The Future of Tourist Assistance

How accepted are social robots in the context of tourist information? A study in Rothenburg ob der Tauber provides new answers to this question.

• Overall, people are still afraid of interacting with robots. Without encouragement, people avoid initiating a conversation with a robot.
• Key drivers of overall satisfaction in the interaction with robots are the accuracy and novelty of recommendations, the robot’s perceived intelligence, and the usefulness of its advice. The robot’s likability plays only a minor role.
• A humanlike robot is rated notably higher in terms of likability. People find the human like version friendlier, more polite, more pleasant, and overall nicer to interact with.

In today’s fast-paced digital world, travelers are bombarded with countless options when planning their trips. From choosing the perfect destination to selecting local activities, the sheer number of choices can be overwhelming. This is where recommendation systems come into play, simplifying the decision-making process by curating options based on individual preferences. A new development in this space is the rise of social robots in places like tourist information centers. These robots don't just answer questions—they engage in meaningful, humanlike conversations, complete with speech technology and emotional expressions through gestures. These capabilities create a more intuitive and engaging experience for users.

However, there's still much to learn about how travelers respond to recommendations from social robots. Do travelers like interacting with robots, and do they value their travel recommendations? Furthermore, what should the interaction with the robot look like? According to the uncanny valley theory, people tend to enjoy interacting with robots that have humanlike qualities, but if a robot looks or behaves too human, it can lead to discomfort. So what’s the sweet spot? Do tourists prefer robots that feel more like humans, or do they enjoy a more robotic interaction? To explore these questions, we collaborated with the Tourism Information Office of Rothenburg ob der Tauber. We developed a social robot using the Furhat platform to provide travel recommendations and tested it with real travelers at the Rothenburg Tourism Information Office.

Considering the requirements of Rothenburg’s Tourism Information Office, the robot was designed to recommend activities tailored to different types of tourists and offer basic information on guided tours in both English and German. Focusing on cultural tourism, Rothenburg’s main attraction, the robot classified visitors based on their interest in cultural experiences and how much culture influenced their decision to visit. For instance, some tourists seek deep cultural engagement, while others prioritize entertainment. To ascertain this information, in the beginning of the conversation, the robot asked the visitor questions about their cultural interest. Based on the answers, it suggested activities that offer either rich cultural immersion or entertainment, ensuring a personalized experience for each visitor.

To explore the impact of different interaction styles, a team of researchers from NIM and the University of Applied Sciences Ansbach developed two distinct versions of the robot: a humanoid and a robotic one. The humanoid version offered a more dynamic experience, engaging in small talk, responding to insults, and providing varied answers. In contrast, the robotic version had a more mechanical appearance, spoke in a monotone voice, and delivered straight-forward, limited responses. This version didn’t accommodate off-topic questions or repeat information, offering a simpler, no-frills interaction. The comparison between these styles helped us understand which approach is more effective for creating satisfying tourist experiences.

Before launching the main study, a pre-test was conducted to fine-tune the social robot’s setup. Based on user and tourism service feedback, several adjustments were made to improve the experience. While research often suggests that robots should explain their functions up front, pretest participants found this process to be too long and bothersome. As a result, the robot's greeting was simplified, explaining its features only when asked. To improve efficiency, the path to obtaining city tour information was streamlined, and the robot's capabilities were expanded to include details on nearby ATMs, toilets, and small talk. Additionally, if the first recommendation didn’t resonate with the user, two more options were provided based on their profile. The humanoid version was enhanced with natural gestures and facial expressions, such as nodding to confirm statements, while conversational intelligence was added to allow the robot to remember and repeat user statements, making interactions more seamless

The experiment took place over two days at the Tourism Information Office in Rothenburg ob der Tauber in the summer of 2023, and 60 random tourists were invited to interact with the social robot. Participants experienced either a robotic or humanoid version of the robot, providing valuable feedback on their experience with the robot in different languages within a questionnaire. Only the most relevant questions were selected to minimize participants' time commitment.

The experiment offered valuable insights into how participants perceived the social robot in a tourism office setting. Overall, the robot’s intelligence and likability were rated highly, and its recommendation accuracy also received positive feedback. However, its ability to suggest novel and unique activities scored slightly lower. Most participants found the system easy to use, with 63% expressing satisfaction with the recommendations they received. When asked if they would use a similar system in the future, 58% of participants responded especially positively, showing interest in adopting such technology again. The perceived usefulness of the recommendations was rated in the middle range, with 67% feeling supported in finding activities they liked. However, only 8% of the respondents felt the robot influenced their preexisting holiday plans.

Key drivers of overall satisfaction were the accuracy and novelty of recommendations, the robot’s perceived intelligence, and the usefulness of its advice. Interestingly, the robot’s likability played only a small role in determining overall satisfaction.

The humanlike and robotic versions of the robot showed minimal differences in performance, with one major exception. The humanlike version was rated notably higher in terms of likability. People not only said they preferred it, but they also found the humanlike version to be friendlier, more polite, more pleasant, and overall nicer to interact with.

The comments offered by participants show that the robot was well received, with 13% praising the robot as "very friendly." Some participants, however, also expressed room for improvement. In particular, 10% felt the conversation wasn't efficient enough, considering their limited time due to existing plans. Additionally, 8% of participants expressed missing human contact during the experience.

Another key insight from the study is that while participants responded positively to receiving tourist recommendations from the social robot once they engaged with it, they were hesitant to approach it on their own. Without encouragement from the organizers, only a small number of people would have initiated a conversation with the robot independently.

Overall, the study indicates that tourists found both versions of the social robot recommender to be satisfying and effective. Most users also expressed that they would use such a service again. However, the recommendations had only a minor impact on their preexisting holiday planning. While the humanlike version was rated higher in terms of likability, this did not significantly influence users’ intentions to continue to engage with the system. The differences between the two versions were minimal, suggesting that a less humanlike design has little effect on the quality of the decision-making support provided.

• For managers, the study shows that social robots don’t need to be highly humanlike to be effective. Rather than investing in complex features, the focus should be on creating a functional and user-friendly experience. Since tourists may not approach the robot on their own, strategies like clear signage or staff prompting can help increase engagement. Overall, social robots are a valuable tool to enhance customer service by offering quick recommendations, though they may not change preexisting travel plans
• For consumers, social robots provide an easy and convenient way to receive personalized activity recommendations. While helpful for discovering local options, these robots may not greatly impact preexisting travel planning. They serve more as practical assistants for exploring activities rather than as full travel guides.
• The study suggests that society is becoming more comfortable with social robots in public spaces. Robots don’t need to be overly humanlike to serve a useful purpose, but the desire for human interaction remains. Social robots can be effective in offering practical help, but they should complement human services, not replace them entirely