Interaction-Driven Virtual Reality Application Design

Transcription

1 Nar s Parés npares@iua.upf.es Ro Parés rpares@iua.upf.es Audiovisual Institute, Universitat Pompeu Fabra, Pg. Cirumval. laió, Barelona, Spain gvirtual Interation-Driven Virtual Reality Appliation Design A Partiular Case: El Ball del Fanalet or Lightpools Abstrat Virtual reality appliation design is usually guided by a ontent-driven strategy, whih gives priority to the appliation s ontent and ontext. In this paper, we shall desribe and study a novel strategy in VR appliation design that is entered on the design of the user interation, regardless of the spei ontent of the appliation. This is espeially useful in reative/artisti appliations of VR. We shall present the spei ase of an artisti VR appliation from whih this strategy has emerged. This VR experiene, El Ball del Fanalet or Lightpools has been suessfully presented at the Miró Foundation in Barelona (Spain). 1 Introdution Presene, Vol. 10, No. 2, April 2001, by the Massahusetts Institute of Tehnology The development of VR appliations has undoubtedly evolved from the development strategies of omputer appliations in general. The phases of the development seem lear when the ontent or the topi on whih the appliation is based drives the whole development proess (Kalawsky, 1993; Waxelblat, 1993; Loef er & Anderson, 1994; Bryson, 1995; Earnshaw, Vine, & Jones, 1995). The topi de nes a ontext, and the ontext is determinant in the hoie of the metaphor used in the interation elements and interfae (Erikson, 1990). These strategies, whih we all ontent-driven strategies, are naturally suited to the development of VR appliations from a wide range of ambits: from sienti simulations to arhitetural designs. Nevertheless, the ontent-driven approah exhibits ertain limitations when experimenting with new approahes to interfae design and exploring spei qualities of VR as an art prodution medium or an audiovisual ommuniation medium (Parés & Parés, 1993, 1994). In this paper, we will de ne the development phases of a general VR appliation as we understand them and then go on to state our de nition for ontent-driven strategy as the ommonly used strategy. We will then de ne the interation-driven strategy and ompare it to the ontent-driven strategy, followed by a real example of interation-driven development in an artisti VR experiene. Finally, we will extrat some onlusions from the differenes between the two strategies. 236 PRESENCE: VOLUME 10, NUMBER 2

2 Parés and Parés Development Phases of a VR Appliation We now will desribe what we onsider to be the most important phases in the development of a VR appliation regardless of the development strategy, in order to study later how these phases are organized differently depending on the appliation strategy that is hosen. These are the phases involved in simulation loop design, the interfaes (physial, logi, and mapping), objet modeling, objet behavior design, and stimuli design. 2.1 The Simulation Loop The simulation loop is the struture of a VR appliation that repeats a series of ations for the duration of the appliation s life. This repetition of ation may determine the time measure of the appliation by de ning yles of a ertain length, or it may be adapted to a spei internal lok that determines the time evolution of the appliation. These ations manage the ontrol of the appliation at every yle, that is, at every instant in time. Its essential parts are Sensor management: read the input data from sensors deteting user movements and deisions. Renew user state: modify the general properties, situation, and status of the virtual subjet, aording to the data aptured by the sensors. Renew objet state: modify the general properties, situation, and status of the virtual objets, from their behavior algorithms and the deisions that these entail. Determine interations: detet whether the modi- ations of the virtual subjet and the virtual objets lead to any interations that an produe a hange in state of the appliation and its elements. Management of the sensorial outputs: generate all the stimuli that the experiene is omposed of (suh as images, sounds, movement, and touh) in order to give oherent feedbak to the user aording to the instantaneous ontext. The simulation loop is strutured around a set of important development phases that inlude identifying the types and soures of data used and the proesses involved in the appliation. 2.2 The Interfaes Interfae design is extremely important beause it de nes a two-way ommuniation between the user and the experiene, allowing the user to pereive and modify the experiene. The stages of interfae design an be divided into three essential parts: deiding whih external hannels will ommuniate with whih internal hannels and how (mapping), determining whih elements will be the external appliation links (physial interfaes), and determining whih elements will be the internal links (logi or software interfaes). A very simple example is the ombination mouse-ursor in a standard Windows environment. The physial interfae is the mouse itself; the software interfae is the ursor; and the mappings are the relation between mousemovement units and ursor-movement units, the relation between the pressing of a mouse button and the ursor s reation seleting a ertain objet, and so forth Mapping. Mapping refers to the way in whih the user s sensorial hannels are linked with the ations that s/he ontrols and the stimuli s/he reeives in the VR experiene. These mathematial links let us de ne what will alter what and how muh. This not only hanges how the environment reats to the user, but also modi es his/her pereption of it. For instane, in the mouse-ursor example, the mapping between mouse displaement and ursor displaement is ruial for pratial reasons. That is, if a large displaement of the mouse is needed to move the ursor one sreen unit, it might make the task of moving sreen objets too slow. On the other hand, if a small displaement of the mouse makes the ursor move a large sreen distane, then the task of moving sreen objets might be

3 238 PRESENCE: VOLUME 10, NUMBER 2 impossible beause of a lak of ontrol. These two extreme ases not only desribe pratial interation issues but also affet how the user pereives the sreen spae (world). In the rst ase, the world seems very large in relation to the user, beause it takes a great effort to move from one plae to another, whereas, in the seond ase, the world seems extremely small, beause the user moves very quikly from one end to the other Physial Interfaes. Mapping requires data input and output through a set of devies that physially onnet the user s output hannels to a software interfae inside the appliation, and data output from the appliation to the input sensorial hannels of the user through a set of displays (not only visual displays). These are the physial interfaes Software (or Logi) Interfaes. In simulation appliations in general, the software interfae is assoiated with the metaphor hosen in the appliation ontext and is a virtual representation of the physial interfae (for example, a virtual hand when a dataglove is used or a virtual steering wheel in a vehile simulator). An interesting exeption to keep in mind is the mouse-ursor example, in whih the ursor has no relation whatsoever either with the mouse or with the desktop metaphor. As we shall see, this idea of a software interfae that is not a diret simulation is enhaned by the new appliation-development strategy that we are presenting, and opens a wide range of possibilities for audiovisual ommuniation and artisti VR appliations. 2.3 Objet Modeling Objet modeling is an important phase in the development of a VR appliation beause objets must be arefully tuned (polygon ount, texture size, and so on) to meet the omputing-power onstraints, the visual requirements, and the de ned interation. In simulation appliations, one may hoose to use highly tuned objets from objet libraries, but, in artisti appliations, one may not be able to use prede ned objets. 2.4 Objet Behavior Objet behavior is a omplex task both beause of its inherent dif ulties (suh as intelligene, reation, and spontaneous ations) and beause of the lak of general and homogeneous tools to develop the behaviors. A similar issue is found here as in objet modeling: that is, simulation appliations may use behavior libraries, whereas artisti appliations probably have to develop spei behaviors. 2.5 Stimuli Both generated and aptured stimuli must meet ertain oherene onditions with respet to the human user. Undesired interations between stimuli may ause instability and disomfort in the user (MCauley & Sharkey, 1992) for example, if the appliation does not offer kinaestheti feedbak in aordane to the visual feedbak. This makes it an espeially deliate phase in the VR appliation-development proess. Artisti appliations might ignore these oherene onditions in their searh for new relations with users mental and physial models. 3 Content-Driven Strategy Now that we have introdued the development phases for VR appliations, let us turn to analyze how these phases are arranged when the ontent drives the development. Content-driven strategy may also be referred to as a top-down strategy, beause it starts at a high abstration level, analyzing the ontent and ontext of the appliation, and works down to the implementation of it. The sequene of stages that we de ne in this strategy inluding implementation, problem evaluation, and user requirement stages are the following. De ne the appliation topi or theme: suh as airraft pilot trainer, laparosopi surgery trainer, aerodynamis analyser. De ne the type of appliation: explorative, ma-

4 Parés and Parés 239 nipulative, or ontributive interation (Parés, 1999). Identify the type of user: expert, ertain knowledge of the topi, general publi, and so on. Identify the neessary virtual objets: landsape objets (vegetation, buildings, et.), internal organs, wing model, et. Identify the data involved: suh as input data, partial results, nal results, and data types. Identify proesses: suh as algorithms, omputations, behaviors, and subjaent laws. Identify input interfaes: suh as type of sensors needed, user interfaes, links, and mappings. Identify output interfaes: suh as results presentation and type of display peripherals. Identify objet-modeling tools needed: geometri modeling, generative algorithms, CAD tools, sanner inputs, et. Identify appliation development tools: suh as programming libraries, developing environments with GUI, and sensor drivers. This type of development strategy is naturally suited to the analysis of almost any omputer appliation that is based on a ertain ontent (theme or topi). Note that some of these stages are not neessarily sequential, but they may be analyzed in parallel with others next to them in the list. 4 Interation-Driven Strategy Although most VR appliations would follow the ontent-driven strategy, there are ertain ases in whih it an be very useful to develop an appliation by onentrating on how the user is to interat with the appliation. In other words, analyzing the interfaes, interation with the elements, and partiipation/ manipulation/ontribution of the user, in suh a way that the obtained results allow a spontaneous emergene of, for example, the spei topi, ontent, aroma, and tone, of the appliation. Suh may be the ase when the developers are interested in testing ertain interation theories through a series of experiments. The resulting experienes are ontextualized only when the appliation has its nal format to be presented to the users. Another ase in whih this might be useful is artisti approahes in whih the artist is trying to onvey onepts and relations through the interation proposed, as opposed to trying to propose a ertain topi or aestheti value. This interation-driven strategy may also be referred to as bottom-up strategy, not beause it is the exat opposite of the ontent-driven (or top-down) strategy, but beause the development begins with identifying the interfaes, sensors, and mappings, and, at the same time, proesses and behaviors, instead of starting at the de nition of the appliation topi. Therefore, the development stages for the interation-driven strategy are rearranged in the following order. Identify input interfaes: suh as type of sensors needed, user interfaes, links, and mappings. Identify output interfaes: suh as results presentation, type of display peripherals, and mappings. Identify the type of user: suh as expert, ertain knowledge of the topi, and general publi. De ne the type of appliation: explorative, manipulative, or ontributive interation. De ne the appliation topi or theme: nd a related metaphor by analyzing the type of interation. Identify proesses: suh as algorithms, omputations, behaviors, and subjaent laws. Identify the neessary virtual objets: objets related to the interation and the metaphor, although they may be ompletely abstrat. Identify the data involved: suh as input data, partial results, nal results, and data types. Identify objet-modeling tools needed: geometri modeling, generative algorithms, et. Identify appliation development tools: suh as programming libraries, developing environments with GUI, and sensor drivers. This strategy failitates the design of software interfaes that do not neessarily represent physial interfaes beause there is no a priori ontent that determines them. This also opens the possibility of de ning mappings that

5 240 PRESENCE: VOLUME 10, NUMBER 2 do not respond to models extrated from our physial surroundings. The freedom in the design of both these elements of a VR appliation enhanes the development of innovative interation, even to the extent of designing experienes in whih the stimuli, whih are fed to the user, do not onform to oherent systems in the sense that a ontent-driven appliation would fore. Here, the metaphor and the appliation ontext for the interfae should emerge from a lose study of the proposed interation, whih skethes analogies traed from this interation to the ontext from whih the metaphor is extrated. 5 El Ball del Fanalet or Lightpools El Ball del Fanalet or Lightpools (Hoberman, Parés, & Parés, 1999), an artwork developed by Perry Hoberman and Galeria Virtual (Parés & Parés, 1995), follows the interation-driven strategy. In fat, the development of this piee made us aware of the existene of this new strategy. We will rst explain what the piee is about in its nal format. Although this might mislead the reader in thinking that it is a ontent-driven appliation, we will later desribe its development stages, whih have led to the struturing of the interationdriven VR appliation-development strategy. 5.1 Desription of the Piee El Ball del Fanalet or Lightpools is a multiuser experiene that uses virtual reality tehnology. It takes plae in a irular arena approximately six meters in diameter, onto whih a real-time, omputer-generated image is projeted from above ( gure 1). Eah user is given one of four fanalets (a paper lantern typial of a Catalan popular dane) ( gures 2 and 3) as they enter the arena, and eah ontains a olored light with a battery pak and a position sensor. The sensor reports its position to a host omputer through ultrasound pulses, allowing eah fanalet s position to be traked in threedimensional spae. The two horizontal dimensions are used to position a olored irle of light projeted onto the oor diretly below eah fanalet. The third dimension (height) is used to determine the size and brightness of the lightpool, so that its behavior mimis the effet of a light soure emanating from the fanalet: as the fanalet is lowered, the pool beomes smaller and more intense; as it is raised, the pool beomes larger and dimmer. The users thus have the impression that the lightpool is projeted diretly from their fanalet, and gives them an immediate, intuitive sense of how to interat with the work. Eah lightpool is a kind of window onto a virtual ground plane that otherwise remains shrouded in darkness. Small, olored tetrahedrons (proto-objets) are spontaneously and randomly generated throughout the arena, glowing brie y like embers before disappearing. Eah proto is mathed in olor to one of the four fanalets (and lightpools). If a user manages to illuminate a proto (during its brief life) with the appropriate lightpool at a suf ient intensity, the proto grows and metamorphoses into an artiulated objet. In effet, the objet appears to feed on light. The objets range from mehanial to biomorphi, abstrat to ornamented. When the objet has reahed suf ient size, it remains stable, and is thereafter under the ontrol of the user ( gures 4 and 5). At this point, if the user fails to derease the intensity of the light (by raising his or her fanalet), the objet grows until it bursts, sattering a new rop of olored protos onto the oor. Other users an then nurture these protos, and the yle starts again. One an objet is grown and has stabilized, it beomes the user s partner and an be trained to dane. Any suf iently rapid sequene of movement of the fanalet is interpreted as a start of this training, whih makes the partner learn every new movement of the fanalet. The partner follows and memorizes this movement until the fanalet remains still, and then starts to repeat it ontinuously on its own (until it is taught a different sequene). Slower movements of the fanalet are interpreted as leading movements, whih the trained partner follows as it danes around the lightpool. By alternating rapid and slow movements of the fanalet, users an teah, dane, and interat with their partner. Although the projeted image is two-dimensional, the partners themselves are represented in three dimensions,

6 Parés and Parés 241 Figure 4. A user growing an objet to obtain a dane partner. Figure 1. Setup of El Ball del Fanalet or Lightpools. Figure 2. Tehni ed Fanalet (paper lantern). Figure 5. A user interating with his/her new partner. Figure 3. The typial Catalan popular dane: El Ball del Fanalet. so that upward and downward motions of the fanalet affet the partner s orientation and sale. Eah partner is omposed of three distint parts (subobjets) that follow the motion of the fanalet at slightly different rates; this gives the partners a somewhat uid appearane; ertain details are revealed only when the partner is in motion (a bit like a peaok showing off its feathers). If the user trains a partner for too long, the partner beomes Figure 6. The fanfare and shower of lights that preede the autonomous dane of the objets and lightpools. bored and leaves the arena through the exit in the railing, leaving the user with an empty lightpool. The user must then nurture a new proto to start anew.

7 242 PRESENCE: VOLUME 10, NUMBER 2 From these interations emerges a omplex, everhanging dane of partiipants and virtual objets. New users an enter at any time (as long as there is an available fanalet), and users an stay as long as they like. Additional spetators an omfortably wath from behind the irular railing that enloses the arena. Several users an share a single fanalet, passing it bak and forth as they wish. The atmosphere is one of a asual but onentrated haos, as users observe and interat with their objets and eah other. Eah event in the piee has an audio omponent as well: a deliate glassy tinkle as protos appear, the whoosh of a pneumati pump as objets grow, whipping air urrents as objets are trained, and so on. A seond, independent, stereo audio trak onsisting of a variety of ballroom dane musi (proessed with huge amounts of reverberation) plays ontinuously and softly in the bakground, as though it is being heard from a great distane. After two or more users have grown and trained stable objets, they may, at any point, bring their fanalets together in a ritual gesture that indiates that they want their objets to join together in a horeographed objetdane. After a brief fanfare and a shower of light ( gure 6), the lightpools and objets leave the user behind and perform a group dane to a driving drumbeat, using the movements that they have been taught as they follow a horeographed path. The lightpools then exit the arena leaving the objets outside, and ome bak to the users to begin the yle again. El Ball del Fanalet or Lightpools is, in part, an investigation of the proposition that, given the urrent state of available tehnology, it makes some sense to projet an image onto the oor, instead of an upright wall. Of the three dimensions (height, width, depth), we are suggesting that height might be the most easily expendable. This strategy allows immersion or entrane into a two-dimensional image. Projeting the image onto the oor also fores new strategies of spatial representation, beause it effetively makes perspetive spae inomprehensible. The projet also suggests a number of possible relationships that an be experiened within a publi spae in an atmosphere that is reminisent of a ballroom or skating rink. We are interested in the area of what might be alled asual or open-ended interativity, an area that remains underexplored. Most media installations are still designed for a single user or for remotely networked users, and most still require users to remain immobile. We are trying to reate the onditions under whih an interative, immersive image an be experiened as a asual, open, soial spae. After its installation at the Miró Foundation in Barelona, this piee has proven to be very suessful in breaking the fear or apprehension that is often found in users who have no previous experiene in VR installations. Although there was no formal psyhologial study of the users reations or appreiation of the piee, after 7,000 users from very different origins and ultures 1 had passed through it in fteen days, we observed that very few had troubles in their interation and that most of them had a suessful experiene with the piee. We believe this is due to the fat that the user is led into the experiene through a very natural and lear metaphor that has smoothly emerged from the designed interfae and whih is well suited to the proposed interation. We onlude this has been favored by the interation-driven strategy as will be seen in setion The Interation-Driven Development of El Ball del Fanalet or Lightpools Now we will desribe the steps taken in developing this piee. Although this should be read as a ontinuous desription, we have tagged the setions orresponding to eah stage in the interation-driven strategy to guide the reader in the mapping of the struture of the interation-driven strategy onto the design proess Identify Input and Output Interfaes. Our main goals were to design a multiuser interfae that would allow the users to engage in a soial interation among themselves and would not alienate them as multiuser VR appliations usually do. 1. The Miró Foundation, being a world-renowned museum, is one of the most visited museums in Barelona and its visitors range from shool trips to tourists from around the world.

8 Parés and Parés 243 In this manner, we developed the idea of an interfae in whih the environment would be projeted onto the oor and would be large enough to hold up to four simultaneous users. We did not want the entire environment to be visible at one within the interation area. We deided the users should be able to see only a delimited area around them, foring them to move about in order to understand the whole environment. Hene, some sort of wireless position sensor would be neessary to trak the users position, allowing the system to reveal the areas of the environment through whih they pass. It was important to us that the users would have to ollaborate with eah other to be able to aquire omplete understanding of the experiene, by sharing their limited personal view with the others Identify the Type of User. At the same time, we wanted the piee to be aessible to a very wide range of people. Therefore, the experiene needed to provide simple interation mehanisms, although the relations derived from their interation ould be rih and diverse De ne of the Type of Appliation. We also wanted the appliation to be an explorative and manipulative experiene, with a ertain degree of ontribution on the users part. These ontributions would be hanneled through the hange in behavior of the objets in the environment De ne of the Appliation Topi or Theme. At this point, we still did not know what the atual appearane of the experiene would be like or whih type of objets we would design. Based on the aforementioned premises, we thought an appropriate metaphor for suh a multiuser experiene would be that of a dane oor. We hose a typial Catalan ouples dane known as El Ball del Fanalet, whih seemed partiularly appropriate for the following reasons: It is a soial dane, hene promoting user interrelations and not only user-environment interation. The ouple in the dane holds a paper lantern (a fanalet) with a lit andle inside, whih must be arefully moved while daning smoothly so as to avoid having the whole lantern ath re. This fanalet was very appropriate as a metaphor for the exploration of the environment through the limited visible area that we wished to impose on the users. This area evolved into the virtual light spot, or lightpool, whih would follow the user around the dane oor. The light projeted by the fanalet ould be used as energy to unleash hanges in the environment as an interation element. With this frame of referene, we deided to searh for all possible dane relations: A single user: The sole fat of having the user exploring the environment would already sketh a dane senario. User with virtual objet: We deided that the de- ned objets should funtion as dane partners for the user. User with user and virtual objet with virtual objet: One eah user has a virtual dane partner, a user might approah another user to dane with him/her and at the same time make their virtual objets dane with eah other. Virtual objet with virtual objet: We deided that there should be a mehanism through whih virtual objets ould start a dane between them, beoming ompletely autonomous from their owner/user. All these deisions led to the nal design of El Ball del Fanalet or Lightpools and its omponents Identify Proesses. The main algorithms developed were the behavior algorithms for the lightpools and the objets (from protos to partners). These were developed through the implementation of nite automata, but we will not desribe them here Identify the Neessary Virtual Objets. We wanted the virtual objets to meet ertain requirements: They should not represent anything in partiular, beause there was no spei topi to whih they

9 244 PRESENCE: VOLUME 10, NUMBER 2 would be related. The dane oor metaphor had set a ontext, but we did not want to represent any partiular dane oor nor de ne any virtual human daners. They should be able to evolve in size and shape. They should exhibit interesting movements that ould be appropriate to the notion of daning in abstrat terms. With these properties in mind, small objets were designed as partiles, so that they would start as protoobjets with the potential of being nurtured to grow to a more de ned shape. These mature objets would beome the dane partners for the users. The partners would follow the lightpool that made them grow with an elasti link, whih would ause a delay in their response to the lightpool s movements. Eah partner would in fat be omposed of three subobjets, eah of whih would show a progressively larger delay suh that the elasti movement would be exaggerated and the partner would break up into its parts while moving and reompose while at rest. This behavior would reate a train of movement that would aentuate the sensation of daning and horeography Identify the Data Involved. The input data was identi ed as the position hanges of the wireless sensors in the fanalets, beause these hanges were to unleash the different reations and states of the lightpools and onsequently of the objets (both protos and partners) Identify Objet-Modeling Tools Needed. We deided a standard modeling tool was to be used Identify Appliation-Development Tools. Beause great exibility and algorithmi power were required, programming with a standard programming language suh as C, enhaned by high-level VR libraries (WorldToolKit from Sense8), was found to be the best approah. As an be seen from the above desription, the development of the experiene (from the oneptual stages to the atual implementation) did not take into aount the ontent of the nal design until late in the proess. This method radially hanges the way developers think of the appliation beause the user interation is the entral issue. To summarize, in this piee we started by de ning the users interation (suh as relations between users (soial aspets), relations between users and environment, and view of a limited area of the environment) and the interfae (the limited area that the user ontrolled). This led us to nd appropriate metaphors that would enhane the interation: a dane oor and light spots. Finally, the metaphors set the ontext for the ontent of the experiene, that is, the nal aspet, feel, and setup. 6 Conlusions We have presented a novel strategy for developing VR appliations that proves to be very powerful in experimental, reative, and artisti VR experienes. The idea behind this new strategy is to onentrate on user interation rather than onentrating on the ontent spei ed by the appliation s topi; hene, we have named it interation-driven strategy as opposed to ontent-driven strategy. In fat, there is no a priori appliation s topi when approahing development in this manner. This provides great design freedom to the developers in the de nition of new interfaes and new interation relationships that allow for experimentation in user approah and reation to VR appliations. This freedom is due to the fat that this strategy does not x the appliation topi and its related metaphor, but rather lets the metaphor emerge naturally from the de ned interation and sets the topi through the metaphor. The interation-driven strategy was found experimentally while developing an artisti multiuser VR appliation, for whih we had imposed upon ourselves the restrition of designing rst how we wanted the user to interat. We worked under this restrition to onentrate on the soial aspets of multiuser VR interation and interfaes. We have desribed both the piee and its development proess and have shown how the interation-driven strategy has enhaned experimentation with interation and interfae aspets in the nal result.

10 Parés and Parés 245 It is lear that one is working under different frames of referene when developing a VR appliation with the ontent-driven strategy than with the interation-driven strategy. The whole bakground positioning is turned upside-down, giving a predominant position to the ontent and topi of the experiene, or to the ways in whih the users approah the experiene and interat, respetively. On evaluating the differene between the two, it is lear that the use of one or the other depends upon the desired nal result. In interation-driven development, taking advantage of the high level of freedom in the interation model of the experiene, one may be muh more reative and innovative. One may de ne types of users that would not be possible if the theme and ontext of the appliation were xed from the beginning. On the other hand, in ontent-driven development, one may be muh more rigorous when de ning requirements, and the lak of exibility in the interation model makes everything more ontrollable under standard and known frames. The ontent-driven model is used in all appliations in the simulation and sienti ambit, whereas the ambit of artisti prodution may hoose to use a ontentdriven strategy for appliations with a spei topi, or an interation-driven strategy for appliations with an important experimental use of interation strategies. It is important, at this point, that many more appliations be developed with this new strategy to be able to evaluate it in greater detail. The two strategies that have been studied here are not the only two that an be obtained from the analysis of the development stages of a VR appliation. Other strategies may be de ned and studied in the future, but we have espeially desribed the interation-driven strategy beause we have obtained it experimentally from our artisti appliation. Aknowledgements We would like to thank Perry Hoberman for sharing his experiene throughout the whole prodution proess of El Ball del Fanalet or Lightpools. Also, we thank the Miró Foundation for its support and ontribution to the prodution and rst installation of the piee. We would also like to thank X. Artigas and R. Franh for their exellent work in developing the piee. Finally, we want to thank espeially Orit Kruglanski for her exellent onstrutive ritiism to both the style and ontent of this artile. Referenes Bryson, S. (1995). Approahes to the suessful design and implementation of VR appliations. In R. A. Earnshaw, J. A. Vine, & H. Jones (Eds.). Virtual reality appliations (pp. 3 15). London: Aademi Press. Earnshaw, R. A., Vine, J. A., & Jones, H. (Eds.). (1995). Virtual reality appliations. London: Aademi Press. Erikson, T. D. (1990). Working with interfae metaphors. In B. Laurel (Ed.), The art of human omputer interfae design (pp ). London: Addison-Wesley. Hoberman, P., Parés, N., & Parés, R. (1999). El Ball del Fanalet or Lightpools. Proeedings of International Conferene on Virtual Systems and Multimedia (pp ). Dundee, U.K.: VSMM 99. Kalawsky, R. S. (1993). The siene of virtual reality and virtual environments. Reading, MA: Addison-Wesley. Loef er, C. E., & Anderson, T. (Eds.). (1994). The virtual reality asebook. New York: Van Nostrand Reinhold. MCauley, M. E., & Sharkey, T. J. (1992). Cybersikness: pereption of self-motion in virtual environments. Presene: Teleoperators and Virtual Environments, 1(3), Parés, N., & Parés, R. (1993). Galeria virtual. Proeedings First Eurographis Workshop on Virtual Environments. Eurographis Tehnial Reports Series (pp ).. (1994). Galeria virtual: A platform for non-limitation in ontemporary art? Proeedings Virtual Reality and Appliations. Leeds, U.K.: British Computer Soiety.. (1995). Galeria virtual: A platform for non-limitation in ontemporary art? In R. Earnshaw, J. Vine, & H. Jones (Eds.), Virtual reality appliations (pp ). London: Aademi Press. Parés, N. (1999). Charaterization, issues and properties of artisti VR appliations. VR as an art prodution medium. Pre-thesis researh work in the audio-visual ommuniation dotoral program. Universitat Pompeu Fabra, Barelona, Spain (in Catalan only). Waxelblat, A. (Ed.). (1993). Virtual reality appliations and explorations. London: Aademi Press.