-
Chapter 1: Artificial Agents and Agency
- University of Michigan Press
- Chapter
- Additional Information
Chapter 1 / Arti‹cial Agents and Agency In developing a legal theory appropriate for arti‹cial agents, the ‹rst tasks are to identify, and then clarify the nature of, the subject of our theorizing. There are two views of the goals of arti‹cial intelligence. From an engineering perspective, as Marvin Minsky noted, it is the “science of making machines do things that would require intelligence if done by men” (1969, v). From a cognitive science perspective, it is to design and build systems that work the way the human mind does (Shanahan 1997, xix). In the former perspective, arti‹cial intelligence is deemed successful along a performative dimension; in the latter, along a theoretical one. The latter embodies Giambattista Vico’s perspective of verum et factum convertuntur , “the true and the made are . . . convertible” (Vico 2000); in such a view, arti‹cial intelligence would be reckoned the laboratory that validates our best science of the human mind. This perspective sometimes shades into the claim arti‹cial intelligence’s success lies in the replication of human capacities such as emotions, the sensations of taste, and selfconsciousness . Here, arti‹cial intelligence is conceived of as building arti‹cial persons, not just designing systems that are “intelligent.” While such a goal may capture the popular imagination, many important accomplishments of modern computer science, in areas not commonly thought of as arti‹cial intelligence, all contribute to the achievement of its engineering goals. The autonomous humanoid robots beloved of the popular imagination merely represent one point in a multidimensional continuum of intelligent automation of human capacities. The familiar spelling and grammar-checker are points in this space, as are the 5 tax-advisor expert system, the intelligent personal assistant, the website shopping program, the chess-playing program, the Warrior military robot (Singer 2009), SONY’s AIBO robotic pet dog,1 and the planetary surveyor robot (Pedersen et al. 2003). The term agent in computer science technical literature represents a broad cluster of technologies and a large research program within arti‹cial intelligence, all concerned with relatively autonomous information -processing systems.2 Agents might be de‹ned as “a piece of software that acts on behalf of its user and tries to meet certain objectives or complete tasks without any direct input or direct supervision from its user” (Borking, van Eck, and Siepel 1991); as “computational systems that inhabit some complex dynamic environment, sense and act autonomously in this environment, and by doing so realize a set of goals or tasks for which they are designed” (Maes 1995); or, as the canonical textbook of the ‹eld puts it, as “anything that can be viewed as perceiving its environment through sensors and acting on that environment through actuators ” (Russell and Norvig 2002, 33). The terms in this last de‹nition are capable of a broad application so as to capture a wide variety of entities (humans, robotic agents, and software agents). A web bot could receive network packets or user keystrokes as sensory inputs and act by reading and writing ‹les and network connections; a sophisticated mobile robot could receive visual and auditory data as sensory input and act on its environment through a range of hardware extensions (Russell and Norvig 2002, 34). Agents may also be classi‹ed on the basis of their functionality, resulting in a more granular taxonomy: ‹ltering agents sift through information directed at users, letting through only information of interest or relevance , or allowing the choice of particularly relevant items; search agents seek out information on behalf of their users; user interface agents monitor and regulate the interaction of their users with information systems ; broker agents mediate between buyers and vendors of products or services; work-›ow agents automate basic of‹ce tasks; system management agents manage the operations of an information system; and problem-solving agents function as expert systems for resolving, or helping to resolve, complex issues (Bygrave 2001). Such a classi‹cation scheme reveals how much of the computerized technology available to the everyday corporate or individual user is agentlike in its functionality. Agents can thus, most perspicuously, be understood as modeling humans’ abilities to act as representatives (Wooldridge and Jennings 1995). 6 / A Legal Theory for Autonomous Arti‹cial Agents [3.237.186.170] Project MUSE (2024-03-28 21:53 GMT) There is a risk of identifying “agents” only with certain research programs in arti‹cial intelligence and not recognizing that “intelligent agents” share deep commonalities with other programs not considered “agents”: “An agent is...