Mind the gap : a critique of human/technology analogies in artificial agents discourse

摘要

CONNECTING HUMANS AND TECHNOLOGYand technologies should or will be bridged.This assumption is evident in the recurrent suggestion that the natural next step in technological development is to move computer technologies closer to humans by endowing them with human-like capabilities.AI researchers have projected images of intelligent agents that 'think' for or with a human user, that make decisions, learn, act autonomously, and anticipate and adapt to human behavior.These future artificial agents, so these researchers say, will be capable of supporting or replacing humans in a wide range of activities.They will profoundly change the way we relate to technologies.By learning about us, electronic 'personal assistants' will be able to schedule meetings, find music and books, organize and book trips and manage our e-mail, according to our preferences, habits and interests (Maes, 1994a).Artificial "teammates" will collaborate with their human partners in disaster response efforts and space missions (Sycara & Sukthankar, 2006).Sociable robots will serve as companions or tutors, once they are able to communicate with us in personal and intuitive ways (Breazeal, 2002).In our future homes and offices, we will be surrounded by sensitive and responsive computerized environments regulated by interacting agents (Aarts, Marzano et al., 2003;F. Zambonelli & V. Parunak, 2003).Although these visions project an enticing future, they raise some important practical and moral concerns.For one, the utopian rhetoric about promises of artificial agents is offset by questions and concerns about the unintended and undesirable effects that the development of these technologies can have.Critics and 'agent researchers' alike have pointed to issues of security, privacy and trust (Luck et al., 2005). 1 Can we trust agents to make decisions concerning our e-mail, our online transactions or the education of our children?Ben Schneiderman, a human-computer interaction (HCI) researcher and a critical observer of agent research, has expressed his concern that the anthropomorphic representations in this field mislead designers and deceive users."It increases anxiety about computer usages, interferes with predictability, reduces user control, and undermines users' responsibility" (Schneiderman & Maes, 1997, p. 56).One of the most ardent critics, Jason Lanier, even contends that the idea of intelligent agents is "both wrong and evil" (1995, p. 66).It misrepresents what computers can do, but more importantly for Lanier, in order for agents to look smart, people make themselves dumb.He warns against the consequences that 1 From now on I will use the term agent researcher to refer to individuals advocating and engaging in research on artificial agents.