Some Ethical Aspects of Agency Machines Based on Artificial Intelligence. By Francesco Amigoni, Viola Schiaffonati, Marco Somalvico

Transcription

1 Some Ethical Aspects of Agency Machines Based on Artificial Intelligence By Francesco Amigoni, Viola Schiaffonati, Marco Somalvico Politecnico di Milano - Artificial Intelligence and Robotics Project Abstract The adoption of increasingly powerful and flexible machines developed within artificial intelligence, like agencies, envisages a novel anthropological framework of man and his relation to machines. The paper aims to present some innovative ethical observations that are stimulated by this new anthropological framework. 1 Introduction The growing impact of computer science on human society stimulates several ethical considerations that are enriched and enhanced when we refer to powerful and flexible machines, called agencies, developed within artificial intelligence. An agency is a machine composed of several cooperating inferential entities called agents, which are computer or robots [Amigoni et al., 1999c]. The purpose of this paper is to present some innovative ethical observations that arise from the adoption of agencies in the conceptual framework that conceives man and his relation to machine, in general, and to agency, in particular. More precisely, after a short introduction to the concept of agency (Section 2) and to the epistemological framework adopted (Section 3), we discuss the ethical aspects of universality (Section 4), equality (Section 5), solidarity (Section 6), and freedom (Section 7). The four ethical concepts are summarized and symbolically represented by the anthroparadigmatic tetrahedron illustrated at the end of the paper (Section 8). In this paper, we use the word man with the meaning of human being in order to preserve the terminology of our previous papers.

2 2 Agency Machine in Artificial Intelligence The preliminary technical description of agency machine is necessary for the appropriate understanding of the ethical consequences that are implicated by this powerful and flexible machine of artificial intelligence. Marvin Minsky [Minsky, 1985] has firstly introduced the idea of agency in order to provide a description of the complex nature and performance of the human mind. According to Minsky, several competing and cooperating models, called paradigms, each one embedded in an agent, are adequate to describe the complexity of human mind, which can not be fully expressed by a single model. The set of interrelated agents constitutes the society of mind, which is also called agency. The idea of agency has been further developed within the field of distributed artificial intelligence, where an agency is considered as a particular multiagent system in which each agent is a computer or a robot with inferential abilities [Amigoni et al., 1999c]. The most important problem to be solved in order to design an agency is the so-called Babel Tower Aporia. It consists in the difficulty encountered in integrating specialized and different agents in a uniform and general cooperation framework. A plausible solution to the Babel Tower Aporia is to structure the architecture of each agent as a couple of semiagents; the first one, called op semiagent, devoted to operate, the second one, called co semiagent, devoted to cooperate. According to this particular architecture, an agency can be constructed by starting from existing heterogeneous agents, considered as mere op semiagents, and by successively integrating them with homogeneous co semiagents. Among the various ways to design and integrate the co semiagent in each op semiagent, we have individuated a novel and particularly advantageous technique that exploits the modern technology of mobile code systems. A unique mobile code system (called Mobile Intelligent Agent or MIA) triggers an evolutionary process, that eventually provides the uniform co semiagents, which together with op semiagents (called, by contrast, Fixed Intelligent Agents or FIAs), constitute the whole agent. MIA travels in the communication network connecting the op semiagents and, when it reaches a FIA, firstly replicates itself and secondly enables each replica to evolve into a common and uniform co semiagent. In this way, an agency is automatically built by starting from heterogeneous FIAs inserted as nodes in a communication framework and by spreading uniformity as the result of travel, replication and evolution of the unique

3 MIA. This novel kind of agency, developed following the described methodology, is called dynamic agency [Amigoni and Somalvico, 1998]. The advantages provided by dynamic agency approach are summarized in the flexibility of the process that results in an agency tuned toward a given goal. Firstly, in a dynamic agency it is possible to address different applications by sending different MIAs, which are related to the applications. In this way each MIA evolves in the most appropriate co semiagents for the corresponding application. Secondly, by means of the travelling MIA, it is conceivable the setting up of an automatic selection (called recruitment) of the op semiagents and, thus, of an automatic composition of the dynamic agency. Finally, flexibility is embedded in the possibility of modifying the composition of a dynamic agency at run-time, namely during the process that solves an application problem. 3 A New Anthropological Outlook of Man In this section we introduce the epistemological framework in which the following considerations are embedded, in order to present a new anthropological outlook of man and his relation to machines. We adopt the empirical inductive deductive paradigm [Amigoni et al., 1999a], inspired by Galileo Galilei works [Galileo Galilei, 1967], which represents a description of scientific method. This paradigm provides an illustration of one of the processes that bring man to know reality, intended as the source of phenomena that humans can perceive and describe by models. The key point of the empirical inductive deductive paradigm consists in the differences between reality (in which phenomena occur) and knowledge of reality (to which models belong). Two different kinds of intelligence act within this paradigm. We call them, according to Henri Bergson s approach [Bergson, 1984], creative intelligence and fabricative intelligence. Creative intelligence includes irrational intellectual activities that can not be modelled, namely that can not be described by models, whereas fabricative intelligence includes rational intellectual activities that can be modelled. Creative intelligence is involved in every situation in which man performs the passage from reality to knowledge of reality and vice versa. Fabricative intelligence, instead, is exclusively performed within knowledge of reality. In this scenario, each machine can be considered as a kind of model of a phenomenon. Since any model is made up by shaping a form within a formalism, in the case of machine the form of the model is the architecture of the machine, whereas the formalism of the model is the set of

4 components (intended as submodels) of the machine. Therefore, a machine is an artificial entity that is a reification of a model of a phenomenon, namely a machine is an artifact that plays the role of operationally describing a model. It is interesting to note that the performance of the unique composite machine is indeed the emulation (the partial reproduction) of the phenomenon whose model has been reified by the machine. In our proposal, machines can exclusively emulate the fabricative intelligence processes in which only rational activities are carried on. In this scenario we propose a new anthropological outlook of man, arising from the novel relation he establishes with the particular class of machines (e.g., computers and robots) that are able to emulate some of his intellectual activities. We consider man as a unique man-mind subject, who performs his intellectual activities in two different sites denoted as poles. The first pole is called man-body pole and represents the natural site of man s body, where both creative and fabricative intellectual activities are performed. The second pole is called man-machine pole and represents the artificial site of machine, where only fabricative intellectual activities are performed. 4 Universality An initial ethical concept is related to the generality of culture that is envisaged by the advent of agencies as man-machine poles. The splitting of the intellectual activities of man and their allocation in two poles represent the bipolar condition of man-mind subject, who can perform his intellectual activities immediately, namely without any medium, in manbody pole or mediately, namely by means of a medium, in man-machine pole. According to the bipolar condition, it is clear that man-machine pole is an expression of human culture. In fact, every machine reflects human culture, since a machine is the result of the whole history of science and technology and of the efforts of contemporary men who work together in order to develop better and better artifacts. Moreover, when the machine is an agency, we can say that it represents human culture, because it roughly emulates a society of men, each one roughly emulated by an agent. Hence, human culture permeates every man, since man-machine pole is the result of human culture and, in the case of agency, it also represents human culture. We call universality the property that expresses the pervasive dissemination of culture within the man-mind subject, not only abstractly within man-body pole, but also concretely within man-machine pole.

5 The passages, in the design and construction of machines, from macrotechnology to microtechnology and from it to nanotechnology further promote the property of universality in a still more concrete sense. In fact, these passages can be considered from four different perspectives. Metric perspective, when we note that the physical dimensions of machines are getting smaller and smaller. Anthropometric perspective, when we note that machines are getting smaller and smaller than human body. Anthropic perspective, when we note that machines are going from emulation of phenomena outside human body to emulation of phenomena inside human body. Anthrotopic perspective, when we note that machines are going to be situated from outside human body to inside human body. Hence, universality of culture is enhanced by inserting man-machine pole, which reflects human culture, inside human body as a concrete entity placed in a specific place. 5 Equality Another ethical concept, called equality, can be introduced by observing the double role that an agent can play due to the particular nature of agency. A single agent can describe both the product and the producer of culture and, in particular, of scientific knowledge. In fact, as we have seen, agency represents the culture of several men (each one associated with an agent) when it describes, as a flexible and renewable machine, the product of human culture. In this case, it is called Product Dynamic Agency (PDA). Moreover, agency enhances each single bipolar man-mind subject, when he acts as producer of culture. In fact, agency, as man-machine pole, represents the practical support for man in performing fabricative intellectual activities and, in this case, it is called Site Dynamic Agency (SDA). The ability of agency to describe both the product and the producer of culture [Amigoni et al., 1999b] is called duality function and envisages the ethical concept of equality. In fact, it is equal to use an agent to describe the product of culture or the producer of culture. The generalization of duality function brings to consider each agent equally usable for addressing different pragmatic goals. This means that equality corresponds to the egalitarian utilizability of agents. In fact, according to the idea of agency, there are many alternative systems to address a given problem, and there is not the winner or the most adequate system for every situation. Since each agent can be equally utilized for satisfying different pragmatics, a situation of cooperation by competition is

6 promoted among heterogeneous but equivalent agents, and hierarchies that are not based on pragmatic considerations are excluded. 6 Solidarity A still further ethical concept here presented involves solidarity that is centered on the circular nature of culture. We call circularity property [Amigoni et al., 1999b] the synergic integration between the agency that represents the product of culture, namely PDA, and the agency that represents the producer of culture, namely SDA. This property is encountered when a strongly flexible agency, like dynamic agency (see Section 2), is adopted as man-machine pole. Circularity property involves the grafting of the agents of newly created PDA within SDA in order to reinforce and enhance the site that is, in this way, ready for beginning a new improved process of knowledge production. The adoption of dynamic agency according to the circular nature of culture envisages and stimulates the ethical concept of solidarity. From a general point of view, in fact, circularity property reflects a scenario in which the results of a service contribute to compose the future server. In particular, circularity property clearly expresses the synergic and solidaristic interconnection between agents (each one roughly emulating a man) that play, in sequence, the roles of service and server. 7 Freedom A last ethical concept, called freedom, is envisaged by the adoption of dynamic agencies. Freedom is related to the bivalent nature of man who, when he interacts with a dynamic agency, can be considered from two different perspectives. A single man can be seen both as the user of dynamic agency and as the designer of the dynamic agency he uses. This is related to the particular nature of dynamic agency that can be automatically built starting from a formalized initial exigency expressed by the user [Amigoni and Villa, 1999]. This leads to identify the user of a dynamic agency with its designer. It is important to note that the correspondence between the user and the designer, encountered adopting a dynamic agency, involves not only the selection of the best (according to the problem to be solved) software system, but also the selection of the best hardware. In a concrete sense, this means that the user becomes the designer because he implicitly selects the best available agents in order to form a dynamic agency able to satisfy his exigency.

7 The coincidence between user role and designer role in a dynamic agency promotes the concept of flexibility that, from an ethical perspective, entails freedom. In fact, by adopting dynamic agencies, man is free to design, develop, build and use machines that address the solutions of his needs. Moreover this ethical quality stands up within a machine, a very complex one, like dynamic agency, where complexity is not against simplicity. More precisely, one of the performances of such complex entity is the performance of ease of design that is offered to the end-user, who is different from a professional designer. Such outstanding and surprising quality symbolically reaffirms the greatest ethical quality of freedom as the prime motor in the development and fulfillment of the human society. 8 Conclusions In this paper we have presented some innovative ethical observations that arise from the adoption of agencies, a novel and powerful class of machines, in the conceptual framework that conceives man as a bipolar subject. Within this framework, the bipolar condition of man is a consequence of the dualistic approach we adopt (see Section 3). In particular, we have considered four ethical concepts - universality, equality, solidarity, and freedom - and we have proposed a correspondence between them and four human dichotomies, which describe a novel anthropological outlook of man and his relation to machine. In particular, we have associated universality to bipolar condition that illustrates the role of man and machine in human intellectual activities. We have associated equality to duality function that illustrates the role of agency both as site of knowledge production process and as representation of the results of knowledge production process. We have associated solidarity to circularity property that reflects the synergic integration and concatenation of dynamic agencies that act as site and as description of results of knowledge production process. Finally, we have associated freedom to the bivalent nature of man that describes the double role of man both as user and as designer of a dynamic agency. In order to summarize our conceptual perspective, in which machines, human dichotomies, and ethical aspects are integrated, we present a scheme called anthroparadigmatic tetrahedron [Adorno, 1999] (see Figure 1). Dynamic agency Bivalent nature Freedom Dynamic agency Circularity property Solidarity Machine Bipolar condition Universality Agency Duality function Equality

8 Figure 1 - The anthroparadigmatic tetrahedron. It illustrates the separation between universality, equality, solidarity, from one side, and freedom, from the other side. In fact, we say that bipolar condition is related to space, since it involves two different sites that are distributed in space. Duality function is related to time, since it involves different roles at different times for a given agent. Circularity property is related to space and time, since it involves integration and concatenation of SDA and PDA that are distributed in space and act as distributed in time. On the other hand, the bivalent nature of man is situated on a higher level and includes space and time, since it unifies them in a more abstract knowledge process which goes beyond space and time. Moreover, the anthroparadigmatic tetrahedron illustrates the relationships among the four dichotomies previously described. In fact, the plane, where the base of tetrahedron stands, expresses the three dichotomies between man and machine. On the other hand, the third dimension, identified by the height of tetrahedron, expresses the dichotomy between man who is (user) and man who knows (designer). Acknowledgements The authors are glad to thank prof. Francesco Adorno for his philosophical and philological contribute to this work. References Adorno, F. (1999), Personal communication. Amigoni, F., Schiaffonati, V., and Somalvico, M. (1999a), Processing and interaction in robotics, Sensors and Actuators A: Physical, 72(1), Amigoni, F., Schiaffonati, V., and Somalvico, M. (1999b), Dynamic agencies and creative scientific discovery, Proceedings of the AISB 99 (the society for the study of Artificial Intelligence and Simulation of Behaviour) Symposium on

9 Artificial Intelligence and Scientific Creativity, Edinburgh, UK, April , Amigoni, F. and Somalvico, M. (1998), Dynamic agencies and multi-robot systems, in Lueth, T., Dillmann, R., Dario, P., and Worn, H. (eds), Distributed Autonomous Robotic Systems 3, Springer-Verlag. Amigoni, F., Somalvico, M., and Zanisi, D. (1999c), A theoretical framework for the conception of agency, International Journal of Intelligent Systems, 14(5), Amigoni, F. and Villa, M. (1999), An algebraic description of agency design, Eleventh European Summer School in Logic, Language and Information (ESSLLI99) Workshop on Foundations and Applications of Collective Agent Based Systems (CABS), Utrecht, The Netherlands, August Bergson. H. (1984), Creative evolution, University Press of America. Galileo Galilei (1967), Dialogue concerning the two chief world systems: Ptolemaic and Copernican, Stillmann Drake. Minsky, M. (1985), The society of mind, Simon & Schuster.