SBHD: Intelligent agents are invading our screens in a quest to improve relations. Michael Evamy goes under cover
Remember that Microsoft ad on TV where the middle-aged woman starting a new secretarial job is shown her desk and, despite never having touched a computer before, has the thing licked within 30 seconds and has audited the entire company by elevenses?
Does this ever happen? I think not. Despite the claims about the intuitive and natural qualities of modern operating systems, it still takes years to master a substantial program; that is, to be able to click the “yes” box and always know what will happen next. Perhaps it is only the young who can claim utter confidence. To the elderly, even the most amenable interface can appear alienating.
Researchers have recognised this and are developing the first generation of on-screen guides that will take human-computer interaction to a new level. These devices bear the Le Carr-esque tag of “intelligent agents”. If they render computers and networks accessible to all, they will boost sales of hardware and software. These friendly go-betweens will also simplify affairs on the Internet and aid the most uninstructed in accessing services.
Microsoft has taken the first step in the intelligent agent race. In January, Bill Gates launched the chummiest-of-chummy slices of software, Bob. Bob is a “social interface” which uses cartoon characters to guide users around a package of eight computer applications, including letter-writing, e-mailing, and accounting. The characters let you know how well you’re doing by squeaking inanities like, “There you go” and “Whaddya know camper”. Microsoft says Bob fulfils users’ need for a deeper relationship with their computer.
Clifford Nass is one of the Stanford professors who helped develop the system. “People deal with their computers on a social level… We see people unconsciously being polite to the computer, applying social biases, and in many ways treating the computer as if it were a person.”
BT Labs at Martlesham is going further by developing intelligent agents for use to communicate across its own internal network. But, in the long term, they are also developing them for use as on-screen servants carrying out tasks for consumers on mass networks, such as the Internet. They could, for instance, gather information to compare holidays. The volume and disparate nature of information on the Internet is fast becoming overwhelming. Agents would remove the legwork.
BT’s agents have been developed to mimic the way humans behave – by trial and error – and to learn from experience in carrying out tasks. BT is staying tight-lipped about how its agents are programmed, but neural networks and game theory have been major influences. What it has yet to discover is the best way for the agents to communicate with the user. BT wants to animate its agents with the kind of verbal and gestural signals that humans use to communicate with each other.
Dr Chris Winter is BT’s project leader on its intelligent agents program. The design of the agents’ on-screen representations and behaviours are all-important, he says. “If everyone is going to be using these information services coming into the home, the interface has to be something they feel comfortable with. If you are asking a machine to do tasks on your behalf, you begin to want feedback about how well it has got on.
“If you’re managing a big telecoms network, you might have an icon that can do the equivalent of taking down Birmingham’s main exchange, but you don’t want to double click and get a prompt saying, `Did you really want to do that? Yes or no?’ You want feedback as to how important that action is. If I go to a human and say, `I want to switch off Birmingham’s main exchange’, he’ll say, `Are you really sure? You’re taking out two million people! I’d really like a reason why…’ And you get an emotional feedback that he is really very unhappy about this.”
Winter describes a system developed at MIT MediaLab that uses cartoon characters to represent pieces of software. One took the form of a small boy who smiled if he was involved and went into a sulk if no-one played with him. “Our work involved giving the icons properties. As you approached one, it might burst out into spikes and spin round, which looks dangerous and discourages you from touching it.
“Teenagers reacted to the smiling boy as they would react to a human smiling – which is very different to the way they approached a computer interface. It didn’t matter how computer literate they were, it wasn’t like having to cope with a computer. It fitted in with our view of the future: that computers will become, from the outside, more like humans.”
Which makes me wonder what people in white coats reckon is “human”. A computer that sat on my desk sulking, grinning and grimacing all day would soon invite treatment of an inhuman nature.
Intelligent agents could, on the other hand, succeed in opening the superhighway to a wider constituency by turning intangible entities like software and processes into palpable presences. Designers can help: one of London’s top product design studios is already involved with an intelligent agent project. There must be opportunities for others, perhaps more used to adding personality to real-life objects, to create enduring characters for this new generation of virtual vassals.