ShadowTouch: a Multi-user Application Selection Interface for Interactive Public Displays

Transcription

1 ShadowTouch: a Multi-user Application Selection Interface for Interactive Public Displays Ivan Elhart, Federico Scacchi, Evangelos Niforatos, Marc Langheinrich Universita della Svizzera italiana (USI), Faculty of Informatics Via Giuseppe Buffi 13, 6900 Lugano, Switzerland firstname.lastname@usi.ch ABSTRACT In order to make public displays more attractive, recent work has begun to explore the concept of multi-application systems that allow passers-by to interactively choose from a set of display applications. However, the use of traditional desktop approaches (e.g., menus or tiles) for choosing an application in such circumstances is limiting, in particular in multi-user setups where multiple passers-by can concurrently interact with different applications on different parts of the screen. In order to explore novel means of application control, we have developed and evaluated a multi-user application selection interface called ShadowTouch. ShadowTouch uses information from a depth-camera to present the available applications overlaid within the bounds of the users on-screen silhouettes (i.e., their shadow ). In this paper we present the design, development, and initial usability evaluation of ShadowTouch. application menus (pop-up or drop-down lists), featuring application names and/or icons as buttons for starting applications. Not surprisingly, docks and lists are also used for application selection in most public display systems today that offer interactive application selection. While docks and menus represent a good starting point, in particular due to the familiarity of users with such mechanisms, they nevertheless entail an important drawback when it comes to multi-user environments: they are designed for single user interaction and are typically located in a fixed area on the screen (e.g., bottom). In display systems that explicitly support multi-user interaction, such an approach makes application selection difficult. Author Keywords Public Displays; Interactive Applications; Dynamic Application Scheduling. ACM Classification Keywords H.5 Information Interfaces and Presentation: Miscellaneous INTRODUCTION In order to make public displays more attractive and appreciated in their environment, previous research has suggested to open them up to both content and control from passers-by [5]. While a number of studies have looked at how individuals and groups of users perform specific tasks on public displays [1, 11, 14], only a small subset of display systems actually allow display users to influence application scheduling, such as e-campus [7], Instant Places [10], and UBI-Hotspots [12]. In desktop and mobile environments, application selection is typically done with the help of selection bars (docks) and/or Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Permissions@acm.org. PerDis 15, June 10-12, 2015, Saarbruecken, Germany c 2015 ACM. ISBN /15/06... $15.00 DOI: Figure 1. The ShadowTouch Application Selection Interface Previous studies showed that a silhouette representation of users can help to attract and invite display users to interact with the screens [11]. Our idea is not only to use silhouettes of users to indicate their presence and ability to interact, but also to use the silhouette representation to offer display and application control mechanisms. The basic principle is to explicitly mark the display area that the user s body occludes with a shadow overlay, in which individual control elements (e.g., application icons) are then shown. While Vogel [21] has previously suggested to use this area for displaying private information, we instead use it to show a private control interface. In the design of such an interface, we want to achieve two goals: 1. to design an application selection interface that allows multiple users to simultaneously influence application scheduling on public displays.

2 2. to design an interface that will be as good as a traditional application selection interface for selecting applications by a single user. In this paper we present the design, development, and initial evaluation of the ShadowTouch interface. In order to understand its potential and usability for public displays, we performed a within-subjects experiment with both individuals and groups (pairs) of users in a lab study. In our experiment we integrated the novel ShadowTouch interface into a public display system that has been deployed at our university since February 2014 [6]. Users interacted both with the standard interface, which features an application menu bar, as well as the new ShadowTouch interface. RELATED WORK A number of projects pioneered the vision of open display networks, in particular integrating interactive applications with digital signage. One of the earliest systems is the e-campus system at Lancaster University, UK. e-campus is a university wide installation of over 30 public displays, featuring a number of interactive applications such as interactive maps, art installations, and e-channels, among others [3, 18]. The e-campus platform allows users to interrupt the regular presentation of applications and content using mobile phones. By setting the device Bluetooth device names, users can not only indicate which application they want to see on the display, but also provide additional personalization parameters [3, 4]. Instant Places is a display platform that features a number of web-based display applications, such as a presence application that shows the profiles of users around the displays, a poster application for publishing user contributed posters, and a pin application that shows content associated with the profiles of display users, among others [10, 15, 19]. As part of the Instant Places display network, Taivan et al. explored the use of mobile phones to create a queue of applications that were requested by multiple users. The study looked at the expectations of individual users in requesting applications and potential conflicts that could happen between the users [20]. In contrast to e-campus and Instant Places, which both require mobile phones for interaction, the UBI-hotspots system features touch display that allow for direct user interaction. UBI-hotspots is a network of indoor and outdoor public displays installed in the city center of Oulu, Finland [8, 13]. The displays combine standard digital signage content (images and videos) with interactive applications by dividing the screen estate into three areas: one showing digital signage ( UBI-channel ), one offering a (touch-enabled) application menu for a sizable number of interactive applications, and a third area for running the actual interactive application. Additionally, users can influence content presentation through their mobile phones [9]. At our university we have deployed a network of four touchenabled public displays that we call USI Displays [6]. Each USI Display features a set of 17 interactive applications. The main goals of the USI Displays are to explore the challenges of dynamic application scheduling that is distributed between Figure 2. A diagram of the overall interface architecture display owners and display users, as well as the acceptance of interactive displays by the various stakeholders. The system allows individual applications to request scheduling information about other running applications as well as to influence when and how other applications are presented on the screen through a set of scheduling API calls. This functionality allows for the easy integration of application selection interfaces and dynamically influencing application scheduling. Our ShadowTouch interface uses the Microsoft Kinect to detect and track passers-by interacting with the display. The use of a depth camera like the Microsoft Kinect as part of a public display interface is not new. Müller et al. used the silhouette representation of users to study how people notice and use interactive displays [11]. The authors showed that the shadow representation is more effective in attracting the attention of display users than avatar-like representations. The distance of users from a display can also be used to present different information on the display. Vogel and Balakrishnan demonstrated a system that adapts its content based on the distance and orientation of users [21]. The authors also suggested that the space on the screen occluded by the user s body can be used to show more personalized information. Brudy et al. used the occluded part of the screen by the user body to show personalized content [2]. By indicating the occluded and visible areas, the authors increased the awareness of users which information on the screen is visible to other display users and passers-by. Shoemaker et al. used the distance and the shadow representation of users to support interactions on large wall displays and visually filter information on displays when used by multiple users [17]. In contrast, we use the shadow area to show personal control interfaces. SYSTEM ARCHITECTURE Figure 2 shows the overall architecture of the developed interface. The system obtains the depth data from a Microsoft Kinect sensor, which we positioned above the touch-enabled display and that we connected to a Mac computer. The image data was processed using the SimpleOpenNI library, which provides access to raw depth data. The depth information is then processed by a Processing sketch. The sketch takes the depth data, identifies and separates silhouettes of display users, and streams the silhouettes information to a web part of the interface. The ShadowTouch interface runs as a separate web-based application on top of the main display process driving the USI displays on our campus [6]. The ShadowTouch application obtains information about other available applications on the

3 display system from the underlying scheduler. The underlying scheduler provides names and icons of all running applications, as well as when they are or can be scheduled for presentation. This allows ShadowTouch to dynamically adapt the list of applications available for launch. Depth Information The SimpleOpenNI library that provides access to the raw depth sensor also provides information about the number and position of people in front of the screen. We overlapped this information with the raw data to identify and differentiate between silhouettes of multiple users. Figure 3 shows the shadow information overlayed with information about areas occluded by different users and their IDs. For each area occluded by a user we calculate a central point, which we then use to arrange the icons of available applications on the display around. Figure 3. Visualization of the depth data Application Selection In the existing USI Displays system all applications are presented inside separate iframes [6]. The visualization part of the display scheduler organizes iframes into a stack where only the topmost iframe or iframes are visible to the users. We placed the shadow applications as the topmost element, thus making it always visible. This is illustrated in Figure 4. ShadowTouch has a transparent background that allows the second most top element or elements to be visible as well. The ShadowTouch interface is made constantly visible in order to allow display users to select and start applications. When a user selects an application using the touch interface by touching the application icon, the scheduler rearranges the iframes and positions the selected application to the second topmost layer. When an application is selected, or when the interface is touched slightly outside the shadow area, the shadow interface minimizes into a short grey bar at the bottom of the screen. The bar still follows the user but does not occlude the applications in the second topmost layer. When the users touches the bar, the shadow interface with all application icons reappears on the screen. STUDY In order to evaluate the usability of the ShadowTouch interface, we ran a brief user study in a controlled lab setting. Figure 4. Visualization of different elements of the interface Study Objectives The main goal of our study was to evaluate how our ShadowTouch interface compares to traditional application selection mechanisms both for individual users as well as for multiple users. Therefore, we had two objectives: 1. to design ShadowTouch to be as usable as our current selection mechanisms for individual users, and 2. to make ShadowTouch more usable for multi-user application selection on public displays than our currently available interface. In addition to usability questions, we also wanted to understand actual interface use and overall user experience of such a novel interface. Participants In total 25 participants evaluated ShadowTouch in our lab. 13 participants performed single user test and 12 participants performed group tests in pairs (6 groups). Out of the 25 participants, 5 were female and 20 male. Participants were recruited at our university using the snow ball sampling method, i.e., initial study participants were asked to recruit further participants. Participants were mostly from the Faculty of Informatics (24). The participants differed by the academic program they were involved in: 1 bachelor student, 9 master students, 8 PhD students, and 7 faculty and staff members. 18 participants had already interacted with the USI Displays that we deployed at our university, two of them did so regularly, at least once a week. The other 7 participants had not yet interacted with USI Displays. The average age of our participants was 27 years (Min = 23, Max = 35, SD = 3.72). Tasks During the experiment we asked participants to perform different tasks on the display. We had a list of eight tasks that participants should perform on the display. The initial set of tasks was identical for all individual and group experiments. Before each experiment we randomly chose two tasks and presented them as examples to the participants, in a random order. All tasks consisted of launching an application on the screen and performing a simple in-application task. We did

4 not limit the period of time for completing each individual task. The eight possible tasks were: 1 - select the Twitter application and read the latest tweet about the university; 2 - select the Picture Taking application and take a photo; 3 - select the Public Transportation application and check when the next bus leaves the university; 4 - select the Picture Gallery application and check the latest picture; 5 - select the News application and read the title of the latest news item; 6 - select the University Staff Directory application and check how many faculties are at the university; 7 - select the University Information application and check how many study programs are available; and 8 - select the Map and find the Informatics building. The applications for tasks 1 and 2 are able to run in the sidebar concurrently to the other applications, which usually take up either the entire screen or, which shown concurrently with the sidebar, about 3/4 of the screen real-estate (see Fig. 4). Apparatus For the experiments in our lab, we used a 46 touch enabled LCD display, operating in landscape mode. The display was attached to a Mac computer running an USI Display client with a set of eight different applications. The screen real estate was divided into two parts: a smaller sidebar display zone on the left and a bigger display zone that is left on the right side. The shadow representation of the users was created using a Microsoft Kinect sensor mounted above the screen and software using OpenNI, SimpleOpenNI library, and Processing. The lab setup that we used to perform the experiments is shown in Figure 5. For the single user interaction we asked participants to perform two randomly selected tasks. For example, we asked a participant to select the Twitter application and read the last tweet about the university and then to select the News application and read the title of the latest news item. For the multi-user interaction we asked both participants to simultaneously perform two randomly selected tasks. Due to lack of applications that can run in the sidebar display zone on the left, we always chose applications 1 or 2 for the participant on the left part of the screen. Te participant on the right side of the screen received one of the remaining tasks. Conditions Each participant, whether in individual or group experiments, performed the previously mentioned tasks using two different interfaces: the regular application selection bar that we use in our deployment at the university, or the ShadowTouch selection interface. We randomly selected the order of the interface used by both individual and group participants. Methodology and Procedure Before the participants performed the tasks, we introduced the system, the purpose of the study, and asked for their consent to record the experiment using a video camera. During the experiment, we asked the participants to think aloud about what they were doing and what their intentions were in performing the tasks. After the experiment, we asked participants to fill out a questionnaire and answer several semi-structured questions about their experience. The questionnaire first asked participants to provide demographic information about themselves and their previous experience with USI Displays, before prompting questions from the Software Usability Scale (SUS) questionnaire that we adapted for the interface [16]. The SUS consisted of 10 questions each with 5 point Likert scale. The SUS provides a single measure of usability that is between 0 and 100. A SUS score above 68 is considered above average. We used the SUS score as the main measure of the usability of our new interface. For the experiments we replaced system with interface in the questions. Figure 5. The laboratory setup for single and multi-user experiments EXPERIMENTS WITH SINGLE USERS Usability After the single user experiments, the bar interface displayed an average SUS score of 79.42, while the ShadowTouch interface had in average the SUS score of Both interfaces can be recognized as good and above average according to the SUS guidelines. However, for the single user experiments, the bar interface turned out to be, in average, more usable. A further analysis of the results reveals that the bar interface had better scores in all questions, both those with positive and negative tone. The average SUS scores for individual questions are shown in Figure 6. Individual participants preferred to use the bar interface (Q1), but they found it to be more complex (Q2) and more cumbersome to use (Q3 with positive tone and Q8 with negative tone). Participants found the bar interface to be harder to learn (Q4 and Q10), but thought it is quicker to learn (Q7). In terms of the integration of various functions (Q5) the bar scored better but with more interface inconsistency (Q6). Individual participants felt more

5 confident in using the bar interface than the ShadowTouch interface (Q9). In addition, we performed an analysis of variance with the SUS score for the two interfaces as dependent variable and the type of the interface (bar and ShadowTouch) as an independent variable. However, no significant main effect was found (F(1,28) =.621, p =.437, h 2 p =.022). Figure 6. Single User Experiments - System Usability Scores [P1] We can interact together? Okay, I see, that s a good point, now it makes sense, this is a very good idea actually [P6] We can do things separately, yeah this is nice. [P9] This is an interface where you can have many people interacting [... ] it s more interesting for two people... EXPERIMENTS WITH MULTIPLE USERS Usability After the multi-user experiments, the bar interface had an average SUS score of 71.96, while the ShadowTouch interface had in average the SUS score of Both interfaces can be recognized as good and above average according to the SUS guidelines. However, for the multi-user experiments, the ShadowTouch interface turned out to be more usable. The average SUS scores for multi-user questions are shown in Figure 7. Comparing to the results with individual participants, there was a change in questions 1, 2, and 9. Participants that interacted with the screen in pairs preferred to use the ShadowTouch interface (Q1), but they found it more complex for multi-user tasks (Q2). Also, participants found the bar interface to be less cumbersome to use (Q3 with positive tone and Q8 with negative tone). For multi-user tasks the bar interface was thought to be harder (Q4 and Q10), but slightly quicker to learn (Q7). In terms of the integration of various functions (Q5) the bar scored better but with more interface inconsistencies (Q6). Individual participants felt more confident using ShadowTouch than the bar interface (Q9). Experience The majority of participants that performed single user experiments liked the bar interface more than ShadowTouch. Those participants explained their choice by indicating that they preferred a static menu instead of a set of icons that move around the screen. [P3] The bar is more static, more organized... However, some participants liked the idea behind the ShadowTouch interface. They liked the feeling of a personal area on the screen, and the curiosity of having a dynamic shadow that reacts to their presence. [P11] It gives you the feeling that it is only for you [... ] it is more personal... Even though participants preferred to use the bar interface, all participants also reported that both interfaces were simple to use and intuitive. All participants were able to perform the experiments without any difficulties. [P7] There was no huge difference between those two... At the end of each experiment, we asked participants if they wanted to try to use the screen again and which interface they would prefer to use. All but one participant chose to try the ShadowTouch interface again. When a researcher joined the participant and the second shadow with application icons appeared on the screen, all participants realized the potential of the ShadowTouch interface for multi-user application selection. Figure 7. Multi-User Experiments - System Usability Scores Similarly to our single user analysis, we performed an analysis of variance with the SUS score as dependent variable and the type of the interface (ShadowTouch and bar) as independent variable. However, no significant main effect was found (F(1,22) =.356, p =.557, h 2 p =.016). Experience In contrast to our single user experiments, participants who took part in multi-user experiments preferred the Shadow- Touch interface over the bar interface. Participants realized

6 that multiple application selection choices allow them to interact simultaneously on the screen without waiting for other users to choose and select applications. [P14]... definitely the first one (ShadowTouch), because each of us have own possibility to launch an application. [P21] I kind like the first (ShadowTouch), because it was more efficient, we could both press some buttons at the same time. [P22] The first one (ShadowTouch) is better for multitasking. Unfortunately, our prototype did not support that all application were available on both sides of the screen. While the system is able to distinguish the positions of users and try to schedule the application in a respective region, only two applications were available on the left sidebar display zone, as mentioned before. After the experiments all participants wanted to try the shadow interface again. During the free to roam time on the display, there were situations when an application opened in the opposite side of the screen. The participants had a feeling that the part of the screen behind their shadow is personal as well. Participants on the left thought that their application would run always in the sidebar, while participants on the right expected them always in the right hand side region of the display. When this did not happen the way they expected, they tried to change positions with the second participant. One participant tried to select an application using the second participant s ShadowTouch, an action that provoked the second participant to defend his part of the screen: Don t touch my shadow, you have your own. [18] When we are standing all like this, two of us, and we have both the silhouette, I get the idea that sort of the left half is my screen and the right half is his screen [... ] If now I click on the Info and it opens here (on the right), I feel like argh! this is mine why it is opening in the other part of the screen? DISCUSSION From the experiments with individual users we found that while both interfaces have good usability scores and participants see the potential of the ShadowTouch interface, they still preferred to select applications through the bar interface. This could be for two possible reasons. First, more than two thirds (18) of our participants had used USI Displays with the same application selection bar at least once before the experiment. Also, due to the interface implementation, ShadowTouch was more reactive to the users movements (i.e., harder to control) than participants expected. The fact that all but one participant chose to use ShadowTouch again after the usability test and explore it more seems to indicate that participants in principle liked the interface. A revised ShadowTouch interface with slower, more predictable responses to user movements could hence be as good as the application selection bar for single user tasks. The experiments with multiple users and SUS scores seem to suggest that participants preferred to use the ShadowTouch interface and felt more confident in using it on the screen. All group participants wanted to try ShadowTouch again in the second round. However, handling application placement on our asymetrical display layout remained a challenge, as participants were struggling with applications that appeared on the opposite side of the screen. Also, without any long-term deployment, judging the influence of a novelty-effect on the interface s use is difficult. Prior work by Peltonen et al. [14] seems to suggest that multi-user interactions around a display usually follow a turn taking pattern, with the lead-user assuming control over the display. However, without adequate means of concurrent application selection control such as ShadowTouch, such patterns might simply be a by-product of current single-user application selection interfaces. CONCLUSION AND FUTURE WORK In this paper we describe our first attempt at building a novel application selection interface for interactive public displays that enables multiple display viewers to concurrently select applications. Using a depth camera, we create a virtual shadow of a close-by viewer and arrange the icons of available applications inside the viewer s shape. This allows the system to place application icons close to an actual user s position, without interferring with other user s display use. In an initial lab experiment looking at single-user interactions, we found no significant usability difference (based on SUS scorse) between our new ShadowTouch interface and the more traditional full-page menu. Moreover, a simple multiuser experiment found that participants appreciated the novel interface as an effective mean to concurrently select applications on a large display. We believe that such multi-user support is the main benefit of the ShadowTouch interface, yet our current setup with its rather small 46-inch screens did not allow us to properly test this. In future work we plan to explore how the ShadowTouch interface can be used with multiple people on significantly larger displays, in particular scaling beyond two concurrent users. Also, we want to continue our analyis of application selection performance in real-world settings by fielding the interface on our university-wide deployment that has been in use since February This should allow us to better investigate how group behavior around a display changes once concurrent application use is possible. Last but not least, it would be interesting to see the idea of ShadowTouch being applied to even larger displays, such as media façades, where concurrent use is the norm, not the exception. REFERENCES 1. C. Ardito, P. Buono, M. F. Costabile, and G. Desolda. Interaction with large displays: A survey. ACM Comput. Surv., 47(3):46:1 46:38, Feb F. Brudy, D. Ledo, S. Greenberg, and A. Butz. Is anyone looking? mitigating shoulder surfing on public displays through awareness and protection. Proceedings of The International Symposium on Pervasive Displays, pages 1:1 1:6, 2014.

7 3. N. Davies, A. Friday, P. Newman, S. Rutlidge, and O. Storz. Using bluetooth device names to support interaction in smart environments. Proceedings of the 7th international conference on Mobile systems, applications, and services, pages , N. Davies, M. Langheinrich, S. Clinch, I. Elhart, A. Friday, T. Kubitza, and B. Surajbali. Personalisation and privacy in future pervasive display networks. Proceedings of the 32th SIGCHI Conference on Human Factors in Computing Systems, CHI 14, Toronto, ON, Canada, April 26 - May 1, N. Davies, M. Langheinrich, R. José, and A. Schmidt. Open display networks: Towards a new communications medium for the 21st century. EEE Computer, vol. 45, Special Issue on Large Interactive Displays. IEEE Computer Society, pages 58 64, I. Elhart, M. Langheinrich, N. Memarovic, and T. Heikkinen. Scheduling interactive and concurrently running applications in pervasive display networks. The International Symposium on Pervasive Displays, PedDis ACM, A. Friday, N. Davies, and C. Efstratiou. Reflections on long-term experiments with public displays. Computer, 45(5):34 41, T. Heikkinen, T. Linden, T. Ojala, H. Kukka, M. Jurmu, and S. Hosio. Lessons learned from the deployment and maintenance of ubi-hotspots. Forth International Conference on Multimedia and Ubiquitous Engineering, S. Hosio, M. Jurmu, H. Kukka, J. Riekki, and T. Ojala. Supporting distributed private and public user interfaces in urban environments. In Proceedings of the Eleventh Workshop on Mobile Computing Systems & Applications, HotMobile 10, pages 25 30, New York, NY, USA, ACM. 10. R. José, N. Otero, S. Izadi, and R. Harper. Instant places: Using bluetooth for situated interaction in public displays. IEEE Pervasive Computing, Vol. 7, No. 4, pages 52 57, J. Müller, R. Walter, G. Bailly, M. Nischt, and F. Alt. Looking glass: a field study on noticing interactivity of a shop window. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI 12, pages , T. Ojala, V. Kostakos, H. Kukka, T. Heikkinen, T. Linden, M. Jurmu, S. Hosio, F. Kruger, and D. Zanni. Multipurpose interactive public displays in the wild: Three years later. Computer, 45(5), 42-49, T. Ojala, H. Kukka, T. Linden, T. Heikkinen, M. Jurmu, S. Hosio, and F. Kruger. Ubi-hotspot 1.0: Large-scale long-term deployment of interactive public displays in a city center Fifth International Conference on Internet and Web Applications and Services (ICIW), pages , P. Peltonen, E. Kurvinen, A. Salovaara, G. Jacucci, T. Ilmonen, J. Evans, A. Oulasvirta, and P. Saarikko. It s mine, don t touch!: Interactions at a large multi-touch display in a city centre. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI 08, pages , New York, NY, USA, ACM. 15. J. Rui, P. Hélder, S. Bruno, M. Ana, and R. Helena. Beyond interaction: tools and practices for situated publication in display networks. Proceedings of the 2012 International Symposium on Pervasive Displays, PerDis 12, J. Sauro and J. R. Lewis. Quantifying the user experience: Practical statistics for user research. Morgan Kaufmann, G. Shoemaker, A. Tang, and K. S. Booth. Shadow reaching: A new perspective on interaction for large displays. In Proceedings of the 20th Annual ACM Symposium on User Interface Software and Technology, UIST 07, pages 53 56, New York, NY, USA, ACM. 18. O. Storz, A. Friday, and N. Davies. Supporting content scheduling on situated public displays. Computers and Graphics, 30(5): , C. Taivan and R. Jose. An application framework for open application development and distribution in pervasive display networks. Proceedings of the 2011th Confederated international conference on the move to meaningful internet systems. Berlin, Heidelberg, 21-25, C. Taivan, R. Jose, B. Silva, I. Elhart, and J. Cordoso. Design considerations for application selection and control in multi-user public displays. Journal of Universal Computer Science, D. Vogel and R. Balakrishnan. Interactive public ambient displays: transitioning from implicit to explicit, public to personal, interaction with multiple users. UIST 04: Proceedings of the 17th annual ACM symposium on User interface software and technology, ACM (2004), pages , 2004.