Software agents: ethical issues concerning agent autonomy
Kane Baxter Bignell · 2005
This paper establishes a need for software agents in an Information Age where even competent computer users suffer information overload. The social consequences of agent use is examined, that to this time have not been adequately addressed by developers or governing bodies. Thus, there is a need for the establishment of a universally accepted, effective code of ethical agent practice in order to avoid agent misuse and threats to individual privacy. Consequently the code would protect users and ensure the future of the technology itself, as people become increasingly informed and suspicious of these intrusions such as spyware agents. The author proposes an Agent Ethical Code of Conduct, noting the possible future developments of such a Code should further research be undertaken in this area. INTRODUCTION TO AGENT TECHNOLOGY: ESTABLISHING A NEED FOR AUTONOMY The Internet, the most complete representation of information in western society, has fundamentally changed the way society utilises and interacts with computers. Business and individuals find themselves in transition to a networked, information society. The continued digitisation of content and networking of business and private computer users forms the basis of this information society where the computer is becoming increasingly ubiquitous now we are all connected through electronic networks. Consequently, the amount of data and software applications is continually increasing and growing in complexity. Users now have an abundance of information at their fingertips, yet the consequential diversity and complexity of how this data is represented can overwhelm even competent computer users finding it difficult to productively search through the abundance of information to find relevant material. Western society’s current technology related dilemma seems to be ‘information overload’ and the resulting stress and orientation loss. Thus, there is a need for assistance in this highly complex, digital environment from autonomous software agents where new methods of information dissemination and filtering are required to manage this overload of information. However, due to agents’ autonomous nature, there are potential hazards with their use. Because of this, agent developments must involve not only attention to technical details, but also to the ethical concerns relating to their resulting impact. In the excitement of new and emergent technology such as agents, Information Technology (IT) developers often forget to examine the impact new technology will have on users. In fact the social dimension of all technology is the driving force and the most important consideration of technology itself (Rieder, 2003). This paper encompasses analysis of the technical aspects of agent development evolving from fields of computer science, together with ethical issues deriving from the humanities and social science fields of IT. The technical analysis of agent software is necessary to form a basis for critical evaluation of the social ramifications of agent implementation, determining whether agents are a viable solution to the well documented problems of information overload, along with how current agent development techniques will affect the way users interact with computers in the future. EVOLUTION OF AGENT TECHNOLOGY: CURRENT APPLICATION OF AGENTS Software agents are one of the fastest growing areas of information technology, however, they are not a new idea. They are developments stemming from research in artificial intelligence (AI) which flourished in the 1980’s and 1990’s. AI is a field of computer science and engineering concerned with the computational understanding of what is commonly called ‘intelligent’ behaviour. To this extent, software agents are currently being developed with the expectation they will one day possess the same level of intelligence as humans, to actively learn from and assist computer users. However, to this point researchers have not been successful in creating an ‘intelligent’ agent, as they still do not know exactly how the human brain functions, yet many authors still wrongly label this category of software as intelligent agents. It is also inappropriate to regard agents as intelligent as measuring and defining intelligence is considered especially subjective due to our own expectations, therefore it is almost impossible to label software as ‘intelligent’. Bradshaw’s (1997, p.16) argument that “one person’s ‘intelligent agent’ is another person’s ‘smart object’; and today’s ‘smart object’ is tomorrow’s ‘dumb program’” emphasises this subjectivity where intelligence is only reflected by the activity or role that it plays in assisting users. Without a direct application or use, software intelligence is meaningless. To date the only software agents in operation use pattern recognition algorithms to assist people performing routine, or personal tasks. These agents can be seen in operation through such electronic commerce sites as amazon.com or Barnes & Noble’s online bookstores (http:// www.bn.com) where user demographics are analysed to give personalised service to users. Pattern recognition agents are also used for e-mail prioritisation, in education and also to aid Internet users in searching for information such as news articles. Microsoft has recently (July, 2004) launched an agent-based news filtering application they call ‘Newsbot’. The agent scours 4,800 news outlets to find the most relevant news headlines that would be of interest to the user (AustralianIT, 2004a). From these applications the author identifies that there is a fundamental difference between software agents currently used over the Internet, compared to the agents which are being developed to replicate human intelligence. SOFTWARE AGENT CHARACTERISTICS: APPLYING A UNIVERSAL DEFINITION It has not yet been possible to agree on a universally acknowledged, comprehensive definition of a software agent, as their developments are influenced by criteria from various research fields, notably artificial intelligence and information communications. Ultimately the name ‘agent’ is an umbrella term which covers many different applications. Other aliases often utilised include personal assistant agents, autonomous agents, information agents, intellibots, chatterbots, databots, knowbots, softbots, userbots, taskbots, personal assistants, avatars, personal agents and network agents. The reason these different labels exist is that agents promise such diverse applications and therefore authors have been hesitant to restrict research by applying a universal 701 E. Chocolate Avenue, Suite 200, Hershey PA 17033, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com ITP5248 IDEA GROUP PUBLISHING This paper appears in Managing Modern Organizations Through Information Technology, Proceedings of the 2005 Information Resources Management Association International Conference, edited by Mehdi Khosrow-Pour. Copyright 2005, Idea Group Inc. 55