Pers Ubiquit Comput (2007) 11: 45–58 DOI 10.1007/s00779-006-0068-5

ORIGINAL ARTICLE

Yanna Vogiazou Æ Bas Raijmakers Æ Erik Geelhoed Josephine Reid Æ Marc Eisenstadt Design for emergence: experiments with a mixed urban playground game

Received: 13 March 2005 / Accepted: 20 September 2005 / Published online: 8 June 2006 Ó Springer-Verlag London Limited 2006

Abstract In this paper we present our work in the design integrating virtual presence with physical presence of ubiquitous social experiences, aiming to foster group through location and proximity information. We focus participation and spontaneous playful behaviours in a on the possible social implications of these technologies city environment. We outline our approach of design for such as the feeling of being part of a group, based pri- emergence: to provide just enough of a game context marily on the sense of others being present in virtual or and challenge for people to be creative, to extend and (both physical and virtual) spaces. enrich the experience of play through their interaction in Recent studies Paulos and Goodman [16] have ex- the real world. CitiTag is our mixed reality testbed, a plored ways to represent and communicate our rela- wireless location-based multiplayer game based on the tionships with people we encounter regularly in our concept of playground ‘tag’. We describe the design and everyday lives, yet hardly ever get to interact with, our implementation of CitiTag and discuss results from two encounters with ‘familiar strangers’ [10]. These studies user studies. have illustrated that our perception of who else is around us is a significant factor for feelings of comfort Keywords Social computing Æ Wireless location based and reassurance in public spaces. Rheingold’s discussion games Æ Mixed reality Æ Emergence Æ Group behaviour of Smart Mobs [20] highlights the overwhelming power of social cohesiveness that can be brought about by knowledge of the presence and location of others in both 1 Introduction real and virtual spaces. We know that messaging tech- nologies (IM, SMS) have been used extensively for The uptake of mobile technologies has undoubtedly purposeless communication, just for the sense of being been changing our communication practices and affect- in touch with other people. For example, the exchange ing our everyday life patterns. One of our key interests is of ‘goodnight’ text messages creates connectedness [21]. presence, as knowing or being aware of other people’s Similarly, Nardi and Whittaker [13] report that people in existence, for example whether one’s friends and col- their study of IM use in the workplace found value in leagues are online or somewhere in a nearby location. simply knowing who else was ‘around’ as they checked With the advent of mobile technologies, presence has their buddy list, without necessarily wanting to interact become a rich concept as well as more ubiquitous, with buddies. These examples show that a sense of presence can be communicated in symbolic ways, via the Y. Vogiazou Æ M. Eisenstadt display of meaningful state information (e.g. availabil- Knowledge Media Institute, The Open University, ity, activity, location, identity etc). MK7 6AA Milton Keynes, UK New opportunities have also emerged for individuals B. Raijmakers (&) and groups to communicate and coordinate their Interaction Design Department, Royal College of Art, activities spontaneously in urban environments. We Kensington Gore, London SW7 2EU, UK know that wirelessly inter-networked groups of humans E-mail: [email protected] can exhibit emergent prediction capabilities [20] and thus demonstrate self-organizing dynamics. In this E. Geelhoed Æ J. Reid Hewlett-Packard Laboratories, Filton Road, context, we have been exploring how spontaneous Bristol BS34 8QZ, UK individual and group behaviours can emerge in the real world through playful collective experiences. Previous Y. Vogiazou online studies [27] have shown that spontaneous Design Department, Goldsmiths College, University of London, London SE14 6NW, UK social behaviours can ‘emerge’ among groups present in 46 multi-user environments, even without explicit and ver- them in the actual design, but encouraging team play at bal communication. The recent Flash Mobs phenome- the same time, we expect to see spontaneous individual non [8] illustrated that people do not hesitate to perform and collective behaviours to emerge. certain acts in public together with many others, which Why design for emergence? Johnson [7] in his book otherwise would have been quite embarrassing. In fact, Emergence has described how artificial intelligence sci- people participating in those events seemed very engaged entists and game designers like Will Right who designed and amused. These acts of spontaneous play have been SimCity have understood and recreated emergent sys- thought-provoking within the context of our research. tems based on simple rules that have a complexity and Play has been inherently social, before the advent of life of their own. In fact, emergence is a particularly communication technologies, as we see everyday on intriguing buzzword for game designers [3] as they try to school playgrounds. Presence enabled technologies cre- incorporate emergent properties in their games to en- ate new prospects for play, for adults as well. In our hance the user experience. Emergent interaction based research these are the boundaries we explore: what kind on simple, high level rules can vary result in a different of engaging social experiences can emerge in the real experience every time, a game that is interesting to play world based on the awareness of individuals participat- more than once. ing in a parallel virtual experience? Does virtual presence Emergence is not only valued in game design, but penetrate physical presence in any way? other design disciplines too, such as computer-aided design. A property of a design that is not represented explicitly at the time of creation is said to be an emergent 2 Design principles property if it can be made explicit [4, 22, 11]. Protocol studies of designers while sketching have shown that We believe in the concept of ‘complexity growing out of unexpected discoveries of emergent shapes can have a simplicity’, as being crucial for the success of mixed significant impact on the course of further design activity reality experiences where the main focus should be on [23, 26]. While the above studies focus on emergence the surrounding environment and people rather than on during the design process itself, we are specifically the technology itself. This concept is summarized in our interested in emergence occurring from the unintended design principles for creating engaging social experiences uses of technology. We have many examples of unin- through the use of mobile technologies below: tended uses of technology, such as, SMS, hijacking Bluetooth phones, radio broadcasts from iPods. Often these unintended uses of technology repurpose the de- 2.1 Lightweight design: keep it as simple as possible sign towards a new direction, incorporating ludic values. Unintended use as a design ‘approach’, can also be seen The goal is to enhance our interaction with the social as a way to open up for the public to take part in the and physical world around us by adding an additional shaping of a public sphere, the cyberspace [24]. While experiential layer. So the game we describe below is there is an economic interest in designing for the unex- designed as a prompt for the real world, real people pected (e.g. SMS revenue, killer apps), it is also benefi- being physically close. cial for the design process as well, as users can become part of the process by pointing out new design directions and providing inspiration for designers. 2.2 Scalability Following from the aforementioned design principles, our research focuses on how play, based on very simple We know that certain collective behaviours, such as for game rules, can lead to more complex interaction in the example the Mexican Wave in the real world [2], can physical world, emergent cooperation and collective only emerge once a critical mass of participants has been behaviour. Experimental studies can provide useful les- achieved. By designing scalable applications, aiming to sons for the design process of future ubiquitous com- involve as many users as possible, we want to explore puting social applications and we describe our testbed what kind of spontaneous interactions could emerge at game next. large scale.

3 The game design 2.3 Design for emergence Our project has been inspired by playground games, This principle summarizes our overall approach. By particularly the simplicity, spontaneity and instant fun employing the right affordances [14], such as using of ‘playground tag’ [15]. In this well-known children’s simple metaphors, related to people’s past experiences game, play emerges at often unpredictable moments, that can be re-interpreted or subverted, we acknowledge when, for example, a group of children walk along the creative aspects of playing, game tactics and bending the street together and one touches another saying the rules. By allowing people to explore strategies and ‘you’re it’ and they then start chasing each other. Tag or cooperation practices themselves, without incorporating tig illustrates, like other playground games, that highly 47 engaging playful interaction is possible even without the specific principles and goals of structured games. As Opie and Opie [15] mention in their book on playground games, play is unrestricted, while games have rules; in the playground there is no need for an umpire, little significance is attached to who wins or loses and it doesn’t even seem to matter if a game is not finished. Mobile technologies enable short, spontaneous and intermitted interactions while on the move [18] and it is precisely these playful interactions we aim for. More- over, we know that mixed reality, location-based expe- riences are characterized by a transition between immersive and non-immersive states [19] caused by either planned events or occurring interruptions such as a happening in the physical world or a system fault. The design challenge here is, unlike fully immersive computer based games, to smoothen these transitions between immersion and non-immersion. Our game is designed to be played on an ad hoc basis, allowing for awareness of the surrounding environment and external to the game stimuli. Playful, temporal social interaction in the real world while participating in the game can become part of the overall experience. Fig. 1 Tag alert We have further developed the ‘tag’ concept to encourage emergent social behaviours in an urban con- The game has been designed with the principle of text. In our design we tried to specify the minimum rules scalability in mind; we foresee it evolving to an every- and structure possible for people to understand the game day experience with the kind of ‘smartphones’ are and its metaphor and for cooperation to emerge. City already appearing. The game has a lightweight design, space is used as a playground and passers-by can motivated by the hypothesis that very simple game become the usual or unusual suspects in a novel expe- rules based on presence states (e.g. I am Green and rience. We named our game CitiTag in reminiscence of ‘tagged’) can result in an enjoyable social experience, the original game. stimulated by real world interaction among players. Therefore, the game application communicates prox- imity rather than location, as the players’ true location 3.1 Design is not revealed and they receive only the name or

CitiTag is a multiplayer, wireless location-based game, played on handheld, iPaq Pocket PCs connected to global positioning system (GPS) and a wireless network. As a player of CitiTag, you belong to either of two teams (Reds or Greens) and you roam the city, trying to find players from the opposite team to ‘tag’. When you get close to someone from the opposite team, you get the opportunity to ‘tag’ them: an alert appears on the screen with a sound (Fig. 1). You tap on the screen with your thumb to ‘tag’ the other person. You can also get ‘tagged’ if someone from the opposite team gets close to you and ‘tags’ you first (Fig. 2). If this happens, you need to try and find a team member in vicinity to set you free, to ‘untag’ you. Sim- ilarly to tagging, players receive an alert whenever a ‘tagged’ team member is nearby that they can rescue (Fig. 3), provided there is no opponent nearby to tag them at the same time. There is a ‘group state’ feature, so players can see how many people are free and how many have been tagged in each team at any time (see top of screen in Figs. 1, 2, 3). They also see how many people they tagged and rescued individually at the end of the game. Fig. 2 The player got tagged 48

3.2 Architecture

The diagram below shows the system architecture for CitiTag [17], which is part of the Mobile Bristol project, by HP Labs Bristol [6]. The mobile bristol application Hull [5] is used to read the location data from the GPS receiver and to pass it via an XML socket, to the Flash Game Client, the program that is used to actually play the game on a Hewlett-Packard, iPaq PocketPC. This program is written in Macromedia Flash and it also connects, wirelessly, to the CitiTag application on the Flash Communication Server. The server incorporates the game logic and sends all connected game clients updates about their state (tagged or free) and nearby devices they can interact with (tag or untag). There is a browser-based Administration client also written in Macromedia Flash which is used to administer and monitor the game (Fig. 4).

4 Experiment design and user studies

Fig. 3 A tagged team member nearby We used a combination of qualitative and quantitative research methods in our empirical studies. Two user trials were carried out: a pilot trial with 9 participants in nickname of a person being really close, whom they an open field space at the Open University campus in can ‘tag’ or rescue. This also helps to discourage people Milton Keynes and a trial with 16 participants in a from looking at the screen too much to pinpoint oth- square in the city centre of Bristol. Participants played ers’ exact location. Instead it serves as the alert aiming the game outdoors for approximately 40 min with to draw attention to the surrounding environment and intervals for restart, debugging or quick briefing when give a clue for players to try to identify the person necessary. There was a briefing session of 15 min before appearing on their screen. The familiar metaphor of the start of each trial when participants received the kit tagging originating from the famous children’s game and instructions on how to use it and how to play the and the idea of the two teams are very simple too, to game. allow people to extend and bring new meanings to the Trying the game in two very different locations al- game as they discover their own cooperation strategies. lowed us to compare the emerging interactions in the Our approach to design for emergence in this game can context of the particular environment where the trial therefore be summarized as follows: simple game took place. We took video of both trials, aiming to interaction rules, providing for interaction with the capture the experience of game play, as opposed to physical world, simple metaphors that relate to players’ simply registering what was happening. Therefore the past experiences.

Fig. 4 CitiTag architecture GPS Receiver Flash Communications Server running Serial data CitiTag application (server-side scripts) PocketPC Wireless Network Mobile Access Bristol Point Application

802.11b wireless XML Socket

Flash Administration Flash Game Client Client 49 cameraman walked or even ran with the players. This GPS positioning at the same time on each device was a allowed us to observe what people did, how they reacted great technical challenge which we dealt with in the best to events in the game, see their face expressions and hear way possible. Planning these experiments in the form of what they were saying to others. Also the close presence events was necessary to achieve a critical mass of players of the cameraman often elicited them to comment on the and test the concept at an appropriate city location. game. On such occasions they freely ventilated their thoughts, very much like protocol analysis where users are asked to ‘‘think aloud’’ while solving problems or 5 Findings performing a task [1]. The material was edited down to 3 min films for each of the two trials. These films al- Our user trials provided results on the fronts we have lowed us to convey the game experience and show some been looking at: emergence, game experience, team crucial observations to others. ‘belongingness’ and design. We describe our findings The trials were followed by group interviews, in below. addition to which all participants completed a ques- tionnaire about their experience with CitiTag. The group interviews lasted 30–45 min and were open discussions 5.1 Emergence loosely structured around the main research themes: experience of gameplay, the game as part of everyday Observation during the trials and video analysis revealed life, group cooperation and strategies, awareness of a range of emergent behaviours, both individual and others and interaction with the device. These discussions collective, which show how our participants explored the were also recorded on video for further analysis. The game, technology and the social context by pushing the questionnaire was designed to collect more detailed data limits of what was available to them. In the Bristol trial, on the research themes as well as the usability of the this process became clear from the log of the game: device and game interface. The questionnaire consisted initially one team quickly won the other, but later, as of graphic rating scale questions [25] interspersed with strategies evolved, the teams learned to fight back, so the open questions. Participants were asked to put a mark game lasted longer and the numbers of tagged people on a 100 mm line between two extremes, the extreme left kept going up and down with often surprising last being 0 mm (‘‘Not at all’’ in many questions, scored as minute victories. Similarly, in the trial at the Open 0) and the extreme right 100 mm (‘‘Very Much’’, scored University, once we introduced nicknames and the game as 100). The statistical results reported in the Findings became more ambiguous, participants spent more time section below concern correlations between two vari- trying to find out who was part of which team, so the ables, denoted by the letter ‘‘r’’, accompanied by the game lasted longer and also became more interesting for relevant degrees of freedom (‘‘df’’) and the likelihood them. (‘‘P’’), indicating how much the results could have come Our observation is that emergence occurred when about by chance, with ‘‘P’’ being significant when people stretched the boundaries of both the virtual and P < 0.05. physical world. Here are some illuminating examples: Participants were recruited via email sent to student and departmental mailing lists. The invitation was open to anyone and although we aimed for a student popu- 5.1.1 Stretching the limits of the technology lation, the majority of our volunteers (10 out of 16) in Bristol were HP Labs employees, eight of which decided In both trials we observed participants really exploring to try the game as an exercise for team building. This the technology and finding what works and what brought an interesting dimension to our research and we doesn’t, so in the Open University for example, we had tried two different variations of the game: one where two people who tried to estimate the range for the tag- team members were not known and part of the challenge ging event to come up by getting close to each other was to find out who was on which team and one where slowly and moving further away, watching when the tag the group was divided into two visible teams, with the event would appear. Often people took the role of the team practicing team building starting together on one ‘amateur scientist’ in both trials, looking at each other side of the square. All participants used nicknames in the screens, trying to solve occurring technical problems game. In our first trial in Milton Keynes we tried two with GPS and wi-fi and discussing how the technology different variations of the game: one with participants’ worked. One participant in the Open University started real names and another one with nicknames. running around trying to pick up any signal. Similar Organizing both of our experiments involved a great explorative behaviours have been observed in ubiquitous effort and detail as everything needed to be coordinated location-based experiences with soundscapes [19], where to perfection for 16 people to be able to play together users were typically engaged with learning how the sys- simultaneously outdoors. For instance, in Bristol, a tem works and experimenting a lot to discover the support ‘crew’ of ten people was necessary to prepare boundaries between areas with different sounds. and run the trial. Our trials were not without problems, What was interesting in CitiTag is how participants in particular, having a stable wi-fi signal and accurate turned the technical difficulties to their advantage, so for 50 example, when they would see that they had a good A couple of participants in the Open University trial signal and everything was working well, they would try also referred to other playground games in the ques- to get close to as many people as possible. At least a tionnaire: one person said it was a kind of cyber cops couple of people in the Bristol trial tried to take and thieves game and another that it reminded her hide advantage of server communication lags by teaming up and seek, only that everybody hides and seeks same in a pair: in this way they were in a more advantaged time. position than a lonely opponent: even if he or she tagged While physical behaviours associated with the classic one of them, the other usually still had enough time to tag game are not part of CitiTag’s design, we observed tag the opponent and then rescue the tagged team that people could extend the metaphor of ‘tag’ by member. This tactic often resulted in clusters of team introducing this physicality in the game. Our conclusion players moving along together in a group. One partici- is therefore that the use of metaphor is helpful in com- pant in Bristol commented: municating a style of game ‘It’s like a game of tag’, but then the actual tactics, behaviours and interaction ‘‘it seems to work! We got about five people in one techniques of the new game are defined through exper- session and kept ourselves released and then kept imentation and trial. walking and we got more people, so it did seem to work’’. 5.1.3 Bending the game rules This example shows that social protocols around group formation, behaviours and movement will occur natu- Most games can have different levels of rules: rules that rally when people are co-located with a shared activity, are enforced by the mechanics of the game, locally through the exploration of simply designed technology. established rules and social and legal rules. For example It is not necessary or desirable to introduce sophisticated if we consider the game of football there is a well technology solutions that interfere with natural social established set of game rules. Local rules that interpret bonding behaviours or emergence of game tactics. For or bend those games rules can be agreed by the players example mechanisms in the system for mapping the and are enforced by a referee or the players themselves. location of team members or forming groups would Social rules such as not allowing ball games in certain most likely be over complex and distract people from areas restrict the venues and the areas that one can play. looking around them and using facial gestures and body A parallel set of rules can be established for mixed language to identify team members. reality games. In CitiTag game rules, enforced by the system, will not allow a person to tag others once he or she has been tagged. Other rules can emerge locally, 5.1.2 Extending the metaphor depending on the context of the game. For example, in our trial at the Open University participants often asked The simple tag metaphor helped people to relate to the explicitly what team others were, making the game too game and use their imagination to extend its meaning obvious and easy, so we introduced a ‘no speaking’ rule. through their interaction. For example, in the Open Social and legal rules exist in the environment, so it University trial, one participant kept his hand up (Fig. 5) would not be socially acceptable to grab non partici- to indicate he was tagged in order to attract attention so pants and hide behind them or to damage or disturb that someone would come and untag him. He said in the features in the environment. interview that he remembered that his classmates did this We observed several cases in which participants in ‘tag’ games in childhood. The same participant also subverted the rules of play in creative ways. One par- went and sat in a corner waiting for people to come by ticipant in Bristol broke the team division game rule by and untag him and indeed, other participants ap- introducing a ‘secret assassin’ behaviour that no one proached him. In Bristol, waving or raising hand when expected (Fig. 6). He logged on to the opposing team tagged to attract attention, was a typical behaviour, and when the game started he switched back to his performed by at least five different people. original team whilst among opponents and tried to tag

Fig. 5 Tagged players raise their hand to indicate they want team members to rescue them at the OU (left) and Bristol (right) trial, respectively 51

Fig. 6 A ‘secret assassin’ tagged puzzled members of the Red team Fig. 7 A passer-by in Bristol has stopped to ask a trial participant in Bristol when they did not expect it about what she is doing and started a conversation as many people as possible before getting tagged. This We observed the effect our user trial had in the social was very surprising because teams had been divided at environment of the square in Bristol. People were curi- that point and as the whole Green team was at the other ous about what was happening as they were watching side of the square were the trial took place, players of the many people with small back packs running or walking Red team could not understand what ‘hit’ them right at all around and some approached players to ask what the beginning of the game. So by breaking the rules, this was happening, starting a conversation (Fig. 7). Other, participant also created an entirely new role in the game. ‘non-participants looked quizzically on’, as one player At the Open University, participants very quickly reported. Interestingly, a couple of girls passing by, subverted the ‘no speaking’ rule by using gestures and when they saw one of the players holding his hand up (to body language to communicate what was happening, for indicate that he was tagged), waived back and greeted example smiling at someone in a cunning way, trying to the player in amusement. understand their team membership, to guess by It became clear with this trial that using the device exchanging facial expressions. One participant noted in and participating in the game automatically authorized the interview an inherent contradiction in the game from people to be a child, but in a socially acceptable manner. the point we introduced the ‘no talking’ rule, between (a) In fact, people used the space as their own playground, wanting to be secretive to obey the rules and (b) wanting often running about, laughing and shouting over dis- to be more sociable to have some fun. We believe that tance. It is important to note that our participants did local rules in mixed reality games will be negotiated and not feel embarrassed to play this game in public, as the defined depending on the situation and the context of average score was 23.25 on a 100 mm scale of 0 (not at play (e.g. how exposed is the location, familiarity with all embarrassed) to 100 (very embarrassed), with 12 other participants). Our observations revealed that people out of 16 ranging between 0 and 40. In fact, one bending the established rules can result in a creative participant commented that playing it in public was activity (e.g. a new role in the game, new forms of actually a great part of it. interaction), therefore the design of such games should Although our trial attracted attention, it did not have allow space for exploration and play with the rules a disruptive impact on the social environment of the themselves. square because players respected the social rules we mentioned above, like not getting in the way of other people’s activities. Skate boarders were also present in 5.1.4 Exploring the physical environment the same location, so the square was a public space allowing for play. Nowdays many public spaces try to Our players in Bristol tried to use the environment to ban the use of mobile phones because they are socially their advantage by hiding behind obstacles when trying intrusive. The design of CitiTag aimed to facilitate to approach another person. A few people also tried to peripheral awareness and we did not observe any cases stay behind a bush for some time. However, hiding is in which engagement with the game and lack of aware- not only physical as there is another form of hiding ness weakened the players’ social responsibility, which is possible in CitiTag; one participant mentioned that if a common risk for mixed reality experiences. This is you go under the bus stop you would lose GPS so you consistent with the points we made earlier about the could not be tracked any more, what we have identified rules—unlike other bendable rules we observed, social as hiding in the virtual world, i.e. still visible by others, rules are clearly less likely to be disrupted, certainly by but not virtually ‘there’. our participants. 52

Fig. 8 Several players ran as either a defensive or an offensive tactic, and their behaviour sometimes provoked others to run after them. Photos from the OU (left) and Bristol (right) trials

5.1.5 Stretching personal limits ‘untaggers’ away from the ‘battlefield’, so that tagged players would know to come back to them. We also In both trials, people pushed the limits of themselves by observed an ‘invincible pair’: two players who sponta- trying to work out strategies both individually and as a neously teamed up and kept rescuing each other. What is team. Frequently observed behaviours were running, special about this case is that they did not agree on the often contagious, introduced by one or two people and strategy explicitly, it emerged spontaneously. Here is followed by others starting running too (Fig. 8). how one of them described it: Participants also tried to surround a person together ‘‘The pairing thing happened just like that. We just (Fig. 9). One participant at the Open University wandered off together and that was it, we didn’t really pretended to be in the game although he had lost GPS, talk about it. It seemed that it was working and we were so he kept walking in the middle of the filed, in a chal- constantly releasing each other—we didn’t need to say lenging way often approaching people suddenly to scare ‘let’s do this’’. them. Another player focused a lot on identifying who was from which team and developed his own ‘spying’ strategy: We have come to the conclusion that our broad principle of design for emergence can be actually translated as ‘‘Whenever I was able to determine if someone was an design for pushing boundaries, based on our observation opponent, I would always try to track their movements that when people push boundaries on all fronts, as through the game and try to catch them off-guard’’. illustrated in this section, then emergent behaviours occur. In Bristol our participants cooperated a lot with each other. For instance, at least eight people reported in the questionnaire that they tried playing the game in pairs in 5.2 Team ‘belongingness’ and cooperation order to rescue each other when tagged. In this respect the game play was very different to that of the OU trial, We asked our participants to rate how important it was where the exposed and limited space did not encourage for them to know how many players from their team strategies to evolve as much. Another strategy players were free and how many had been tagged, in order to tried as a team in Bristol was to keep one or two official investigate the team aspect of the game. In the OU trial most participants found it very important (mean score 80.3), with 7 out of 9 ratings ranging between 70 and 100 on a 100 mm scale ranging from 0 (not at all important) to 100 (very important). Similarly in Bristol, most par- ticipants found it also very important (mean score 80.68), with 10 out of 16 replies ranging between 80 and 100. This result shows that CitiTag is clearly a team game, even if players start playing without knowing who their team-mates are. We also asked them to rate how strongly they felt part of their group (Green or Red). In Bristol eight out of nine responses to the question varied between 40 and 85 with a mean score of 57.5. This shows that many participants gave a quite neutral response, some felt part of the Greens or Reds a lot while others less. In Bristol, we had a similar result again with most ratings ranging between 40 and 100 with a mean score of 62.56. So al- Fig. 9 Two players at the OU trial try to surround an opponent to tag him. In this case they did not succeed; the opponent tagged one though our participants thought that it was very of them first important to know the group state (how many were free 53 and how many tagged), they did not associate as strongly part of their group also found this group state strongly with the team identity. When asked whether awareness information very useful. We can conclude they would defect to the opposite team, 10 out of 20 that minimal ‘group state’ awareness information, along respondents in total in both trials would stay loyal to the lines of ‘there are 3 free Greens out there’, appears to their team, while the other half would join the opposite be beneficial for team belongingness and sufficient to team to explore game play possibilities or to win. This evoke the sense of group participation in a shared mixed might also be the result of the team identity not being reality experience. consistent: players changed teams in the different game A particularly interesting significant correlation was sessions during the trial. found in Bristol between awareness of team members For the Bristol trial we found a significant correlation and the importance of knowing the group state (Fig. 10) between feeling part of the group (Reds or (r = 0.588, df = 14, P < 0.025). So, people who were Greens) and the usefulness of the figures on the top of more aware of their team members also said it was the device’s screen, used in the game to convey the group important for them to know how many players were free state, i.e. how many players of each group were tagged and how many had been tagged, the group state, as we and how many were still free (r = 0.794, df = 15, can see in Fig. 11. P < 0.001). This indicates that people who felt more Our participants in both trials commented during the interviews that they needed variable levels of proximity information, displaying both players they could perform actions on (e.g. tag, untag) as well as players in vicinity, but not yet in their immediate ‘radar’ field. There is an indication that those players who were more aware of their team members, were also ‘team players’, they cooperated with others and tried to work out strategies. So for example, three players who were highly aware of their team members (96, 87 and 85, respectively), they all reported in the questionnaire that they tried out cooperation strategies in pairs, going along together to rescue each other. About four-five participants in both trials mentioned in the interview explicitly that they were trying to free their team mates. For example, a female participant in the OU trial would check the figures and if her team was losing she would try to look around for people to free. Another female participant in Bristol had the same ‘team player’ attitude: ‘‘I’d look at the numbers, and if the numbers were getting high, I’d go and find them and untag Fig. 10 Correlation between ‘group belongingness’ and ‘usefulness them’’. Some people could even get a ‘savior’ status, of group state figures’ in the Bristol trial for example a participant in Bristol commented: ‘‘There was a guy, I think from HP, he was really good cause he actively searched us out and he would come and release us all and then we would run back in’’. So players who concentrated on the team aspect of the game also wanted to have up to date information about the state of each team. We also found in the Bristol trial that team awareness is significantly correlated to amusement (r = 0.589, df = 14, P < 0.05), awareness of other people (r = 0.557, df = 14, P < 0.025) and the importance assigned to wining (r = 0.659, df = 14, P < 0.01). The fact that, individuals who were keen on winning were also good team players and tried to collaborate, is illu- minating, as this is not necessarily the case in games in general, where hard competition goes along with indi- vidual pursuit of winning. These results allow us to draw conclusions that ‘team aware’ individuals are good Cititag players, that CitiTag is a team game and that ‘group state’ awareness infor- mation is important for ‘team belongingness’ and for Fig. 11 Correlation between ‘team awareness’ and the importance of knowing the group state, in the Bristol trial cooperation to emerge. 54

5.3 Issues affecting the user experience concluded that an urban environment is more appro- priate for this type of game. Unlike football, a game that We have identified several factors that enhance or grew up in the street with a strong physical element, but hamper the experience of playing CitiTag and we discuss restricted to specialized venues or playground locations, them below: a mixed reality game can be socially acceptable and blend in the environment. The design of such games needs to maximize the need to be aware of the envi- 5.3.1 Location ronment and the people around, rather than assume their presence as irrelevant to the game experience. Our user trials revealed that the experience of playing CitiTag can vary depending on the location where the game is being played. We noticed significant difference 5.3.2 Group dynamics and the social aspect between the quick, action game sessions in the OU trial and the more strategic ones, with evolving cooperation People also make the experience different every time, and group convergence in the city environment in the individual behaviours emerge and other players respond Bristol trial. Players were able to move peripherally, hide accordingly, group dynamics constantly defining game- and work out strategies with others in the city, actions play. A great part of the overall experience is the social that were much more difficult to perform at the Open aspect and strategy in the game. One participant at the University trial, because the field area was small and OU commented that ‘it is nice to belong to a team’ and completely exposed. that ‘it is even more interesting when you don’t know who In the OU trial participants did not find the location is in your team and who is an opponent’. Another par- particularly suitable for this game (mean score was 54.2). ticipant said that ‘it is different every time you play it!’.In Eight out of nine participants agreed during the inter- the OU trial participants said they preferred playing view that the location was not particularly suitable and with nicknames rather than with their real names. They explained that they felt too exposed and they needed explained that because they all knew each other game- more space, obstacles, players and other people around. play was too easy and finished quickly, whereas it They wished to explore further the possibilities of play, became more intriguing with nicknames, when they tried collaboration and strategies in the city. In fact, we figuring out who was who. In Bristol, participants also identified a significant negative correlation between liked trying to find out who was part of which team as ‘enjoyment’ and ‘suitability of location’ (r = À0.831, well as playing with a known team, with some people df =8,P < .01), which shows that the more our par- having a preference for one over the other. This suggests ticipants enjoyed the game, the more they thought that that the game has two different yet complementary the location was not suitable for it. dimensions: an initial phase of exploration where the In Bristol, however, we found the reverse result: a player is alone, trying to find team players and tag significant correlation between ‘enjoyment’ and ‘suit- opponents, much like a spy, and then a second phase ability of location’, so the more players enjoyed the when the team is known and people cooperate against game, the more they thought the location was suitable the opposing team. When analysing both experiments, (r = 0.555, df = 15, P < 0.025). Questionnaire com- we found a significant correlation between enjoyment ments indicate that eight people were positive about the and feeling social across both experiments (r = 0.542, location (they rated its suitability above 60) and thought df = 24, P < 0.01). This shows that in our game, the it was good to have other people and objects around. social aspect is particularly strong in the overall expe- Two people felt too exposed and wanted more places to rience. hide and three people found the location too busy and Our participants suggested some ways they would noisy. One participant liked the fact that he had to study like to see CitiTag evolve into a more strategic game, the environment. During the interview people were po- leveraging social skills, for example though matching sitive when discussing the location and even suggested profiles and dynamic team formation based on the other city areas which would be more convenient, such profile. They wanted more decision making and trade- as pedestrianised areas with no traffic and less noise and offs, in particular, to be able to choose themselves areas ‘‘with lots of concrete blocks you can hide’’. between tagging an opponent or rescuing a team mate, Playing CitiTag in a public space, rather than in a as opposed to this being done automatically by the game-specific location (e.g. like paintball) is important system with the tagging event having priority over an and a big part of the game’s identity, as one participant untagging event. Some design ideas they mentioned in Bristol mentioned: ‘‘I was trying to compare it with were: universal ‘untagger’, shields, ‘roving medics’, re- other ‘tag’ types of games, if you think of laserquest, wards, time limit for being tagged etc. Such features where you go into n environment of dark lights and people could keep both the novice and experiences players en- wearing suits, walking around with guns, but here it was a gaged and possibly encourage people to undertake dif- form of tag game which actually you could play and it ferent roles. The element of getting to know your wasn’t awkward to play in public’’. We therefore have opponents or friends was flagged as particularly inter- 55 esting to explore through this kind of game, which can be combined nicely with socializing and meeting people.

5.3.3 The real world as an interface

Our results suggest that engagement in CitiTag is asso- ciated with the physical and the social aspects of the real world. In particular, in the OU trial we found a signif- icant correlation between ‘enjoyment’ and ‘usefulness of sound alerts’ (r = 0.979, df =8,P = <0.001), so the more a player enjoyed the game, the more useful they found the sound alerts. The use of audio was very much appreciated in both trials. Audio cues were good awareness mechanisms as they often liberated people from looking at the screen too much, which took away attention from the actual game interface—the real world. In Bristol, one participant suggested an example of the kind of interaction we should be aiming for through the use of audio: the scary sound in the film Aliens. The participant described this sound as ‘a warning that something was coming close: it was Fig. 12 Correlation between ‘awareness of other players’ and becoming louder and more frequent as the ‘alien’ was ‘enjoyment’ across both trials approaching, so as you could not see anything, it was very scary just with the sound’. This example highlights the potential of sound as an awareness feature, 5.3.4 The mixed reality challenge: a mismatch between enhancing the player experience. overlayed and what users expect to In the Bristol trial we also asked our participants to see in the real world rate their awareness of other people in the square, not participating in the game and we found that their aver- Bristol participants were frustrated by the fact that they age awareness of non participants was good (59.8) with 8 would see people really close to them and expect game out of 16 participants rating it between 60 and 100. This events to come up on screen (i.e. ‘tag’ ‘untag’ the other indicates that despite occurring distractions with tech- player), but there would be nothing new displayed or the nical problems, the design facilitated awareness of the events would come up with a delay. So the game did not social environment, which is necessary in order to correspond to the immediate environment as promptly maintain the game play in public on socially accepted as they would have expected. This was due to GPS er- levels. rors and wi-fi loss and we believe that it is a typical and We found a significant correlation in the Bristol trial significant problem for mixed reality experiences. Once between awareness of other players and enjoyment we have provided a link between an overlayed reality (r = 0.694, df = 14, P < 0.01). This result indicates and the real world, people expect to see the connection that participants who were more aware of other players, between the two. If what they see with their own eyes is enjoyed the game more than others who focused on the not reflected in their device with a relevant timely alert, screen, either trying to solve technical problems or fig- their expectation is not satisfied and this decreases uring out what was happening, rather than observing the enjoyment and hampers the game experience. During people around them. For example, three players who did the trial we also observed that this same factor caused not look as much at the screen as others (their ratings on several strategies and attempts for collaboration to fail, the 100 mm scale were 68, 46 and 66) were very aware of resulting in even more frustration. For example, when other players (85, 99 and 91, respectively). They also two people stayed back to serve the team as ‘official enjoyed the game (75, 74 and 60, respectively) and untaggers’ and later a tagged team player approached mainly commented on the technical difficulties with GPS them, neither of them got the opportunity to free him, and wi-fi as negative aspects in their experience. because he had lost GPS coverage. Therefore his device Similarly, a significant correlation between awareness was not detectable despite the fact that he was standing of other players and enjoyment was also found across right next to them and this caused disappointment to the both the OU and Bristol experiments (r = 0.640, tagged player. This mismatch, what we called the mixed df = 23, P < 0.01) (Fig. 12), so the game and device reality challenge, also made it more difficult to identify should promote and encourage this kind of interaction. people in the physical world when their nicknames were These results confirm our lightweight design principle as displayed on the device. In the questionnaire most par- the right approach for this type of mixed reality (phys- ticipants reported that it was very difficult (mean 26.2 ical and virtual) social game. with 11 answers ranging between 0 and 30) to identify a 56 particular person by their nickname displayed with the The lightweight design approach, based on simple tag/untag alert. rules and presence states, is crucial, ensuring that the This experience factor made us consider other design device and game interface stimulate real world interac- directions, such as the need to rely less on accuracy and tion without adding overheads. The fundamental pre- to employ a level of abstraction, whenever possible. Our mise is to enhance interaction with others in the real conclusion is that if the technology can not provide an world by adding another layer of reality, so interaction efficient and accurate enough match between the virtual in the form of alerts with sound, as opposed to a rich, and physical world, abstracting the sense of proximity fully immersive computer game experience has proved to and location would be helpful and then user expecta- be effective for this kind of physical, mixed reality game. tions would be less binding. In such cases, the design Sound cues for example, were very much appreciated should be less explicit and allow different possible and often liberated people from looking at the screen too interpretations, rather than try to bridge both worlds. much. We identified that presence awareness is impor- tant, both at the level of the immediate environment (e.g. who is around or near me) but also for the overall state 5.3.5 Persistence for everyday play of the game (e.g. how many of my team are free). Simplicity is also important to ensure scalability. The While our experiments provided a sense of what it was game interactions can scale up and there is no indication like to play the game a few times, we would need to run from our preliminary studies why it could not be played longer term trials in order to find out how this experi- across a whole city if it were widely available. ence could blend with everyday life. Our participants We also identified factors that influence the experi- expressed a preference for introducing features into ence of playing CitiTag: location, group dynamics, using CitiTag that would enable them to re-join ongoing the real world as a game interface and the correspon- games on multiple separate occasions, in other words to dence between the game reality and what actually hap- add persistence capabilities. There are in fact two ways pens in the real world. These observations introduce they could envisage this game being played: interesting challenges and trade-offs and open up (a) Spontaneous, as ‘turn up in the park and play it’ or opportunities for further work in the design of mixed (b) You ‘build a number of tags’ during the day with a reality experiences. reward system. Finally, our studies brought to light an additional design implication for future mixed reality applications: Most people in both trials thought it would be more to try and introduce, whenever possible, a level of engaging as a persistent, real time experience in the form abstraction (e.g. less fine grained presence states) when of an event, which for example, goes ‘live’ for some time matching a virtual concept (e.g. a presence state like at certain locations during the day, maintaining spon- location or proximity) with the real world user experi- taneity and group interaction. Persistent scores and re- ence. In this way, if, for some reason, the superimposed wards were also mentioned as a desired feature. reality does not correspond with accuracy to what the Our results here suggest that playing CitiTag is an user really sees or experiences, the breakdown in engaging social experience on the ‘real world interface’, expectations will not be so palpable and will not hamper which can vary as it is stirred by group dynamics and the overall experience. interaction among players and it depends on factors like Considering a future version of CitiTag as part of our location and the match between virtual and physical everyday life, we can imagine it being played on an ad reality. hoc basis, starting when, for example, a critical mass of registered players shares the same location in a public space. Then, they could all receive an invitation to 6 Discussion and future work participate and gameplay would emerge, much like the original children’s playground tag. One of the most Overall, our two user trials proved the basic concept of interesting innovations of CitiTag is that it challenges CitiTag and showed that group behaviours and playful, the distinction between time dedicated to play and rich social interaction can emerge from a simple game casual daily activities. In order to test CitiTag as part of based on presence ‘states’. everyday life we would need a much large user base, so Revisiting our design principles in sect. 2, we have that at different moments there is a sufficient number of come to the conclusion that it is indeed possible to design players to enable the development of gameplay. for emergence, as indicated by the range of emergent Imagine CitiTag on a large scale, played across a individual and group actions in our user studies. All whole city and it immediately becomes a new concept, these examples were the result of people pushing the which we name ‘UrbanSwarm’. UrbanSwarm would boundaries of what was available to them: the game, the aim to further leverage social skills and collaborative technology, the physical environment, the metaphor and play, focusing on the group participation factor. Here is the participating players. In this context, design for a scenario of how we envisage interaction through emergence can be defined as design for stretching limits. UrbanSwarm to emerge: 57

‘‘You are a commuter walking through Euston station locations? We would like this concept to inspire further in London and your train has been delayed. You sud- interdisciplinary research in the design of ubiquitous denly receive a challenge on your mobile phone computing games, crossing the domains of computer informing you that there are 40 Green UrbanSwarmers science, design and social science. and 25 Red UrbanSwarmers distributed nearby. You have already signed up as a Red player, so you decide to Acknowledgments CitiTag has been developed jointly by The Open take up the challenge and participate even though you University’s Knowledge Media Institute (KMI) and the Mobile Bristol team at HP Labs Bristol. The authors would like to thank are outnumbered this time. Using the awareness fea- the following people for their significant input to the implementa- tures on your device and carefully observing the people tion of CitiTag and their assistance with the user trials: Kevin around you, you manage to find another four Reds. You Quick and Jon Linney (KMi), who programmed the multiplayer team up to follow unsuspecting Greens and spread Red networking capabilities, Richard Hull (HP Labs), Ben Clayton, Tom Melamed (University of Bristol), Paul Marsh who provided around. Then a large group of people emerges from the the Mobile Bristol GPS location-based support for the game, Peter crowd walking quickly in your direction: they are Scott (KMi) for overall supervision and support for the project, Greens! Instinctively, you all run to avoid them and rush Lewis McCann (KMi), Stuart Martin and John Honniball towards the Underground to ‘hide’ in the wireless sig- (University of Bristol) who provided technical support during the nal-free area. The Green swarm won’t follow you there. trial and all our participants. You relax and chat with the people you just played with, exchanging your UrbanSwarmer details. Two of them are taking the same train as you and it is now time to References go!’’ 1. Ericsson KA, Simon HA (1996) Protocol analysis: using verbal reports as data. Oxford University Press, Cambridge, MA 2. Farkas I, Helbing D et al (2002) Mexican waves in an excitable In this scenario we describe an everyday experience: medium. Nature 419(6903):131–132 players can drop-in and dropout of a game depending 3. Garneau P-A (2002) Emergence: making games deeper. At http://www.pagtech.com/Articles/Emergence.html on their current activities (e.g. taking advantage of 4. Gero JS (1994) Towards a model of exploration in computer- ‘dead’ waiting time). Notice the emergence: the Reds aided design. In Gero, Tyugu (eds) Formal design methods for team up to form a larger group and move altogether. computer-aided design. North-Holland: Amsterdam, pp 271– They also use the environment creatively: the London 291 Underground becomes a ‘virtual’ hiding place, while still 5. Hull R (2002) Mobile bristol application framework (white paper), Technology & Lifestyle Integration Programme, Hew- being physically visible. This game would be based on lett-Packard Laboratories, Bristol. Available at http:// the following parameters, based on the findings of the www.mobilebristol.com/PDF/MobileBristol(W)-2002–01.pdf CitiTag studies: 6. Hull R, Clayton BT, Melamed T (2004) Rapid authoring of mediascapes. In: Proceedings of UbiComp 2004. The 6th (1) Team play and tactics: by clustering and swarming international conference on . Notting- the city together in groups, UrbanSwarm players ham, England 7. Johnson S (2001) Emergence: the connected lives of ants, would acquire ‘group power’ and become stronger brains, cities and software. Penguin Books, Baltimore against their opponents. Trade-offs and challenges 8. Kahn R, Kellner D (2004) New media and internet activism: introduced in the game would encourage coopera- from the ‘Battle of Seattle’ to blogging. New Media Soci tive play, strategies and vary the experience. 6(1):87–95 (2) Presence awareness: information on nearby players 9. Macromedia website, http://www.macromedia.com/ 1995–2004 10. Milgram S (1977) The individual in a social world: essays and displayed as simply as possible for the mobile con- experiments. Addison-Wesley, Reading text would provide ‘just enough’ to harness players’ 11. Mitchell WJ (1993) A computational view of design creativity. imagination, to encourage them to interact with In: Gero M (ed) Modelling creativity and knowledge-based people around them and to explore the environment creative design. Lawrence Erlbaum, pp 25–42 12. Mobile Bristol http://www.mobilebristol.com/ in creative ways. 13. Nardi B, Whittaker S et al (2000) Interaction and outeraction: (3) Persistence and drop-in experiences: players would be instant messaging in action. CSCW’2000, ACM Press able to participate in the game spontaneously, 14. Norman D (2002) The design of everyday things. Basic Books, depending on their activity and mood. To enable a London 15. Opie I, Opie P (1969) Children’s games in street and play- sense of continuity, presence in the game and par- ground. Clarendon Press, Oxford ticipation in the community of players we would 16. Paulos E, Goodman E (2004) The familiar stranger: anxiety, develop scenarios that introduce history and social comfort, and play in public places, ACM SIGCHI reputation or rumour in the game. 17. Quick K, Vogiazou Y (2004) CitiTag multiplayer infrastruc- ture, Tech Report KMi-TR-138, Knowledge Media Institute, The deployment of this game as a real world appli- The Open University, Milton Keynes, UK, http://www.kmi.o- cation would bring valuable insight into the kind of pen.ac.uk/publications/techreports-text.cfm 18. Raby F (2000) Project No.26765: flirt. RCA, London implications this can have on patterns of everyday life. 19. Reid J, Geelhoed E, Hull R, Cater K, Clayton B (2005) Parallel Would people create social bonds with others they meet world’s: immersion in location-based experiences. CHI, Port- through UrbanSwarm? Would social groups or Urban- land Swarm ‘gangs’ emerge, regularly meeting at particular 20. Rheingold H (2002) Smart Mobs: the next social revolution. Perseus, Cambridge 58

21. Rettie R (2003) Connectedness, awareness and social presence. 26. Suwa M, Gero JS, Purcell T (1999) Unexpected discoveries and In: 6th international presence workshop, Aalborg S-invention of design requirements: a key to creative designs. In 22. Saunders R, Gero JS (2001) Designing for interest and novelty, Gero, Maher (ed) Computational models of creative design IV. motivating design agents. CAAD Futures Key centre of design computing and cognition. The University 23. Scho¨n DA, Wiggins G (1992) Kinds of seeing and their func- of Sydney, Sydney, pp 297–320 tions in designing. Des Stud 13:135–156 27. Vogiazou Y, Eisenstadt M, Dzbor M, Komzak J (2005) From 24. Stolterman E (2002) Uninteded use, shaping the network Buddyspace to CitiTag: large-scale symbolic presence for society, patterns for participation, action, and change. Seattle community building and spontaneous play. ACM symposium 25. Stone H, Sidel J, Oliver S, Woolsey A, Singleton RC (1974) on applied computing, Santa Fe, New Mexico, pp 13–17 Sensory evaluation by quantitative descriptive analysis. Food Technol 28(11):24–34