Cognitive automata and the law: electronic contracting and the intentionality of software agents

When speaking on general terms of an “agent”, I shall refer to any entity capable of autonomous action in general, following the AI terminology. In the law, on the contrary, the term “agent” usually denotes someone who acts on behalf of another. Both meanings, however, are relevant, since I shall consider autonomously electronic entities acting on behalf of their users. For useful links to research projects and companies dealing with intelligent agents, see http://​www.​aima.​cs.​berkeley.​edu/​ai.​htmlagent. For an approach to artificial intelligence based on the idea of an agent, see Russell and Norvig (2003). For roadmap on agent.-based technologies (though limited to year 2005) see Luck et al. (2005).

The thesis that the law gives some relevance to cognitive states does not entail that such states are always decisive: the law often needs to take into account, besides the perspective of the author of the act, also the way in which the act is understood by the counterpart and by third parties, as well as various pragmatic constraints. For instance. the law of contracts gives some limited relevance to the cognitive states of the author of a contractual declaration (by requiring that the party should intentionally make such declaration, in the awareness of its effects, and by allowing the contract to be annulled when certain mistakes were made in coming to a determination or in expressing it) even though such relevance may be overridden by further considerations, such as protecting the reliance or the counterpart (who justifiably assumed that the contract was regularly formed), or reducing litigation and facilitating the work of the judges (which may suggest that only in exceptional circumstances judges should override the text of the contract and its conventional linguistic meaning). The latter view is often said to characterise the British tradition, see Devlin (1962).

For a collection of some important contributions to the philosophy of mind, cf. Cummins and Dellarosa-Cummins (2000); for a basic introduction, cf. Davies (1998).

See Peczenik (2006, 79), according to whom legal justification should be as much as possible “philosophically neutral” and jurists should “avoid commitment to strong philosophical theories and prefer weak philosophical theories.”

I apologise for this trivialisation of the complex problem of evolution. For a “philosophical” introduction to Darwin’s ideas, see Dawkins (1989) and Dennett (1996).

For example, Nozick (1993) introduces the notion of a function on the basis of the concept of an homeostatic system, which he characterises as follows: “[An homeostatic system] maintains the value of one of its state variables V within a certain range in a certain environment, so that when V is caused to deviate some distance (but not an arbitrary long distance) outside that range, the values of the other variables compensate for this, being modified so as to bring V back within the specified range.” As examples of homeostasis, consider how an increase of bodily temperature may lead to sweating, which lowers the temperature, or how an increase in the temperature of a house may start air conditioning, which goes on until the temperature has fallen to the established level. Nozick defines then the notion of a function as follows: “Z is a function of X, when Z is a consequence (effect, result, property) of X and X’s producing Z is itself the goal state of some homeostatic mechanism M ..., and X was produced or is maintained by this homeostatic mechanism M (through its pursuit of the goal: X’s producing X)”. According to this definition we may say, for example that the function of thermostats is that of keeping temperature within the specified range, since stabilising temperature is the result which is obtained through the process of designing and building thermostats, a process which tends to make so that thermostats are build which can stabilise temperature (thermostats designers constantly endeavour to improve the performance of the thermostats they design). In the same way, the function of lungs is that of absorbing oxygen, since this is the result produced and maintained by natural evolution, which tends to make so that lungs are able to absorb oxygen (through the survival, and therefore reproduction, of the individuals whose lungs can absorb oxygen).

On idea that combinations of humans and artificial entities may represent a new kind of hybrid subjectivity see Teubner (2006), referring to Latour (2005).

For interesting considerations on the attribution of agency and intentionality to artificial entities, an attribution which—contrary to the approach here taken—is based on communicative capacity rather than on purposive rationality, and more on socially shared presumptions than on the nature of the concerned entities, see Teubner (2006).

On the legal aspects of contracts made by SAs, there is already a significant literature, see for instance Kerr (1999); Lerouge (2000); Bellia (2001); Weitzenboeck (2001); van Haentjens (2002); De Miglio et al. (2002); Weitzenboeck (2004); Wettig and Zehendner (2004); Kafeza et al. (2005); Barfield (2005); Andrade et al. (2007); Balke and Torsten (2008).

UCITA has been very controversial (especially since it allows contracts to override consumer protection rules) and been adopted so far only by two States, while being rejected by the American Law Institute.

Further significant legal issues concern the SAs’ use in virtual enterprises (cf. Cevenini 2002), in on-line dispute resolution (cf. Chiti and Peruginelli 2002; Gouimenou 2002), or in police investigations (Burkhard 2006; Abel 2009).

The idea that relying on automated helpers may compromise human values is at the centre of the work of Isaac Asimov (see for instance, Asimov 1996). For an optimistic view concerning the impact of robots on human society, see Brooks (2002).