Cycling with Robots: How Long-Term Interaction Experience with Automated Shuttle Buses Shapes Cyclist Attitudes

Understanding how people's attitudes and trust toward robots change over time based on prolonged interaction is crucial for their successful integration into real-world environments. However, long-term deployments of robotic systems in natural settings are relatively rare. In this work, we administered a survey to bicyclists (N = 94) who shared the road with automated shuttle buses for up to three years on their daily commute. The results of the survey show that cyclists with more interaction experience reported more negative attitudes toward the presence of the buses in the shared space, referring to their low usage as one of the main reasons. This finding underscores the significance of perceived usefulness in shaping long-term acceptance of automated services.


INTRODUCTION
Imagine cycling to work every day and encountering an autonomous shuttle bus (ASB) in the middle of the cycle lane causing you to swerve.How would you feel about this technology after several weeks, months, or years?Small, semi-autonomous shuttle buses (i.e., a type of "robots on the road" [8]) have received attention as feasible urban trafc solutions of the future due to their ability to provide individualized transportation services outside of regular business hours, as well as servicing remote areas of cities and meeting the mobility requirements of individuals with special needs [6,11].ASBs often operate in shared spaces where vulnerable road users (VRUs) like pedestrians and cyclists interact with them daily over a long time.Cyclists represent a unique VRU category as they travel with higher velocity than pedestrians, which requires them to make quicker yielding decisions while at the same time being more vulnerable than passengers of vehicles [1,9].
While cyclists and pedestrians already face the challenge of coordinating in trafc, the introduction of a third party into their shared space might cause disruptions to the current social system of VRU interaction [5,16], especially if the third party is an autonomous artifcial agent whose actions might be difcult to predict [21].Interaction conficts (e.g., stopping, swerving) could potentially reduce the VRU's acceptance of those systems in their shared space over time.Hence, there is a need to study the long-term efects of introducing ASBs into spaces shared on cyclists' attitudes and trust.Do people get used to the presence of the buses and come to accept and trust them (e.g., due to habituation and familiarity) or will their attitudes become less positive over time compared to their initial expectations (e.g., due to novelty efects)?
The present paper reports a study of cyclists' experiences from interacting with ASBs that have been in continuous operation on a university campus for over three years (see Figure 1).To investigate the efect of interaction experience on user trust and attitude, two groups of student cyclists who regularly interact with the ASBs on their daily commute to the university campus were compared: one group with an average of two years of interaction experience and one group consisting of students that had recently arrived on campus and had an average of fve months interaction experience.We administered a survey among the cyclists, assessing their general attitudes towards the buses and trust in the buses' ability to drive safely, and asking them to report potentially dangerous incidents involving the buses and to propose solutions to conficts that might arise between cyclists and the buses.The survey contained items with pre-defned response alternatives as well as free-text response formats allowing participants to elaborate on their answers, such as why they like or dislike the buses, and to provide detailed descriptions of incidents and design solutions to counteract them.

THEORETICAL BACKGROUND 2.1 Conficts in Shared Spaces
Conficts may arise when new technologies disrupt pre-existing socio-technical systems' previously seamless operation [5].The current balance between various road users, such as bicycles and pedestrians, may be disrupted with the introduction of ASBs if they are not properly designed for human interaction and therefore difcult to integrate into the socio-technical system [16].Such disruptions can happen on two levels: on the psychological level, where they may prevent the natural coordination between VRUs, and on the infrastructural level, where ASBs operating on bicycle lanes may cause trajectory conficts and dangerous swerving [17].
Although the perspectives of pedestrians and passengers of autonomous vehicles have received increased attention recently [7,12,18], cyclists' perceptions of the long-term integration of autonomous vehicles into their interaction space have received less attention [3,20].To close this research gap, we focused on cyclists as a unique VRU, as considering the requirements and experiences of all involved parties is essential to the successful integration of ASBs in shared spaces.

Long-Term Efects on Trust and Attitudes
Studies of long-term interactions with autonomous systems in natural settings are important but rare [10,19].Novelty and familiarity efects have been shown to exert a strong infuence over people's attitudes and behaviour towards autonomous systems such as humanoid service robots [13] or autonomous vehicles [15].When encountering an autonomous system for the frst time, people might have relatively high expectations and positive attitudes towards the technology (i.e., novelty efect).However, when expectations are not met, resulting disappointment and negative attitudes can afect user acceptance as well as interaction quality [13].In contrast, if the interaction meets the expectations, liking, acceptance, and trust in robots have been found to increase as people become more familiar with them [2,4,14].Hence, it is important to understand how attitudes toward autonomous systems change over time as people become more experienced in interacting with them.

METHOD 3.1 Sample
Recruitment was conducted twice: frst before the start of the 2023 fall semester when new students arrived at the university, targeting individuals with more experience interacting with the buses; and then again after the semester began, targeting individuals with less interaction experience with the buses.Ninety-seven participants were recruited via social media, fyers on campus and personal approach.Participants were informed about the survey before providing their consent to participate.Monetary incentives were not provided.Inclusion criteria were legal age and regular cycling on campus.Three participants in the inexperienced group did not meet the inclusion criteria.
The fnal sample size was 94.The experienced sample consisted of 50 university students and employees who had encountered the buses on their campus commute regularly for an average of two years (average of 728 days, = 383, 95% CI[618; 837]).The inexperienced sample consisted of 44 university students and faculty who had encountered the buses for an average of seven months (average of 206 days, = 374, 95% CI[92; 319]).As desired, the encounter duration difered signifcantly between the groups, (92) = 6.667, < .001,with a mean diference of 522 days or approximately 1.5 years of experience with the buses.Naturally, experienced participants were on average older ( = 26.32,= 8.08) than participants in the inexperienced group ( = 21.70,= 3.06), (64.34) = 3.57, < .001.
There were no statistically signifcant diferences in reported occupation, gender, or frequency of encountering the automated buses on campus between the experienced and inexperienced groups.Ninety percent of the total sample were university students, 6% university faculty, and 1% "other".Fifty-three percent identifed as women, 46% as men and 1% as "other".The majority reported encountering the buses on campus at least once a week (57%) or once per day (36%).

Environment and Equipment
The automated bus service operates at the main campus of Linköping University, Sweden, as part of a research project and the ride is free of charge (see https://ridethefuture.se/in-english/).ASBs and cyclists share the cycle path that stretches alongside the main campus buildings.The pedestrian path is adjacent to the cycle path and the two are not structurally separated, which leads to pedestrians sometimes walking on the cycle path.The feet comprises two EasyMile EZ10 Gen 2 vehicles.The buses have a maximum speed of 16 km/h and are equipped with automatic braking systems that detect nearby objects.They are supervised by a safety driver who monitors the autonomous operation of the buses and can intervene if required.Over the three-year deployment of the bus service, there have been no reported collisions or accidents.

Measures
The survey was hosted on a university survey tool (https://sunet.artologik.net)and took about fve to ten minutes to complete.Informed consent was obtained from every participant.The survey included the following questions about attitudes and trust in the safety of the buses.The questions started with "As a cyclist, . . ." and were answered in the format(s) indicated in parentheses.
(1) How do you feel about the presence of the automated buses on campus?(do not like them at all (-2), do not like them (-1), neutral (0), like them (1), like them a lot ( 2)) (2) How do you feel about the automated buses driving on the bicycle lane?(very bad, bad, neutral, good, very good, other; free text) (3) Do you think the overall presence of the automated buses on campus is good or bad? (good, because . . ., bad, because . . .; free text) (4) How much do you trust the automated buses to drive safely?
(not at all (-2), not very much (-1), partially (0), quite a lot (1), completely (2); less safe than a human driver (-1), equally safe as a human driver (0), safer than a human driver (1)) In addition, survey participants were asked to report any observed or directly experienced incidents involving the buses that were hazardous or potentially hazardous, and to state any potential solutions that they might have that could facilitate interactions between cyclists and the automated buses.

General Attitude
Experienced participants, whose average group answer was Neutral (Mdn = 0, IQR = 1) were generally more negative than inexperienced participants regarding the buses, whose average group answer was "Like them" (Mdn = 1, IQR = 1), U = 1392, z = 2.513, p = .012.Experienced participants were also somewhat more negative than the inexperienced participants about the buses driving on the bicycle lane, although the average answer was Neutral (Mdn = 0, IQR = 1) in both groups, U = 1392, z = 2.513, p = .012.
Concerning whether the cyclists experienced the buses' presence as good or bad, there was a signifcant diference in proportions of .21( 2 (1) = 4.2, = .039)between the experienced (48%) and inexperienced participants (27%) for selecting "Bad" as answer option.The most commonly stated reasons across both groups why the buses' presence on campus was good overall were related to the fulflment of transportation needs (14 participants), importance to research and education (7), and the promotion of technological innovation (7).The most commonly stated reasons why the buses' presence on campus was bad overall were related to shared space conficts caused by the presence of buses (26 participants), emphasizing their obstruction on the cycle path, potential dangers for cyclists and pedestrians, and the inconvenience they create in crowded areas.Moreover, fve participants in the experienced group referred to the low usage of the buses, for example stating that "Bad, they are in the road and nobody rides them", "Bad, don't think they are used by many people at all", and "Good, but a bit unnecessary, fun but hardly used, always looks empty".This suggests that the perceived usefulness of the buses might afect cyclists' long-term acceptance.

Trust in Safe Operation
The experienced and inexperienced participants reported a similar high level of trust in the safety of the buses.The average answer to the question to what extent they, as a cyclist, trust the self-driving buses to drive safely was Quite a lot in both groups (Mdn = 1, IQR = 1).Moreover, participants in both groups considered the buses to drive equally safe as a human driver, with the average answer in both groups being Equally safe as a human driver (Mdn = 0, IQR = 1).

Experienced Incidents
Six participants in the experienced group and one in the inexperienced group reported having seen or been involved in an interesting or (potentially) dangerous incident with the buses.Five of the seven reported incidents occurred in a crowded environment and one during slippery road conditions.The most commonly reported emotions experienced in connection with the reported incidents were irritation (4), frustration (4), and feeling worried (2).Moreover, one participant reported experiencing anger, one relief and one surprise.Three of the seven participants who reported having experienced an incident provided descriptions of the incidents (see Table 1).

Proposed Solutions
Thirty-two of the 94 participants proposed solutions to avoid conficts between cyclists and the buses.The most commonly proposed solutions in both groups (in order of the number of participants who proposed them) were to: introduce separate lanes for cyclists and buses, employ various external human-machine interfaces to improve communication from the buses to cyclists, widen the shared roadway, and make various improvements to the trafc environment infrastructure.See Table 2 for a complete list of proposed solutions and their frequency.

Do not like them 7%
Neutral 41% Like them 32% Like them a lot 20%

Inexperienced
Do not like them at all 8% Do not like them 20% Neutral 32%

Like them 36%
Like them a lot 4%

Experienced
Figure 2: Pie charts depicting the proportions of self-reported general attitudes toward the automated shuttle buses among interaction experienced (n = 50) and inexperienced participants (n = 44).

DISCUSSION
This study investigated diferences in attitudes and trust between two groups of cyclists with diferent levels of interaction experience with ASBs.Having the opportunity to investigate the efect of a 3-year exposure time of cyclists to these autonomous buses, we found diferences in attitudes between the groups of cyclists but not in their trust.
The results showed that one-third of the group with more interaction experience reported negative attitudes toward the buses in general.These participants emphasized the lack of practicality of the bus service, citing its low passenger usage.The experienced cyclist group also reported more incidents with the buses.The incidents typically involved a harsh braking bus on a crowded walkway, potentially indicating that the cyclists did not expect the buses to brake.Participants proposed infrastructural solutions (e.g., separate lanes, wider roads), design solutions (e.g., interfaces), or limiting the operation of the buses during lunchtime when the campus is crowded, to avoid conficts between cyclists and the buses.
These fndings underscore the importance of studying the perceptions and attitudes of people with signifcant experience interacting with autonomous systems deployed in real-world settings.Nevertheless, cyclists in both groups were positive regarding the presence of the buses on the university campus overall, referring to the fulflment of transportation needs, importance to research and education, and promotion of technological innovation as positive aspects of the service, and also perceived the buses as generally safe.

Practical Implications and Future Work
The study showed that cyclists can adapt quickly to autonomous systems in their vicinity, show high trust in those systems to drive safely, and did not report any accidents or other types of severe incidents involving them.The incidents that occurred were when the ASB braked suddenly and harshly which should be addressed in the motion pattern design of the ASB to make the buses' movements predictable and ensure a transparent system design.
Most of the participants in the experienced group reported negative attitudes towards the buses based on the low usage of the buses that mostly drive empty and hence take up space on the cyclist road without transporting people.The low usage might only become rienced users in our study stem from the low usage of the system and if cyclists would be more accepting of ABSs in their interaction space when more people take the bus.

Limitations
Although we checked for demographic group diferences, other infuencing factors such as attitudes towards robots and automated technology might have infuenced the results.It is possible that the new students on campus were less confdent in expressing their opinions towards the buses than the experienced students.These limitations could be addressed in future studies by assessing pre-existing attitudes towards robots and technology, as well as the tendency to provide socially desirable answers.To control for the infuence of extraneous between-group variables, a repeatedmeasures design would be feasible where one group of cyclists is followed over several months as they become more experienced with the ASBs.Although we described the buses as "automated" in the survey, we did not assess whether participants had perceived the buses as fully autonomous up to that point.It is possible that some of the participants perceived the buses as manually driven, given that a safety driver is present onboard.This could have infuenced the trust measures and needs to be taken into account when interpreting the results.Additionally, the bus deployment is advertised as a research project at the university instead of as a commercial service operated by a local public transport company.As some participants directly referred to the bus service as a research platform in their stated reasons for their positive attitude towards the buses, this needs to be considered for future deployments when the buses become incorporated into general public transport and such associations may fade.

ACKNOWLEDGMENTS
This work was supported by ELLIIT, the Excellence Center at Linköping-Lund in Information Technology (https://elliit.se/), and a Swedish Research Council (VR) grant (2022-04602) on "Social Cognition in Human-Robot Interaction".We would also like to thank Philipp Hock for the design of the title fgure.

Table 1 :
Descriptions of Experienced Incidents

Table 2 :
Proposed Solutions to Human-Vehicle Conficts