to predict perceived safety of both the Wall Following and the Shortest Path behaviours from scores for perceived comfort and being predictable, drawing attention to helping to increase perceived safety of the person around the robot. Therefore, perceived safety may be heavily influenced by the robot being predictable to the person.

6 Limitations and Future Work

This experiment represents a step as part of a multistage research process to better understand the perception of mobile robots that follow a wall to navigate in office environments. There are some known limitations in the use of video-based studies to evaluate robot behaviour. Participants rated their perceived scores of the robot, and this could be subject to different interpretations of the robots role if seen in person. The video set also had a clear range of view to show the intended robot behaviour in the scenario rather than the limited viewpoint that people in the scenario would have themselves. A video set also allows people to view more of the scene to facilitate more time to observe and predict the robots potential trajectory based on the navigation style. This study also involved people rating the robots behaviour from an external camera viewpoint, with a research laboratory member who was aware of the intended trajectory of the robot before filming the scenario. To help mitigate this limitation, the video was designed to provide a neutral and direct approach to the task, and the actor was instructed to reach their goal with minimal hesitation towards the robot, helping reduce behavioural cues that participants may have rated against instead of rating their perception of the robots’ behaviour. The use of a video set does provide some notable strengths for exploring this research design. This format further supports the perception of social acceptance from mobile robots compared to other video-based studies on the topic [21,22,23, 28]. This includes the capacity to collect a broad sample from other countries and to provide further explanation of robot behaviour for each scenario presented in the video set. It is possible that the behaviour may score even higher on metrics such as being predictable for similar scenarios that have not yet been encountered, particularly if people can predict how the robot will move around. In relation to the experimental design, repeated exposure with the behaviour can help build confidence in how the robot works around people, which can lead to more favourable ratings after a longer exposure [33], but the use of changing the behaviour presented first was performed to help reduce this outcome landing on one navigation style. Lastly, more research should be conducted on how cultural differences affect people’s perceptions of the robot and its navigation behaviours, and this study will provide the basis for future work.

7 Conclusion

The aim of this study was to investigate the social acceptability of using Wall Following as a robot navigation strategy compared to Shortest Path. The findings indicated that individuals perceived the Wall Following strategy as being more comfortable, safe, and predictable, highlighting its use as a significant navigation technique for future investigations around social acceptance. This includes social attribution scores, such as compassion. The findings of this study have important implications for future mobile robot design, given the current popularity of Shortest Path planning algorithms that use distance as the main criterion. This includes the limited use and exploration of Wall Following behaviours for mobile robots. This study indicates that simple robot behaviour, such as wall following, might be more readily preferred by people with limited experience in robotics, but at the expense of robot efficiency for the robot to reach its end goal. This trade-off should be carefully considered, including what the intended preferred outcome of the robot, to be efficient or well accepted. Future studies could further expand this to include an in-person wall-following behaviour test to explore how these findings translate into in-person evaluations. This would help to distinguish whether the scores remain consistent for an in-person test or change once the participant is involved walking beside the robot. Furthermore, participants were only exposed to robot behaviours for a total of ten minutes, and opinions of how the robot moves may change with long-term exposure, including when the robot has a time limit to reach the goal or encounters an obstacle. Lastly, future mobile navigation behaviour might instead represent a blend of both behaviours with wall following when humans are detected and shortest path when no humans are in the known trajectory, but this would need to be further investigated to ensure that the mix of navigation styles does not cause more hassle or discomfort for people.