Artificial human companions may be any kind of hardware or software creation designed to give companionship to a person. These can include digital pets, such as the popular Tamagotchi, or robots, such as the well-known Sony AIBO. Virtual companions can be used as a form of entertainment, or they can be medical or functional, to assist the elderly in maintaining an acceptable standard of life.
Senior citizens make up an increasing percentage of the population in the Western nations, and, according to Judith Masthoff of the University of Brighton, they tend to live alone and have a limited social network. Studies also show that those elderly living in such circumstances have an increased risk of developing depression and dementia and have a shorter life span than more socially connected seniors.
It has been known to gerontologists for some time that pets, particularly those such as cats and dogs that exhibit a range of behaviors and emotions, help prevent depression in the elderly. Studies also show some beneficial results from electronic pets such as Sony's Aibo and Omron's NeCoRo; however, the therapeutic value of such artificial pets remains limited by the capabilities of technology.
Television viewing among the elderly represents a significant percentage of how their waking hours are spent, and the percentage increases directly with age. Seniors typically watch TV to avoid loneliness; yet TV limits social interaction, thus creating a vicious circle.
It is Masthoff's contention that it is possible to develop an interactive, personalized form of television that would allow the viewer to engage in natural conversation and learn from these conversations.
Such applications have been with us for decades. The earliest, such as the "psychologist" program ELIZA, did little more than identify key words and feed them back to the user, but even forty years ago Colby's PARRY program at Stanford university, far superior to ELIZA, exhibited many of the features researchers now seek to put into a dialog system, above all some form of emotional response and having something "it wants to say", rather than being completely passive like ELIZA. The Internet now has a wide range of chatterbots but they are no more advanced, in terms of plausibility as conversationalists, than the systems of forty years ago and most users tire of them after a couple of exchanges. Meanwhile, two developments have advanced the field in different ways: first, the Loebner Prize, an annual competition for the best computer conversationalist, substantially advanced performance. Its winners could be considered the best chatterbots, but even they never approach a human level of capacity as can be seen from the site.
Secondly a great deal of industrial and academic research has gone into effective conversationalists, usually for specific tasks, such as selling rail or airline tickets. The core issue in all such systems is the dialog manager which is the element of system that determines what the system should say next and so appear intelligent or compliant with the task at hand. This research, along with work on computing emotion, speech research and Embodied Conversational Agents (ECAs) has led to the beginnings of more companionable systems, particularly for the elderly. The EU supported Companions Project is a 4-year, 15-site project to build such companions, based at the University of Sheffield.