Tourist Trip Recommender Systems (RSs) suggest points of interest (POIs) and combine them along enjoyable routes. Integrating public displays into the recommendation process promises to overcome the limitations of mobile devices, such as small screens, thereby enriching the user experience of a tourist trip RS. However, in practice, public displays are rarely integrated in this manner. In this paper, we show how a mobile RS for tourist trips can be adapted to public displays and propose a Distributed User Interface (DUI) approach where the RS is distributed among both public and private devices. The results of a preliminary user study indicate that integrating public displays is perceived as attractive and novel; however, people remain concerned about privacy issues when using a public display. Public displays become more interesting when used for group recommendations; thus, we outline how our proposed approaches can be integrated into a group RS.
• Information systems → Recommender systems; • Humancentered computing → Interaction techniques; Touch screens;
Recommender Systems (RSs) have been applied in various domains;
however they are particularly popular in tourism where they allow
users to receive suggestions for points of interest (POIs) or tourist
trips comprising multiple attractions [
Public displays at touristic spots are one solution to overcome
these limitations. Currently, public displays are used to display
primarily static content, such as maps and timetables. The next
step in public display research is to provide personalized content
tailored to individual information needs by allowing the user to
interact with the display and share their preferences. However,
using public displays for personalized recommendations raises
privacy issues, and some people are reluctant to use them because
of social embarrassment [
Our overall goal is to integrate public displays into tourism RSs.
Previously, we developed TourRec, a tourist trip RSs that
recommends sequences of POIs along enjoyable routes between a start
point and a destination and adapts the routes according to user
preferences and constraints [
The remainder of this paper is organized as follows: In Section 2 we present background information and related work. We introduce our approaches to integrate public displays into a tourist trip RSs and the results of a preliminary user study in Section 3. In addition, we explain how our ideas can be used in group RSs. Conclusions and suggestions for future work are given in Section 4. 2
Tourism RSs can recommend diferent travel-related items, such
as POIs, travel plans, and tourist trips [
In the tourism domain, RSs are typically developed for mobile
devices or desktop clients. Another idea to provide personalized
recommendations to people who are already traveling is deployment
1https://play.google.com/store/apps/details?id=de.tum.in.cm.tourrec
on public displays, such as information kiosks in trains stations,
airports, and touristic areas. Public displays vary in size from small
television screens to display static information, such as timetables,
to large and interactive multi-user wall displays [
Even though public displays have many advantages compared
to mobile devices, such as screen size, social embarrassment and
privacy concerns prevent people from interacting with them. People
may be uncomfortable entering sensitive data in a publicly available
device [
In this section, we describe TourRec and show how it can be adapted to public displays. We also introduce a DUI approach that combines private and public devices. In addition, we summarize the results of a preliminary user study performed to evaluate these approaches and explain how such approaches can be integrated into a group RSs. 3.1
TourRec is a mobile tourist trip RS that combines multiple POIs
along enjoyable routes [
To generate tourist trips based on user queries, TourRec communicates with a backend we developed for this purpose. The backend architecture is modular and scalable, which allows us to add and evaluate new clients, recommendation algorithms, and data sources. 3.2
Public displays have become increasingly common in touristic areas; however, they are still not used for personalized recommendations. The potential advantages are obvious. The user does not need their own device with Internet connection while traveling. Larger displays can facilitate orientation in an unknown area and support the selection of a suitable recommendation when all relevant data, such as POI information, a map, and context data, are displayed on a single UI. Furthermore, a public display can facilitate decision making when used by a group because the recommendation can be viewed by all members of the group. More advanced approaches allow the user to modify the recommendation directly on the public display and send it to the personal device. These advantages represent our motivation to integrate public displays into our TourRec application.
In this work, we use a kiosk system equipped with a 55-inch multi-touch screen in portrait orientation. Similar tourism information kiosks can be found in many touristic areas. We tried to keep the changes to the smartphone’s UIs to a minimum so that the only independent variable tested in our user study was the interaction type rather than other changes in the layout. Thus, the public display application applies the same layout but attempts to benefit from the larger display area wherever possible.
Figure 1b shows a tourist trip recommendation on the public display. Again, the final recommendation is presented both on a map and as a list of POIs. However, the public display variant takes advantage of the larger screen and displays both modes simultaneously. The map and list are displayed on the top and bottom of the screen, respectively.
We used the AngularJS framework to implement the public display application. The kiosk system runs Windows 10 and the application can be accessed via any web browser. 3.3
The DUI approach distributes the recommendation process among the smartphone and the public display. The two main reasons for this approach are: (i) users can keep sensitive data on their private device but view the recommendation on a large display, and (ii) users can prepare a route request prior to traveling and display a recommendation on a public display as required.
We decided to use a QR code for the pairing between the
smartphone and public display because it has been shown that this
method provides high usability in similar scenarios [
After the user formulates a route request, the extended smartphone application allows the user to send the recommendation to a public display. The user must scan the QR code using the smartphone’s camera to transmit the request to an intermediary server application we have developed. The public display fetches the route request from the intermediary server application. To identify the correct smartphone, each request is labeled with a unique ID that is also encoded in the QR code. After the public display receives the request, it forwards it to the backend and receives a recommendation, which is then presented to the user.
The smartphone and public display applications are the same as in Sections 3.1 and 3.2; however, they are extended by the pairing feature. The intermediary server application is a web service implemented in node.js.
We conducted a preliminary user study to obtain initial feedback on the proposed integrations of public displays.
3.4.1 Goals and Setup. We evaluated the three variants of the single-user RS relative to user experience, execution time of the selected tasks, and comfortability of use in a public space. The user study followed a within-group design. We allowed the participants to test the prototypes in random order to avoid biased results due to the learning efect. The participants were asked to execute three tasks for each interaction technique: (i) create a route between two predefined POIs, (ii) create a route between two predefined POIs with their own travel preferences, and (iii) create a route from the current location to a predefined destination.
The participants were asked to fill out a User Experience
Questionnaire (UEQ) after every interaction technique. The UEQ is a
semantic diferential with 26 items grouped into six user
experience aspects: attractiveness, perspicuity, eficiency, dependability,
stimulation, and novelty [
In total, 16 people participated in the user study. All participants were bachelor or master’s degree students or had recently graduated. Overall, the participants had rather limited experience with interactive public displays, e.g., 50% of the participants had never used a similar system previously.
3.4.2 Results and Discussion. We performed statistical tests where applicable to determine whether the performance of the interaction techniques difered significantly at the 5% level relative to any of the aforementioned aspects. We used Repeated Measures Analysis of Variance (ANOVA) when the results were distributed normally and the Friedman test in other cases. The Shapiro-Wilk test for normality was performed to select the correct significance test. In case of a significant diference, we performed a post-hoc test to identify where the diference occurred, i.e., between interaction techniques.
Figure 2 shows the results of all prototypes relative to the six UEQ aspects. As can be seen, the attractiveness of all prototypes is considered excellent, which means that it is among the 10% best results of the benchmark data set. However, perspicuity is significantly higher for the stand-alone smartphone mode compared to the DUI approach (α = 0.002). Many people are familiar with using smartphone applications. Hence, it is easier for them to get familiar with the stand-alone smartphone variant than a hybrid approach. For dependability, the diference between the stand-alone smartphone and public display modes is significant ( α = 0.006), which means that the participants felt more in control of the interaction when using a smartphone than a public display. Moreover, the public display’s dependability score was below average compared to the benchmark dataset because the public display scored very low for the Secure vs Insecure item. Thus, further efort to protect user data and prevent shoulder-surfing is required. Our DUI approach appears to be a promising solution because its dependability is similar to the stand-alone smartphone variant. Furthermore, the DUI approach demonstrates the highest novelty, which means that this approach feels the most innovative and creative. However, this diference is not yet significant.
Table 1 shows the average execution times for each task and prototype. The execution times of Task 1 are significantly shorter for both the stand-alone smartphone mode (α = 0.007) and the standalone public display mode (α = 0.015) than the DUI approach. Task 3, which requires the user to give the system access to their current location, is significantly faster on the smartphone than on the public display (α = 0.002) and for the DUI approach (α = 0.003). There is no significant diference between the execution times of Task 2 which included entering the travel preferences before requesting a recommendation.
The analysis of execution time shows that there is nearly no diference between the public display and smartphone interaction techniques. This is surprising because many participants had no previous experience with interactive public displays.
Comfortability using a smartphone in a public place is significantly higher than when using a public display (α = 0.005) and using the DUI approach (α = 0.005). During the study, 75% of participants explained that using two devices is a disadvantage and too complex because they could obtain the same recommendation using a single device. However, 25% of participants emphasized that preparing the route recommendation in advance, e.g., by entering route parameters on the smartphone, while waiting in line to use the public display could be a significant advantage in practical use. 3.5
The results of the preliminary user study show that integration of public displays into a tourist trip RS is perceived as attractive and novel. However, the advantages of a hybrid approach are less appreciated when used by single users. The feedback received indicates that public displays could become more valuable when a group of users attempts to agree on a tourist trip.
The simplest ways to find a group recommendation is to use only
a single smartphone or allow one group member to use the public
display on behalf of the group; however, this requires the group
members to agree on the group’s travel preferences in advance. A
more sophisticated approach uses one smartphone per user, thereby
allowing each user to independently enter their travel preferences.
In this case, the preferences of all users are merged automatically by
the RS using a social choice strategy [
When a public display is available, no Internet-connected
device is required. The public display variant presented in Section 3.2
can be used by a group if the group’s preferences are entered by
a single group member. Furthermore, we suggest an extension to
our DUI approach where each user enters their preferences using
their personal device and the recommendation is displayed on a
mutual screen. Thus, this approach uses the same UIs as the
previous prototypes. However, the collected preferences are aggregated
automatically before the public display shows the recommendation
to the group. This approach combines diferent advantages of the
previous solutions, i.e., users do not have to reveal travel
preferences and the mutual display facilitates discussions among group
members, which helps the group determine consensus [
To summarize, the strategies we suggest for a group RS can be distinguished by the following dimensions.
• Small screen vs. large screen 4https://developers.google.com/nearby/ • One user enters the preferences for the group vs. every user enters the preferences separately • User preferences are revealed to the group vs. preferences are hidden from the group • The recommendation is displayed on a mutual screen vs. the recommendation is displayed on each individual’s device Our goal is to compare these diferent approaches to determine which specifications under which conditions facilitate the process of finding a tourist trip for a group. For this purpose, we plan to conduct user studies with diferent group types, such as families, friends, or colleagues. The results will show us which approaches perform best relative to diferent usability criteria and if there are any diferences depending on the type of user and group. 4
In this paper, we have shown how public displays can be integrated into a tourist trip RS. We adapted a smartphone application to public displays and extended it with a pairing functionality to realize a DUI approach. In a preliminary user study, very high attractiveness was demonstrated by all approaches. However, the results of our preliminary study show that public displays provide limited advantage when used by single users. As a result, we have outlined how our approaches can be extended to enable group recommendations. In future, we will implement these extensions and compare the diferent approaches in larger user studies with real groups to evaluate how they support tourist trip recommendations in a group context.
Integrating public displays into a tourist trip RS ofers tourists
many advantages compared to mobile devices; however, privacy
concerns relative to using a public display remain. The DUI
approach presented in this work allows the user to keep sensitive
data, such as travel preferences, on the private device while
beneifting from the public display. This is particularly important in a
group recommendation scenario where the users want to share a
mutual display but not reveal their personal preferences to other
group members. However, further eforts are needed to protect
the data on the public display, such as the actual recommendation.
Diferent approaches to prevent shoulder-surfing, such as blacking
out parts of the display or mirroring a passerby’s position and
orientation to warn the user have been developed [
One main limitation of our preliminary study is the fact that it was conducted as a lab study. The advantages of a large display could become clearer when the user is actually traveling. On the other hand, social embarrassment and privacy concerns could become a significantly larger issue when passersby are present. Hence, in a real-world scenario, the users could accept the presented DUI approach more easily compared to the public display variant.
This work is part of the TUM Living Lab Connected Mobility (TUM LLCM) project and has been funded by the Bavarian Ministry of Economic Afairs, Energy and Technology (StMWi) through the Center Digitisation.Bavaria, an initiative of the Bavarian State Government.