A Legal Theory for Autonomous Artificial Agents

Chapter 3: Attribution of Knowledge to Artificial Agents and Their Principals

Chapter 3 / Attribution of Knowledge to Arti‹cial Agents and Their Principals The modern corporation’s ever-growing presence on the Internet and its dependence on sophisticated information technology means arti‹cial agents are increasingly engaged in tasks that make them the acquirers, processors, and transmitters of information relevant to their principals’ business activities. The legal status of this information—in particular, whether it can be considered the knowledge of the agent’s corporate principal—is crucially important. It is relevant to whether arti‹cial agents can receive notice from their principal’s clients for purposes of contracting, or whether a corporation can be proven to have guilty knowledge or “scienter” for criminal law purposes by virtue of facts known only to its arti‹cial agents. A coherent approach to the question of how knowledge may be attributed to arti‹cial agents and their legal principals will buttress the thesis that, as far as possible, arti‹cial agents can and should be treated like human legal agents. Such an analysis would address the similarities and dissimilarities between arti‹cial agents and human ones, and ideally permit the application of legal principles governing the attribution of knowledge of legal agents to their principals. This application requires a realistic account of corporate knowledge: it cannot sensibly be con‹ned to what is known by human agents of a corporation. In order to re›ect the reality of information management in contemporary commerce, corporate knowledge must also be held to embrace knowledge known only to arti‹cial agents, such as bots, auction agents, spiders, website shopping agents, and the like, that are essential for the commercial viability of the Internet-dependent corporation. 71 The legal principles governing the determination of corporate knowledge and the attribution of agents’ knowledge to their principals have grown and changed to accommodate a variety of fact patterns, management practices, and organizational strategies. What has not changed is the underlying axiom that the knowledge of a principal is crucial in determining his liability or culpability for a state of affairs. The doctrines developed in this chapter, then, will be increasingly relevant as a new kind of legal agent, one with enhanced capacities for information storage and processing, takes on an increasing amount of corporate functionality, thus involving the corporate principal in increasing potential liability. Attributing Knowledge to Arti‹cial Agents Attributing knowledge to humans is a task we engage in all the time. Considering how we may attribute it to arti‹cial agents is revelatory: it shows how a commonplace activity conceals an old, seemingly intractable philosophical problem, and how humans and synthetic information -processing entities like arti‹cial agents can be interestingly similar in terms of cognitive activities. Consider the claim “Joaquin knows Ann Arbor is the capital of Michigan.” This is an attribution of propositional knowledge, of the form “Entity X knows proposition p,” distinct from knowledge of a place or a person (knowledge by acquaintance) (Russell 1984), and knowing how to perform a task (knowledge by competence) (Ryle 2000). Philosophers have long pondered the necessary and suf‹cient conditions under which such a claim would be true, that is, when every case in which X knows p is one in which the analysis’ conditions are met (necessity) and when every case that meets the conditions of the analysis is a case in which X knows p (suf‹ciency). Plato’s Theatetus famously analyzed knowledge as justi‹ed true belief: that is, X knows p if and only if p is true; X believes p; and X is justi‹ed in believing p. The ‹rst condition captures the intuition an agent could not know a false proposition; the second the intuition the agent must bear a particular intentional attitude—that of belief—to the proposition in question; the third condition rules out epistemic luck, so that accidental cases of acquiring true belief do not count as knowledge. For instance, a person might believe there are aliens on the moon through reading a supermarket tabloid. Even if aliens were to be discovered on the lunar sur72 / A Legal Theory for Autonomous Arti‹cial Agents 3.1. face, we would be reluctant to attribute knowledge of that fact to her, for her belief does not appear to bear the right epistemic relation to the proposition “There are aliens on the moon.” The dif‹culty with Plato’s analysis is that justi‹ed true belief might fail to...