In vivo, in vitro, in silico, in virtuo 1 Introduction - Indico [PDF]

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

1

In vivo, in vitro, in silico, in virtuo Jacques TISSEAU UEB-ENIB, LISyC, Brest University European Center for Virtual Reality CERV, 25 rue Claude Chappe, 29280, Plouzan´e, France [email protected]

1

Introduction Complex systems are everywhere, whether in biology - the greatest source of inspiration - or in physics, data processing or economics. New analytical methods are vital if we want to study these systems in detail without relying on simplifications which can misrepresent or distort them. [...] To study complex systems, we could try to apply conventional analytical methods: writing up a set of equations which are supposed to describe the overall behavior of the system, and solving those equations. But generally speaking, we don’t know how to write them. Zwirn H., La complexit´ e, science du XXIe`me si` ecle ?, [Zwirn 03]

The systems we want to model are increasingly complex. This complexity is essentially due to the diversity of components, the diversity of structures and the diversity of interactions put into play. Therefore, a complex system is, on the face of it, an environment which is open (components appear/disappear dynamically), heterogeneous (varied behaviors and morphologies) and made up of mobile composite entities distributed over space and in varying numbers over time. These components, amongst which humans and their free will often play a determining role, can be structured into different levels, either known from the start or emerging as they evolve due to multiple interactions between the components. The interactions themselves may be of different types and operate on different spatial and time scales. But as the quotation heading this section suggests, no theory is currently capable of formalizing this complexity, and in fact there are no a priori methods of formal proof like those which exist in highly formalized models. Without such formal proof, we must fall back on experimenting on a system as it develops, in order to be able to experimentally validate it afterwards. Today, virtual reality provides a conceptual, methodological and experimental framework which is well adapted to imagining, modeling and experimenting on this complexity. Virtual reality is a scientific and technical field using information technology and behavioral interfaces to simulate the behavior of 3D entities in a virtual world. They interact with each other and with one or more users in real time, through pseudo-natural immersion via sensory motor channels. Fuchs P., Arnaldi B., Tisseau J., La r´ ealit´ e virtuelle et ses applications, [Fuchs et al. 03]

According to the definition above, derived from VR work carried out by the French scientific community, virtual reality simulation therefore allows true interaction with the modeled system. Like a biologist performing in vitro experiments, this simulation lets us observe phenomena as if we had a virtual microscope which can be moved and oriented as we wish, with a choice of focal points. The cre-actor user (spectator-actor-creator) can thus focus on observing a specific type of behavior, a subsystem’s activity or the overall activity of the system. The user can interrupt the phenomenon at any time, accurately focus on the bodies present and the interactions underway, and then restart the simulation where it was stopped. At any time, by using sensory-motor behavioral interfaces, the user can disturb the system by modifying the property (status, behavior) of an element or by removing elements or adding new ones. This means the user can test a specific behavior, or more generally speaking, an idea, and immediately observe the consequences on the system under operation. Virtual reality puts the user at the heart of the virtual laboratory, bringing him closer to the methods of experimental science while providing access to digital methods. In virtuo (in the virtual) is a newly coined expression constructed by analogy to adverbial phrases from Latin such as in vivo (in the living) and in vitro (in glass). Biologists often use the expression in silico (in silicon) to describe computer calculations, however, in silico fails to conjure up human participation in the world of digital models being run: which is why we prefer the expression in virtuo, whose common root provides a reminder of the experimental conditions of virtual reality. Tisseau J., R´ ealit´ e virtuelle : autonomie in virtuo, [Tisseau 01]

Going beyond simply observing a digital model being performed on a computer, the user can test its reactivity and adaptability while it is running, thus taking advantage of the behavioral nature of digital models. We call this new type of experiment in virtuo experimentation. An in virtuo experiment is one

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

2

carried out in a virtual world of interacting digital models in which a human being is taking part. Virtual reality fully involves the user in the simulation, which is closely akin to the participatory design approach [Schuler et al. 93], preferring to consider users as human actors rather than human factors [Bannon 91]. In VR, this sort of participatory simulation implements various types of models (multi-model) from different fields of expertise (multidisciplinary). It is often complex, since its overall behavior depends as much on the behavior of the models themselves as on the interactions between models. Lastly, it must include the free will of the human user making use of the online models. In virtuo experimentation also involves an experience which just analyzing digital results does not provide. Between a priori formal proof and a posteriori validations, there is room today for virtual reality experienced by the user, who can thus go beyond generally accepted ideas to ideas grounded in experience. We suggest that this in virtuo experimentation be placed at the heart of the 21st century laboratory, so that the inherent complexity of systems, whether natural or artificial, can be made intelligible. Microscope, telescope: these words evoke the great scientific penetrations of the infinitely small and the infinitely great. [...] Today, we are confronted with another infinite: the infinitely complex. And this time we have no instrument to use. We have only our brain - our intelligence and our reason - to attack the immense complexity of life and society. [...] We need, then, a new instrument. [...] I shall call this instrument the macroscope (from macro, great, and skopein, to observe).The macroscope is unlike other tools. None of that has occurred by chance. It is a symbolic instrument made of a number of methods and techniques borrowed from very different disciplines. It would be useless to search for it in laboratories and research centers, yet countless people use it today in the most varied fields. The macroscope can be considered the symbol of a new way of seeing, understanding and acting. De Rosnay J., Le macroscope : vers une vision globale, [De Rosnay 75]

Although modeled systems are increasingly complex, the formalism which could truly describe their complexity is still lacking today. Only virtual reality enables this complexity to be experienced. Therefore, we must further explore the relations between virtual reality and theories of complexity, so that virtual reality becomes an instrument to investigate complexity, as in the “macroscope” imagined by Jo¨el de Rosnay in the 1970s. But we prefer the term of “virtuoscope” to macrosope, since it reminds us that these systems are studied, first and foremost, through the models we make of them and experiment on in our virtual laboratories. In the long term, the virtuoscope project should provide scientists from all fields with methods and instruments to study complex systems within virtual laboratories by implementing the in virtuo experiments that VR can provide. To support this flagship project, we shall begin by drafting the foundations of this virtual laboratory of the 21st century, taking virtual reality as a base to build on (Section 2). We can then use this new construction’s epistemological position to shed new light on the now-classical methods of scientific reasoning (Section 3). Finally, we will highlight the main stakes, whether scientific, methodological, technological or societal (Section 4).

2

The virtuoscope

The virtuoscope will provide online use of digital models, based on VR concepts, models and tools (Section 2.1). Virtual reality is based on two principles, i.e., presence and autonomy (Section 2.2). Implementation of these two principles is made possible by making models autonomous (Section 2.3) in the IT framework of multi-agent systems. The latter’s simulations become participatory by fully involving users and their free will.

2.1

Making use of models

Models mainly operate through the three modes of perception, experimentation and modification. Each provides a different way of mediating reality [Tisseau et al. 98b].

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

3

Perception of a model : The model is perceived through sensory mediation: the model’s activity is observed through all of the user’s senses. The same is true for spectators in a dynamic cinema theater, watching a hemispheric screen with a sound-surround system from a seat mounted on hydraulic or pneumatic jacks, who really feels as if they are taking part and immersed in the animated film they are watching, even though they cannot modify its course. The quality of sensorial renderings and their synchronization is essential here, since this is a field where real-time animation excels. The most widely accepted definition of ”to animate” is to put into movement. In the more specific field of animated film, to animate means to give the impression of movement by running through an ordered collection of images in (drawings, photos, computer-generated images, etc.). These images are produced by applying a model for object movement in the scene shown. Experimenting on the model : Experimentation on the model brings mediation of the action into play: meaning that the user tests the model’s reactivity using the appropriate manipulators. It is like a fighter pilot using a flight simulator, whose training mainly focuses on learning how the aircraft is going to react. Based on the principle of action and reaction, here, the emphasis is on the timebased rendering’s quality: which is what interactive simulation does best. The standard meaning of simulation is to make something which is not real seem real. In scientific fields, simulation is an experiment on a model. It can be used to test the quality and internal coherence of a model by comparing its results to those obtained by experimenting on a modeled system. It is increasingly implemented today to study complex systems where humans play a role, used both to train operators and study users’ reactions. In these human-in-the-loop simulations, operators thus provide their own behavioral model which can interact with the other models. Modifying the model : Mental mediation means that the user modifies the model himself, with access to the same level of expressiveness as the modeler. This also applies to an operator who partially reconfigures a system, while the rest of it remains operational. This is the fast growing field of interactive prototyping and online modeling, in which ease of intervention and ability to express oneself are vital. To attain this level of expressiveness, the user generally has the same interfaces, and above all the same language, as the modeler at his disposal. Mediation of the mind thus is thus achieved through language mediation. In this way, the related fields of real-time animation, interactive simulation and online modeling represent three aspects of model operation. The three of them provide the three-fold mediation of what is real which virtual reality requires. They define three levels of interactivity. • Real-time animation is the baseline level of interactivity between the user and the model being run. The user is subjected to the model, since the user cannot act on any of its parameters, but is simply its spectator. • Interactive simulation is the first level of interactivity because the user has access to some of the model’s parameters. Thus, the user plays the role of actor in the simulation. • In on-line modeling, the models are themselves the parameters of the system: a higher level of interaction is reached. The user himself, by modifying the model as it runs, takes part in creating this model and thus becoming a cre-actor (creator-actor). The user can interact with the image, using the appropriate behavioral interfaces. However, within the world of models, the user can only observe or do what the system controls through its peripheral device drivers providing the indispensable links between man and machine. This means that the system is in charge of the user’s sensory-motor mediation and that this mediation is modeled within the system in one way or another. The user’s only true freedom lies in his decision-related choices (mental mediation) which are restricted by the system’s limitations in terms of observation and action. So taking account of a user must be made explicit by using a specific model of an avatar to represent him within the system. At the least, this avatar will be placed in the virtual environment in order to define the view point required to render the various senses. It has virtual sensors and actuators (for vision [Renault 90] and hearing [Noser et Thalmann 95], and hand grips [Kallmann et Thalmann 99]) to interact with other models. The data collected by the avatar’s virtual sensors are transmitted in real time to the user by

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

4

the device drivers, while the user’s orders are sent in the opposite direction to the avatar’s actuators. There are also means of communication to communicate with other avatars, which thus reinforce the sensory-motor capacities by enabling it to receive and emit language-type data. There may be no visualization whatsoever of the avatar (BrickNet [Singh 94]), or it may be limited to a simple, textured, but unstructured 3D primitive (MASSIVE [Benford 95]), part of a poly-articulated rigid-segment system (DIVE [Carlsson et Hagsand 93]), or a more realistic display which takes sophisticated behaviors like gestures and facial expressions into account (VLNET [Capin et al. 97]). When this display is available, it makes it easier to identify avatars and the non-verbal communication between them. Thanks to the user’s explicit modeling, three main types of interaction can coexist with the digital model universe: • model-to-model interactions like collisions or bonding; • model-to-avatar interactions that enable sensory-motor mediation between a model and a user; • avatar-to-avatar interactions which allow avatars to meet in a virtual environment shared by several users (televirtuality [Qu´eau 93a]). The user’s status in virtual reality is different from his status in scientific simulations of numerical computations or interactive simulation using training simulators. In scientific simulation, the user sets the model parameters beforehand, and interprets the results of the calculation afterwards. In the case of scientific visualization system, he can observe how the computations change, possibly using VR sensorial devices [Bryson 96], but remains enslaved to the model. Scientific simulation systems are model-centered systems, since science models seek to give reality universal representations which are distinct from individual impressions. In this case, the user acts as a spectator. On the contrary, interactive simulation systems are essentially user-centered, to give the user all the means he needs to control and pilot the model: the model must remain the user’s slave. The user goes from simple spectator to actor. By introducing the avatar concept, virtual reality places the user on the same conceptual level as the model. This replaces the master-slave relationship by an equal-to-equal relationships and increased model autonomy, and consequently greater autonomy for the user. Through a three-fold mediation of senses, action and language, the user becomes the true cre-actor of the models being run.

2.2

Presence and autonomy Geppetto took his tools and began to cut and shape the wood into a puppet. ”What shall I call him?” he said to himself. ”I think I’ll call him Pinocchio.” [...] Having found a name for his puppet, he set to work in good earnest to make first his hair, then his forehead and then his eyes. The eyes being finished, image his astonishment when Geppetto noticed that they moved and stared fixedly at him. [...] he then took the puppet under the arms and placed him on the floor to teach him to walk. Pinocchio’s legs were so stiff that he could not move them, but Geppetto led him by the hand and showed him how to put one foot before the other. When his legs were more limber, Pinocchio began to walk by himself and run around the room. He came to the open door and with one leap, was out into the street and was gone. Carlo Collodi, The Adventures of Pinocchio, [Collodi 1883]

Virtually reality has historically focused on the concept of the user being present inside virtual worlds [Tisseau et N´ed´elec 03]. Robotics experts were already using terms such as telepresence [Sheridan 87], telesymbiosis [Vertut et Coiffet 85] or even tele-existence [Tachi et al. 89], to describe the impression an operator can have of being immersed, the impression of being present in the place where the robot is working, although he is manipulating it remotely [Johnsen et Corliss 71, Minsky 80]. So the first VR studies were naturally turned toward designing and creating behavioral interfaces which favored the immersion of the user and his abilities of interaction in a virtual universe [Fuchs et al. 03]. These interfaces can characterize the presence of the user inside virtual worlds [Slater et al. 94, Schloerb 95, Witmer et Singer 98, Morineau 00] and shed light on the philosophical reflections about this feeling of ubiquity [Flach et Holden 98, Zahorik et Jenison 98]. To the notion of the user’s presence, we’ve added the notion of autonomy of the models which make up and structure the virtual universe [Tisseau 01]. An object’s behavior will be considered as autonomous, it if can adapt to unknown changes in its environment: meaning it must have the means to perceive, act and coordinate perceptions and actions, to be able to react realistically to these changes. This notion of autonomy is essential to combine the behavioral rendering required for VR with the multi-sensory rendering of graphic data. In fact, on a daily basis, we run up against a reality which confronts and

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

5

resists us and is governed by its own laws and not ours, which is, in a word, autonomous. Virtual reality is freed from its origins by moving towards making the digital models it manipulates autonomous, in order to populate the realistic universes that computer-generated images already let us see, hear and touch, with autonomous entities. Thus, VR is refocused on the digital models it uses just as much as on the behavioral interfaces needed to use them [Tisseau et al. 98b]. So, we can characterize a virtual reality application according to these two criteria of presence and autonomy, with presence being characterized by criteria of immersion and interaction. In this way, an application can be represented by a point in a 3D reference scheme (Figure 1): immersion / interaction / autonomy, with standardized axes between 0 (criterion entirely absent) and 1 (criterion entirely present). Within this immersion / interaction / autonomy reference, 3D cinema (1,0,0) corresponds to a typical immersion application, and a video game (0,1,0) to a typical interactive application, while a flight simulator (1,1,0) offers the user both immersion and interaction. A computer virus (0,0,1) is a typical example of an autonomous application which escapes his creator but is not, however, controlled by the user. Virtual theater (1,0,1) lets a user be immersed as an observer, free to move about in a scene played by autonomous virtual actors, but unable to influence their behaviors; on the contrary, interactive fiction (0,1,1) allows a non-immersed user to interact with autonomous actors. So a typical VR application (1,1,1) allows an immersed user to interact with autonomous virtual entities; hence, the user of this sort of application takes full part in the artificial life in these realistic universes made up of autonomous models.

autonomy

interactive fiction

computer virus virtual reality virtual theatre video game

interaction 3D cinema

immersion

presence flight simulator

Figure 1: Presence and autonomy in virtual reality This conception of virtual reality ties in with Collodi’s old dream. The latter, as we saw in the quotation at the beginning of this section 2.2 made his famous puppet an autonomous entity fulfilling the life of his creator. The steps Geppetto took to reach his goal were the same as those we have observed in virtual reality. He began by identifying him (I’ll call him Pinocchio), then turned to his appearance ([he] started by making his hair, then his forehead) and then made him sensors and actuators (then his eyes [...]). Next, he defined his behaviors Geppetto led him by the hand and showed him how to put one foot before the other) to make him autonomous (Pinocchio began to walk by himself) and finally, he could only see that making a model autonomous leads to the creator’s losing control of his model (he leaped into the street and was gone). In this way, just like the famous Italian puppet (Figure 2), models which have become autonomous take part in inventing their virtual worlds. When humans are freed in part from controlling their models, they will gain in autonomy themselves, participating in this virtual reality as spectator (observing the model’s activity), actor (experimenting on the model by testing its reactivity) and creator (modifying the model to adapt it to their needs by defining how proactive it is).

2.3

Making models autonomous Virtual objects, just like the space they are in, are actors, and agents. Endowed with a memory, they have functions to process information and an autonomy which is regulated by their programs. Virtual worlds are constantly being crossed by a strange, intermediate, artificial life. Each entity, each object and each agent can assimilated to an expert system with its own rules of behavior, which it applies or adapts in response to environmental changes or modifications in the rules or metarules which govern the virtual world. Qu´ eau P., Le virtuel, vertus et vertiges, [Qu´ eau 93b]

Enabling a model’s autonomy means giving it the means for perception and action inside its own environment, as well as a decision making module letting it adapt its reactions to stimuli which can be

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

a. user-spectator

b. user-actor

6

c. user-creator

Pinocchio, model of the Child, can also be used as a metaphor to describe the three dimensions of virtual reality. immersion The Italian puppet is a physical model: so the user-spectator has no difficulty in immersing himself in his world (since he is made of the very wood we use for firewood). interaction The puppet’s designer provided the appropriate manipulator to control its movement. Thus, thanks to the puppeteer’s control bar, the user-actor can interact with the puppet and test its reactivity. autonomy The user-creator has modified the model by making it autonomous.

Figure 2: Pinocchio metaphor and virtual reality external or internal. In rendering models autonomous, we are guided by three points: autonomy in essence, autonomy through necessity and autonomy due to ignorance. Autonomy in essence characterizes living organisms, from a cell to a human being. Avatars are not the only models which can perceive and act within their digital environments: any model which is supposed to represent a living being must be given this sort of sensory-motor interface. Animats, for instance, are artificial animals whose laws of functioning are inspired by those of animals [Wilson 85]. Like an avatar, an animat is located in an environment; it has sensors to acquire information and effectors enabling it to act within this environment. Unlike an avatar which must be controlled by a human user, an animat must control itself in order to coordinate its perceptions and actions [Meyer et Guillot 91]. This control can be innate, i.e. pre-programmed [Beer 90], but will more often be acquired in the animat approach, in order to simulate how behaviors adapted for survival in changing environments come about. Therefore, research in this very active field1 [Guillot et Meyer 00] mainly concerns the study of learning (epigenesis) [Barto et Sutton 81], development (ontogenesis) [Kodjabachian et Meyer 98] and evolution (phylogenesis) [Cliff et al. 93] of the control architecture. Animating the virtual creatures obtained using these different approaches provides a highly illustrative example of these adaptive behaviors [Sims 94]. Modeling virtual actors falls under the same approach [Thalmann 96]. Thus, by rendering a model associated with an organism autonomous, we can more faithfully account for the autonomy observed in the organism in question. Autonomy through necessity involves instantly taking changes in the environment into account by organisms or mechanisms alike. Physically modeling mechanisms usually depends on solving differential equation systems. Solving them requires knowledge about the boundary conditions which limit movement; and yet, in reality, these conditions can change constantly, whether or not the causes are known (interactions, disturbances, modifications in the environment). Therefore, the model must be capable of perceiving the changes in order to adapt its behavior while running. This is all the truer when humans are present in the system, since, though their avatars, they can provoke changes that were entirely unforeseeable at the outset. The example of sand in an hourglass can illustrate this point. The physical simulation of granular media is usually based on micromechanical interactions between spheres of varying hardness. To visualize flows of about one second takes several hours of computation, making these simulations unsuitable for virtual reality constraints [Herrmann et Luding 98]. Modeling using larger grains (mesoscopic level) based on specific 1 From Animals to Animats (Simulation of Adaptive Behavior: www.adaptive-behavior.org/conf): biennial conferences held since 1990.

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

7

masses which are linked to each other through the appropriate interactions, leads to satisfactory, but not interactive, visualizations [Luciani 00]. Another approach takes separate, large grains of sand, which individually detect collisions (elastic impacts) and are gravity-sensitive (free fall). It can simulate the flow of sand in the hourglass, but also adapt in real time to the hourglass being turned over, or a hole being created in it [Harrouet 00]. Thus, rendering any model autonomous will enable it to react to unforeseen situations appearing as it runs, and which are due to changes in the environment caused by the activity of other models. Autonomy due to ignorance reveals our present inability to describe the behavior of complex systems using the reductionistic methods of the analytical approach. A complex system is an open system made up of a heterogeneous group of atomic or composite entities. The behavior of the group is the result of the individual behavior of these entities and their various interactions within an environment which is also active. Depending on the school of thought, the group’s behavior is considered either as being organized with respect to an aim, called teleological behavior [Le Moigne 77], or as the result of the system’s self-organization, which would be called emergence [Morin 77]. The lack of models for the overall behavior of complex systems leads to control being distributed at the system component level and thus, making the components’ models autonomous. In this case, the simultaneous evolution of these components provides better understanding of the overall system’s behavior. Thus, a set of autonomous models interacting inside the same space contributes to the study and experimentation of complex systems. Making models autonomous, whether in essence, though necessity or from ignorance, helps populate virtual environments with an artificial life which reinforces the impression of reality. The following three modes are used to make entities autonomous: sensory-motor, decisional and operational modes. This is effectively based on sensory-motor autonomy, since each entity has sensors and effectors which enable it to be informed about and act on its environment. It also relies on decisional autonomy, since each entity decides according to its own personality (background, intentions, state and perceptions). Finally, it requires autonomy of performance: the controller of each entity’s operation is independent from the controllers of the other entities. In fact, the notion of autonomous entity overlaps that of an agent in the individual-centered approach of multi-agent systems. The first studies on multi-agent systems were carried out in the 1980’s. They were based on the asynchronous aspect of actors’ language interactions in distributed artificial intelligence [Hewitt 77], and on the individual-centered approach of artificial life [Langton 86], as well as on the autonomy of mobile robots [Brooks 86]. Two major trends are currently structuring this field of study, i.e., either attention is focused on agents as such (intelligent, rational or cognitive agents [Wooldridge et al. 95]), or the interactions between agents and collective aspects take precedence (reactive agents and multi-agent systems [Ferber 95]). This means we can find agents that reason on the basis of beliefs, desires or intentions, called BDI, [Georgeff et Lansky 87]), more emotional agents as in some video games (Creatures [Grand et Cli 98]), or purely reactive, stimulus/response type agents as in insect societies (MANTA [Drogoul 93]). In any case, these systems differ from symbolic planning models of classic Artificial Intelligence (STRIPS [Fikes et Nilsson 71]), since they accept that a sophisticated behavior can emerge from interactions between more reactive agents located in an active environment [Brooks 91].

3

An epistemological perspective Created from two words with opposing meanings, the expression virtual realities is absurd. If someone were to talk to you about the color black-white, he would seem, combining one word with its opposite, to be confused about what he wanted to say. Of course, in the strictest sense, virtual and real are not opposites. The virtual, from the Latin virtus (virtue or force), is latent in reality, which contains all the necessary prerequisites for it to be achieved. But then, what could a reality containing all conditions for its achievement in itself possibly be? From that point of view, the expression is even more inept. Cadoz C., Les r´ ealit´ es virtuelles, [Cadoz 94]

The emergence of the notion of virtual reality illustrates how dynamic interdisciplinary exchanges are between computer graphics, computer-aided design, simulation, teleoperations, audiovisual techniques and telecommunications. . . [Tisseau et N´ed´elec 03]. But as the philosopher Gaston Bachelard stressed in

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

8

his epistemological study on the formation of the scientific mind [Bachelard 38], scientific or technological progress must confront a number of epistemological obstacles. Amongst them, virtual reality will have to overcome at least the verbal obstacle (a false explanation obtained using an explanatory word), since its name itself is meaningless at the outset, while referring to the intuitive notion of reality, one of the earliest and constituent notions of the human mind. The English expression virtual reality was first suggested in July 1989 at a trade show2 , by Jaron Lanier who was the head of the VPL Research firm, specialized in immersion peripheral devices at the time. He coined the term within his company’s advertising and marketing strategy, without trying to provide a specific definition for it. According to the BBC English Dictionary (HaperCollins Publishers, 1992), Virtual: means that something is so nearly true that for most purposes it can be regarded as true, also means that something has all the effects and consequences of a particular thing, but is not officially recognized as being that thing. Therefore, a virtual reality is a quasi-reality, which looks and behaves like a reality, but is not: it is more like a substitute or ersatz reality. The direct translation of this English term into French gives us “r´ealit´e virtuelle”, which as emphasized in the quotation at the beginning of this introduction, is an absurd and inept expression. Indeed, the French dictionary Le Petit Robert (Editions Le Robert, 1992), defines Virtuel: as having within itself the necessary prerequisites for its achievement. So virtual reality would be a reality containing all the necessary conditions for its achievement, which is the least we’d expect from reality! We see that in going from English to French, the term of virtual reality has become ambiguous. There is a rhetorical processed called an oxymoron, which consists in putting together two words which seem incompatible or contradictory (expressions like living dead, clair obscur, or an eloquent silence are examples of this). This type of construction gives unexpectedness to an expression, which, we must agree, is more mediaoriented than scientific. Other expressions like cyber-space [Gibson 84], artificial reality [Krueger 83], virtual environment [Ellis 91], or virtual world [Holloway 92], have also been put forward, but a rapid web search indicates that the antonym virtual reality remains in widespread use. Another point of view considers that reality is what exists in itself, independently of whether we can perceive it or see it or not (Dictionnaire historique de la langue fran¸caise, Robert, 2000). This means that reality is a representation of what is real, i.e., a model - and virtual reality would therefore be a virtual representation, which is the least we’d expect from a model! The term virtual reality then becomes a pleonasm [Mellet 04]. Between oxymoron and pleonasm, these ambiguities blur distinctions and create a true epistemological stumbling block for the scientific development of this new field of study. It is up to scientists and professionals in the field in question to make the effort of epistemological clarification, in order to remove ambiguities and clarify its status as a scientific discipline (concepts, models, tools), especially with respect to closely related fields like modelling, simulation and animation. So a simulation typology must take the new way of experimenting on models, which is in virtuo experimentation (Section 3.1) into account, as well as highlighting complementarities between the different ways of modeling the same phenomenon (Section 3.2).

3.1

Types of simulations Beyond the physical experiment lies another, which is abundantly practiced at a higher level - the thought experiment. The project inventor, the building of castles in the air, the novelist, the author of social or technical utopia all experiment in their thoughts. But so do the down-to-earth merchant, the inventor or the serious researcher. They all think about the circumstances, and link a prudent approach and expectation to their perception; ie, they run a thought experiment. [...] Our representations are more easily and conveniently to hand than are physical facts. We experiment with thought, so to speak, at a lower cost. And so we mustn’t be surprised if the thought experiment often precedes and prepares a physical experiment. Mach E., Erkenntniss und Irrtum, [Mach 1905]

The main qualities of a model, i.e. an artificial representation of an object or a phenomenon, are based on its abilities to describe, suggest, explain, predict and simulate. Simulating the model, or experimenting on it, consists in testing the behavior of the representation under the effect of actions which can be exerted on the model. The simulation outcomes then become hypotheses that we try to verify by designing experiments on a single prototype of the real system. Thus rationalized, these are truly in vivo experiments. Traditionally, it is considered that there are four main kinds of models, i.e., perceptive, formal, analogical and digital models. Their experimentation has currently led to five main families of simulation: 2 Texpo’89

in San Francisco (USA)

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

9

in petto intuition resulting from perception models, in abstracto reasoning within formal models, in vitro experiments on analogical models, in silico computations on digital models, and now in virtuo experiments on these digital models. In petto intuitions Simulation of a perceptive model correspond to in petto intuitions that spring from our imagination and the sensory perception that we have of the system being studied. This enables perceptions to be tested on the real system. Inspirations and associations of ideas and heuristics, neither codified nor reasoned, lead to the forming of mental images with a power of suggestion. The scientific approach will try to rationalize these first impressions, whereas artistic creation will draw digital or analogical works of art from them. But it is often the suggestive nature of the perceptive model which triggers creative moments which lead to an invention or discovery [Vidal 84], as stated by Alfred Wegener, the father of continental drift and seafloor spreading which gave rise to the plate tectonics theory in the late 1960s. The first idea of continental translation symmetry came to mind in 1910. While looking at a map of the Earth, I was suddenly struck by the similarity of the coasts of the Atlantic, but I didn’t dwell on it at all at first, because I thought such translations were implausible. Alfred Wegener, La gen` ese des continents et des oc´ eans, 1937

In abstracto reasoning The simulation of a formal model relies on in abstracto reasoning conducted within the framework of a theory. Reasoning supplies predictions which can be tested on the real system Galle’s discovery of the planet Neptune in 1846 using Adams and Le Verrier’s theoretical predictions illustrates this approach within the framework of the gravitational perturbation theory with the two-body approximation in celestial mechanics. Likewise, in particle physics, the discovery of the intermediate vector bosons W+, W? et Z0 in 1983 had been predicted a few years earlier by the theory of electroweak interactions. Hence, from the infinitely large to the infinitely small, the predictive nature of formal models has proved to be very fruitful in a large number of scientific fields. In vitro experiments Simulation of an analogical model relies on in vitro experimentation on a sample or mock-up which is built by analogy with the real system. Similarities between the mock-up and the system improve the understanding of the system being studied. Aircraft mock-up trials in wind tunnels enable aerodynamicists to better characterize the flow of air around obstacles, by studying the similarity coefficients that Reynolds and Mach introduced at the end of the 19th century. Likewise, the analogy of the heart as a pump in physiology allowed Harvey to demonstrate that blood circulation followed the laws of hydraulics (1628). Thus, the explanatory nature of analogical models has always been used, more or less anthropocentrically to bring the unknown into the realm of the known. In silico computations Simulation of a digital model means running a program which is supposed to represent the system being modeled. In silico calculations provide results which are compared to measurements taken on the real system. Numerically solving mathematical equation systems is the most usual way of using digital modeling. Indeed, the analytical determination of solutions often comes up against difficulties which are due just as much to the characteristics of the equations to be solved (non-linearity, matching) as to the complexity of boundary conditions and the need to take very different space and time scales into account. Studying chemical reactions kinetics, calculating the deformation of a solid under thermo-mechanical constraints or characterizing an antenna’s electromagnetic radiation are classic examples of deploying differential equation systems on computers. Therefore, the digital model obtained by discretizing the theoretical model has become a vital tool to surpass theoretical limitations, but is still often considered as a stopgap solution. In virtuo experimentation More recently, the possibility of interacting with a program being run has opened a new path to in virtuo experimenting on digital models. It has become possible to perturb a model while it is running, dynamically modify the boundary conditions, and eliminate or add elements during the simulation. This gives digital models a virtual mock-up status, which is infinitely more malleable than the real mock-up used in analogical modeling. Flight simulators or video games are forerunners of virtual reality systems which become necessary when recourse to

10

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

direct experimentation becomes difficult or even impossible. This can be for various reasons: hostile environments, lack of accessibility, space-time, budget or ethical constraints. Thus, in virtuo experimenting now enables us to render certain of the thought experiments mentioned at the beginning of this section 3.1 in a tangible form, where until very recently, they were confined to imagination alone. In fact, these different simulation modes are complementary and all or part of them can be implemented to appraise and understand a given phenomenon.

3.2

Modeling and simulation “We only reason on the basis of model” (P. Val´ ery). But how do we develop the models we base our reasoning on? By models, we mean the intelligible artificial, symbolic representations of situations in which we intervene: modeling is both identifying and formulating a few problems by constructing their wording, and trying to solve them by reasoning through simulations. By making the model-wording work, we try to produce models-solutions. Modeling and simulation, thought and reasoning, are the two inseparable sides to any deliberation. Le Moigne J.L., La mod´ elisation des syt` emes complexes, [Le Moigne 90]

Whether we are literary or scientific minded, artists or engineers, study of an actual phenomenon begins with our sensorial information (Figure 3). These impressions, confronted with our personal imagination, inspire us with in petto intuitions which are expressed as perceptions. theories Formal models

tat

digitization

programs

m

predictions

p im

en lem

formalization

in abstracto reasoning ion

ria l

iza

tio

n

analogies

prototypes

Digital models

ate

models

Real phenomena

in virtuo experiments

in vivo experiments

dig

rea tio

n

in vitro experiments

similitudes

on

impressions

perceptions

behaviors

ita lc

Analogical models

ti rea lc

g alo an

ica

imaginary Perceptual models in petto intuitions

field

modeling

Model simulation

understanding

Figure 3: Phenomena modeling, simulation and understanding Only in the second phase will the scientific approach seek to formalize these initial perceptions to create a representation as free of individual illusions as possible. In abstracto reasoning developed in the framework of an appropriate theory and which is usually based on a logical-deductive approach, will thus lead to predictions about the phenomenon being studied. In vivo experiments on the real system can also be conducted to compare these predictions to experimental results. But in many real situations, the formal approach alone, which is reductionistic in essence, cannot account for the complexity of the phenomenon studied. Therefore, we can have recourse to analogies

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

11

to carry out experiments on real mock-ups. These mock-ups or models can be obtained by scale analogies (like scale models) or by formal analogies (thermal ? electric type). Results from these in vitro experiments will then be adapted to the real phenomenon by similarity (scale or conversion factors). Today, recourse to digital methods and computer programs opens up another path for simulating a formal model for which no analytical solution is available. Here we distinguish between the in silico computation and in virtuo experimentation by the absence or presence of a human in the simulation loop. In virtuo experiments allow the user to manipulate an actual virtual mock-up and to feel, or not, his first impressions, where in silico calculations would supply only numerical results. In some cases, we try to use simulation to explain not a real phenomenon, but an idea. To materialize his idea, an artist goes directly from the perceptive model of his imagination to an analogical or digital work of art: this is the act of artistic creation which will be analogical or digital depending on the medium used. Whereas the engineer will go from the formal model of his theoretical framework to a real or virtual mock-up, i.e., the act of technological design made tangible in analog or digital form. This model or mock-up, whether real or virtual, becomes a singular prototype for a new real phenomenon which can in turn be studied through new experiments. So the scientist enters an interactive process of modeling/simulation which enables him to clarify his idea and thus refine and improve the various associated models. The notion of a model as a representation of reality is based on two metaphors, one of them artistic and the other legal. The legal metaphor of delegation (the elected official represents the people, the papal nuncio represents the Pope, and the ambassador the Head of state) suggests the idea of replacement: the model substitutes for reality. The artistic metaphor of realization (a play is given in public, artistic inspiration is represented by a work) suggests the idea of presence: the model is a reality. Consequently, while complementing our now classic means of investigation, i.e., in vivo and in vitro experimentation, or in silico computations, experimenting on a digital model in virtuo ensures it a true presence and opens new fields of exploration, investigation and understanding of reality.

4

The stakes Better to make use of the new possibility of standing back a bit from the intellectual imperialism of science, not to deny its importance or its interest or attempt to overshadow it, which is impossible, but on the contrary, to try at last to look at it from a distance at assign it its appropriate place in the cultural landscape. It would be pathetic to deny the efficacy and scope of scientific knowledge, it would be absurd to refuse to use its instruments of thought. Even at that, we must decide what to do with it. L´ evy-Leblond J.M., Aux contraires, [L´ evy-leblond 96]

For scientists, describing, explaining, predicting and simulating the behaviors of complex systems, whether natural or artificial, was the major challenge of the 21st century. Virtual reality will help meet this challenge, whose related stakes are as scientific (Section 4.1), methodological (Section 4.2), and technological (Section 4.3) as they are societal (Section 4.4).

4.1

Scientific stakes Modern science is pursuing the desire for knowledge which already existing in the Neolithic. It is obviously deeper, more extensive, because we have other concepts at our disposal, but the desire to know our surroundings is the same. Previously, we thought that we Newton’s law and two or three others, we could understand it. Nature was perceived as being fundamentally simple, whereas today, we see that it is fundamentally complex. [...] I remember reading years ago a book by the American physicist Richard Feynman, called The Character of Physical Law. Feynman thought the world could be compared to a huge chess game, where complexity was only apparent and where each move was simple; once you knew the rules of the game, you could decipher the world. And yet, what is more complex than a proton? A proton is made up of quarks and these quarks interact via gluons and all sorts of things. Nature has ceased to be simple. Prigogine I., Entretien avec Ilya Prigogine, cit´ e dans [Benkirane 02]

Nature has ceased to be simple and the complex systems constituting it - open systems made up of numerous entities in interaction - are dynamic systems which are perpetually changing. To understand them, we need to model them. Modeling is the action of intentionally, through the composition of symbols, designing and building models which could render a seemingly complex phenomenon intelligible. It also amplifies the reasoning of the actor who plans to deliberately intervene within the phenomenon, i.e., the reasoning which aims to anticipate the consequences of possible action plans [Le Moigne 90].

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

12

Modeling complex systems raises numerous theoretical questions which must be answered. What modeling approach should be adopted? How can the modeling intention be explained in the model? How can a complex system and its multi-level organization be characterized? How should the notion of emerging overall behaviors based on individual behaviors be addressed? How can models with different space-time dynamics be integrated? How should the free will of the human taking part in these systems be taken into account? How can the singularity of a complex system be compared to the requirement of reproducibility of the scientific method? How should models be validated? What are the links between reality and virtuality? and so on. In virtual reality, the human is in the loop in multi-sensory interaction with multi-model and multidisciplinary systems. This raises 3 main questions: 1. What is the place and the role of humans in these virtual environments? 2. What is the place and the role of the virtual entities moving in these environments? 3. What devices or systems should be deployed to ensure quasi-natural interactions between the human who is immersed in the environment and the entities populating it? Answering the questions in this non exhaustive list will require defining a new methodological approach that is suited to studying complex systems.

4.2

Methodological stakes The more complex a science is, the more important it is, in fact, to establish a proper experimental critique, in order to obtain facts which are comparable and free of sources of error. [...] The complete scientist is one who masters both theory and practice. Although he must fully possess the art of establishing experimental facts, which are the materials of science, he must also clearly account for the scientific principles which direct our reasoning in the highly varied environment of experimental study of natural phenomena. It would be impossible to separate the two; head and hand. A skillful hand without the head to direct it is a blind instrument; the head without the hand to do things remains powerless. Bernard C., Introduction ` a l’´ etude de la m´ edecine exp´ erimentale, [Bernard 1865]

In order to establish a proper experimental critique, study and validation of complex system models require their simulation. In the framework of virtual reality, simulation becomes participatory and proposes in virtuo experimentation of digital models while they are running. In virtuo experimentation should be conducted from multi-model modeling requiring the cohabitation or integration of the various models, while allowing a multi-level (local/overall) analysis of the system being experimented on. It should enable a cognitivism/constructivism dialogue between formal approaches and experimental approaches in order to validate the models experimented. Virtual reality, which enables users to experience sensory-motor activities in artificial worlds, instruments Varela’s enactive approach where cognition is not only representation, but also embodied action: a system’s intelligibility is based as much on praxis in situation as on pure information processing [Varela et al. 93].

4.3

Technological stakes Every computer program is a model, hatched in the mind, of a real or mental process. These processes, arising from human experience and thought, are huge in number, intricate in detail, and at any time only partially understood. They are modeled to our permanent satisfaction rarely by our computer programs. Thus even though our programs are carefully handcrafted discrete collections of symbols, mosaics of interlocking functions, they continually evolve. We change them as our perception of the model becomes deeper, larger, more generalized, until the model ultimately attains a metastable place within still another model we are struggling with. The source of the exhilaration associated with computer programming is the continual unfolding within the mind and on the computer of mechanisms expressed as programs and the explosion of perception they generate. If art interprets our dreams, the computer executes them in the guise of programs! Abelson H., Sussman G.J., Sussman J., Structure and interpretation of computer programs, [Abelson et al. 85]

If art interprets our dreams, the computer executes them in the guise of programs and, in order to experiment on them in virtuo, they need instrumentation. The virtuoscope, which gives VR digital models their instruments, must of course have a large scale computing infrastructure available, a true virtual lab bench for the 21st century. Creating this virtual lab bench must ensure real-time, reliable and almost natural interoperability of systems to enable believable sensory-motor experiences on multi-model and multi-disciplinary systems.

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

13

That way, we will have new tools for collective activity in virtual environments with less risk and lower costs at our disposal. The virtuoscope’s users will then be able to extract themselves from their own spatial-temporal spaces to interact and participate within the same virtual universe in an artificial life simulating the reality or the imagination of the modeler.

4.4

Societal stakes Virtual worlds must be realized, in other words, one must endeavor to update what is virtually present in them, to know the intelligible models that structure them and the ideas the develop them. The ”fundamental virtue” of virtual worlds is to have been designed with an end in mind. It is this end that must be realized, actuated, whether the application is industrial, spatial, medical, artistic, playful or philosophical. Images of virtuality must help us reveal the reality of virtuality, which is of an intelligible order and of proportional intelligibility to the goal pursued, whether theoretical or practical, utilitarian or contemplative. Qu´ eau Ph., Le virtuel, vertus et vertiges, [Qu´ eau 93b]

The “fundamental virtue” of virtual worlds is to have been designed with an end in mind. Indeed, the understanding of complex systems and their experimentation instrumented within the virtuoscope should make allow complexity to be mastered in all its societal dimensions. • In technological terms (assisted design, virtual mock-up, simulation, etc.), in virtuo experimenting on virtual models or mock-ups takes place at each stage in the life of a new product - from disposable razor to nuclear power plant -: from the idea of a product from its eventual disassembly to its design, prototyping and its maintenance. • In cultural terms (virtual museum, virtual theater, interactive fiction, participatory arts, etc.), in virtuo experimentation places the spectator at the heart of the work of art, enabling him to become actor and creator (cre-actor). • In educational terms (virtual training environments, preparing missions in hostile environments, etc.) in virtuo experimentation returns learning of know-how and interpersonal skills to its proper place. • In health care terms (surgical operations, therapies, bio feedback, etc.) in virtuo experimentation lets patients play an active part, in full cooperation with the caregivers. • In political terms (decision-making aid for urban planning, emergency services, environmental protection, etc.), in virtuo experimentation provides deciders with the means to better assess the different scenarios being considered and test the different possibilities by acting in the virtual environment. In virtual reality, decision also becomes a simulation of action, as suggested by the neurophysiologist Alain Berthoz. Therefore, decision is not just reason, it is also action. It is never a purely intellectual process, a logic game that we can put in an equation. Decision involves reflection, of course, but it already carries within it, while including the elements of the past, the act which it will lead to. Berthoz A., La d´ ecision, [Berthoz 03]

These few examples, selected amongst others, clearly illustrate the importance of the stakes related to the understanding, study, experimentation, instrumentation and mastery of complex systems. In virtuo experimentation in virtual reality thus appears to be one of the best ways to grasp this complexity. There are two illusions which divert our minds from the issue of complex thought, and which must be dispelled. The first is the belief that complexity leads to the elimination of simplicity. Of course, wherever simplifying thought falters complexity does indeed appear. However, it incorporates all that can bring order, clearness, distinction and precision to knowledge. Whereas simplifying thought splits up the complexity of reality, complex thought includes simplifying ways of thinking, insofar as possible. But it rejects the mutilating, simplistic, onesizing, and in the final analysis, blinding consequences of simplification, simplification which takes itself for the reflection of what is real in reality. The second illusion involves confusing complexity with completeness. Indeed, the ambition of complex thought is to account for the linkages between subject fields which disjunctive thinking (itself one of the main aspects of simplifying thought) breaks down; this thinking isolates what it distinguishes, and obscures all that could link, interact or interfere. In this sense, complex thought aspires to multidimensional knowledge. Yet, from the outset, it recognizes the impossibility of complete knowledge: for one of complexity’s axioms is that omniscience is impossible, even in theory. Morin E., Introduction ` a la pens´ ee complexe, [Morin 90]

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

14

References [Abelson et al. 85] Abelson H., Sumangerssman G.J., Sussman J., Structure and interpretation of computer programs (1985), traduction fran¸caise : InterEditions, 1989 [Bachelard 38]

Bachelard G., La formation de l’esprit scientifique (1938), Vrin, Paris, 1972

[Bannon 91]

Bannon L.J., From human factors to human actors : the role of psychology and humancomputer interaction studies in system-design, dans [Greenbaum et al. 91]:25-44, 1991

[Barto et Sutton 81] Barto A.G., Sutton R.S., Landmark learning : an illustration of associative search, Biological Cybernetics 42:1-8, 1981 [Benford 95]

Benford S., Bowers J., Fahlen L.E., Greehalgh C., Mariani J., Rodden T., Networked virtual reality and cooperative work, Presence 4(4):364-386, 1995

[Beer 90]

Beer R.D., Intelligence as adaptive behavior : an experiment in computational neuroethology, Academic Press, San Diego, 1990

[Benkirane 02]

Benkirane R., La complexit´e, vertiges et promesses, Le Pommier, 2002

[Bernard 1865] Bernard C., Introduction ` a l’´etude de la m´edecine exp´erimentale (1865), GarnierFlammarion, 1966 [Berthoz 03]

Berthoz A., La d´ecision, Odile Jacob, 2003

[Brooks 86]

Brooks R.A., A robust layered control system for a mobile robot, IEEE Journal of Robotics and Automation 2(1):14-23, 1986

[Brooks 91]

Brooks R.A., Intelligence without representation, Artificial Intelligence 47:139-159, 1991

[Bryson 96]

Bryson S., Virtual reality in scientific visualization, Communications of the ACM 39(5):62-71, 1996

[Cadoz 94]

Cadoz C., Les r´ealit´es virtuelles, Collection DOMINOS, Flammarion, Paris, 1994

[Capin et al. 97] Capin T.K., Pandzic I.S., Noser H., Magnenat-Thalmann N., Thalmann D., Virtual human representation and communication in VLNET networked virtual environments, IEEE Computer Graphics and Applications 17(2):42-53, 1997 [Carlsson et Hagsand 93] Carlsson C., Hagsand O., DIVE – A platform for multi-user virtual environments, Computers and Graphics 17(6):663-669, 1993 [Cliff et al. 93]

Cliff D., Harvey I., Husbands P., Explorations in evolutionary robotics, Adaptive Behavior 2(1):73-110, 1993

[Collodi 1883]

Collodi C., Les aventures de Pinocchio (1883), Editions Mille et une Nuits, Paris, 1997

[De Rosnay 75] De Rosnay J., Le macroscope : vers une vision globale, Editions du Seuil, Paris, 1975 [Drogoul 93]

Drogoul A., De la simulation multi-agents ` a la r´esolution collective de probl`emes. Une ´etude de l’´emergence de structures d’organisation dans les syst`emes multi-agents, Th`ese de Doctorat, Universit´e Paris 6 (France), 1993

[Ellis 91]

Ellis S.R., Nature and origin of virtual environments : a bibliographic essay, Computing Systems in Engineering 2(4):321-347, 1991

[Ferber 95]

Ferber J., Les syst`emes multi-agents : vers une intelligence collective, InterEditions, Paris, 1995

[Fikes et Nilsson 71] Fikes R.E., Nilsson N., STRIPS : a new approach to the application of theorem proving to problem solving, Artificial Intelligence 5(2):189-208, 1971

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

15

[Flach et Holden 98] Flach J.M., Holden J.G., The reality of experience : Gibson’s way, Presence 7(1):9095, 1998 [Fuchs 96]

Fuchs P., Les interfaces de la r´ealit´e virtuelle, Les presses de l’Ecole des Mines, Paris, 1996

[Fuchs et al. 03] Fuchs Ph., Arnaldi B., Tisseau J., La r´ealit´e virtuelle et ses applications, dans Le trait´e de la r´ealit´e virtuelle, 2e`me ´edition, volume 1, chapitre 1, pages 3-52, Les Presses de l’Ecole des Mines de Paris, 2003. [Georgeff et Lansky 87] Georgeff M.P., Lansky A.L., Reactive reasonning and planning, Proceedings AAAI’87: 677-682, 1987 [Gibson 84]

Gibson W., Neuromancer, Ace Books, New York, 1984

[G¨obel et Neugebauer 93] G¨obel M., Neugebauer J., The virtual reality demonstration centre, Computer and Graphics 17(6):627-631, 1993 [Guillot et Meyer 00] Guillot A., Meyer J.A., From SAB94 to SAB2000 : What’s New, Animat?, Proceedings From Animals to Animats’00 6:1-10, 2000 [Grand et Cli 98] Grand S., Cli D., Creatures : entertainment software agents with articial life, Autonomous Agents and Multi-Agent Systems 1(1):39-57, 1998 [Greenbaum et al. 91] Greenbaum J., Kyng M. (´editeurs), Design at work : cooperative design of computer systems, Lawrence Erlbaum Associates, Hillsdale, 1991 [Hallyn 04]

Hallyn F., Les structures rh´etoriques de la science, Seuil, 2004

[Harrouet 00]

Harrouet F., oRis : s’immerger par le langage pour le prototypage d’univers virtuels ` a base d’entit´es autonomes, Th`ese de Doctorat, Universit´e de Bretagne Occidentale, Brest (France), 2000

[Herrmann et Luding 98] Herrmann H.J., Luding S., Review article : Modeling granular media on the computer, Continuum Mechanics and Thermodynamics 10(4):189-231, 1998 [Hewitt 77]

Hewitt C., Viewing control structures as patterns of message passing, Artificial Intelligence 8(3):323-364, 1977

[Holloway 92]

Holloway R., Fuchs H., Robinett W., Virtual-worlds research at the University of North Carolina at Chapel Hill, Computer Graphics’92, 15:1-10, 1992

[Johnsen et Corliss 71] Johnsen E.G., Corliss W.R., Human factors applications in teleoperator design and operation, Wiley, New York, 1971 [Kallmann et Thalmann 99] Kallmann M., Thalmann D., A behavioral interface to simulate agent-object interactions in real time, Proceedings Computer Animation’99:138-146, 1999 [Kodjabachian et Meyer 98] Kodjabachian J.,Meyer J.A., Evolution and development of neural controllers for locomotion, gradient-following, and obstacle-avoidance in artificial insects, IEEE Transactions on Neural Networks 9:796-812, 1998 [Krueger 83]

Krueger M.W., Artificial Reality, Addison-Wesley, New York, 1983

[Langton 86]

Langton C.G., Studying artificial life with cellular automata, Physica D22:120-149, 1986

[Le Moigne 77] Le Moigne J-L., La th´eorie du syst`eme g´en´eral : th´eorie de la mod´elisation, Presses Universitaires de France, Paris, 1977 [Le Moigne 90] Le Moigne J-L., La mod´elisation des syst`emes complexes, Collection AFCET Syst`emes, Bordas, Paris, 1990

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

16

[L´evy-leblond 96] L´evy-Leblond J.M., Aux contraires. L’exercice de la pens´ee et la pratique de la science, NRF essais, Gallimard, 1996 [Luciani 00]

Luciani A., From granular avalanches to fluid turbulences through oozing pastes : a mesoscopic physically-based particle model, Proceedings Graphicon’00 10:282-289, 2000

[Mach 1905]

Mach E., Erkenntniss und Irrtum, (1905), Wissenschaftliche Buchegesellschaft, 1976, cit´e dans [Hallyn 04]

[Mellet 04]

Mellet d’Huart D., De l’intention ` a l’attention. Contributions ` a une d´emarche de conception d’environnements virtuels pour apprendre ` a partir d’un mod`ele de l’(´en)action., Th`ese de Doctorat, Universit´e du Maine, Le Mans (France), 2004

[Meyer et Guillot 91] Meyer J.A., Guillot A., Simulation of adaptive behavior in animats : review and prospect, Proceedings From Animals to Animats’91 1:2-14, 1991 [Minsky 80]

Minsky M., Telepresence, Omni 6:45-51, 1980

[Morin 77]

Morin E., La m´ethode, Tome 1 : la nature de la nature, Editions du Seuil, Paris, 1977

[Morin 90]

Morin E., Introduction ` a la pens´ee complexe, ESF ´editeur, Paris, 1990

[Morineau 00]

Morineau T., Context effect on problem solving during a first immersion in a virtual environment, Current Psychology of Cognition 19:533-555, 2000

[Noser et Thalmann 95] Noser H., Thalmann D., Synthetic vision and audition for digital actors, Proceedings Eurographics’95:325-336, 1995. [Platon -375]

Platon, La r´epublique, Livre VII (375 avant J.C.), Traduction de E. Chambry, Biblioth`eque M´ediations, Editions Gonthier, Paris, 1969

[Qu´eau 93a]

Qu´eau Ph., Televirtuality : the merging of telecommunications and virtual reality, Computers and Graphics 17(6):691-693, 1993

[Qu´eau 93b]

Qu´eau Ph., Le virtuel, vertus et vertiges, Collection Milieux, Champ vallon, Seyssel, 1993

[Renault 90]

Renault O., Magnenat-Thalmann N., Thalmann D., A vision-based approach to behavioural animation, Journal of Visualization and Computer Animation 1(1):18-21, 1990

[Schloerb 95]

Schloerb D.W., A quantitative measure of telepresence, Presence 4(1):64-80, 1995

[Schuler et al. 93] Schuler D., Namioka A. (editors), Participatory Design : Principles and Practices, Lawrence Erlbaum Associates, Hillsdale, 1993 [Sheridan 87]

Sheridan T.B., Teleoperation, telepresence, and telerobotics : research needs, Proceedings Human Factors in Automated and Robotic Space Systems’87:279-291, 1987

[Sims 94]

Sims K., Evolving 3D morphology and behavior by competition, Artificial Life 4:28-39, 1994

[Singh 94]

Singh G., Serra L., Ping W., Hern N., BrickNet : a software toolkit for network-based virtual worlds, Presence 3(1):19-34, 1994

[Slater et al. 94] Slater M., Usoh M., Steed A., Depth of presence in virtual environments, Presence 3(2):130-144, 1994 [Tachi et al. 89] Tachi S., Arai H., Maeda T., Robotic tele-existence, Proceedings NASA Space Telerobotics’89 3:171-180, 1989 [Thalmann 96]

Thalmann D., A new generation of synthetic actors : the interactive perceptive actors, Proceedings Pacific Graphics’96:200-219, 1996

Tisseau J., In vivo, in vitro, in silico, in virtuo, 1st Workshop on SMA in Biology at meso or macroscopic scales, Paris, july 2, 2008.

17

[Tisseau et al. 98b] Tisseau J., Reignier P., Harrouet F., Exploitation de mod`eles et R´ealit´e Virtuelle, Actes GTAS’98:13-22, 1998 [Tisseau 01]

Tisseau J., R´ealit´e virtuelle : autonomie in virtuo, Habilitation `a Diriger des Recherches, Universit´e de Rennes I (France), 6 d´ecembre, 2001

[Tisseau et N´ed´elec 03] Tisseau J., N´ed´elec A., R´ealit´e virtuelle : un contexte historique interdisciplinaire, Revue internationale de CFAO et d’infographie 17(3-4):263-278, 2003. [Varela et al. 93] Varela F., Thomson E., Rosch E., L’inscription corporelle de l’esprit, Seuil, 1993 [Vertut et Coiffet 85] Vertut J., Coiffet P., Les robots, Tome 3 : t´el´eop´eration, Herm`es, Paris, 1985 [Vidal 84]

Vidal F., L’instant cr´eatif, Flammarion, Paris, 1984

[Wilson 85]

Wilson S.W., Knowledge growth in an artificial animal, Proceedings Genetic Algorithms and their Applications’85:16-23, 1985

[Witmer et Singer 98] Witmer B.G., Singer M.J., Measuring presence in virtual environments : a presence questionnaire, Presence 7(3):225-240, 1998 [Wooldridge et al. 95] Wooldridge M., Jennings N.R., Intelligent agents : theory and practice, The Knowledge Engineering Review 10(2):115-152, 1995 [Zahorik et Jenison 98] Zahorik P., Jenison R.L., Presence as being-in-the-world, Presence 7(1):78-89, 1998 [Zwirn 03]

Zwirn H., La complexit´e, science du XXIe`me si`ecle ?, dans Pour La Science, num´ero sp´ecial ”La complexit´e”, d´ecembre 2003