Listen to this article
It is 120 years since Alexander Graham Bell patented the telephone, while next year will be the PC’s 25th birthday.
Yet these technologies, which are so mature in many ways, are still hopeless at fitting in with our lives. Unimportant telephone calls distract us from more pressing tasks, while computers notify us about e-mail and instant messages when we’re concentrating on other things.
Could modern systems be made more considerate by understanding our working patterns and adapting their behaviour to suit?
Eric Horvitz thinks so. The senior manager and research area manager at Microsoft Research has worked for years on building context-sensitive IT systems that know when to leave us alone until we have finished what we are doing.
“We have systems that calculate the cost of interrupting someone in dollar terms. How much would a person be willing to pay to avoid an interruption of this type in this situation?” says Horvitz. “We can also forecast when that person will have a lower cost in the future, to help schedule communications.”
Horvitz’s system uses artificial intelligence to guess when a person may be contactable and then tells callers how long they can expect to wait. While not yet commercial, Microsoft trialled this technology using Bestcom, a system managing telephone calls, for 12,000 employees at its Redmond headquarters.
The idea sounds promising in theory, but is not yet for sale. Telling your computer how you like to work takes considerable effort, warns Kara Pernice Coyne, research director at the Nielsen Norman Group, a consultancy that advises on user interface design.
For example, what should callers be told in the event of a personal emergency, and should different callers be given different levels of information? Who should calls be routed to and how does that change for different callers? Whose e-mails and calls should be delivered at all costs?
“People hate customising, so they really have to see the benefit of doing it,” she says, explaining that such systems will use user profiles to try and capture that information in one place. “The profiles must be generic enough to create them easily, but when they’re too generic they don’t work properly.”
Defining profiles and rules that tell systems how you want to be contacted is just the start, says Martin Isley, director of Accenture’s technology labs. More sophisticated systems would guess about your behaviour by analysing data that the system collects about you.
“Let’s say every time you’ve been in a meeting and I’ve called you, you’ve always been short with me and asked me not to call again,” says Isley. “Somehow the system needs to sense that and create a rule that you can’t be interrupted in that case.”
Perhaps the system would realise that those calls only lasted a few seconds, and make a rule based on that information. Accenture is also researching systems that analyse emotions during phone calls to help inform such decisions.
Inferring new rules from our behaviour might seem like science fiction but most human activity is depressingly predictable.
For example Reality Mining, a project run by MIT’s Media Lab, used data from participants’ mobile phones to build a profile of their behaviour. Using data such as cellphone tower ID and Bluetooth connections made to nearby phones, researchers found that they could predict the movements of individuals with 90 per cent accuracy. That type of data could make contextual computing much more realistic.
While mining existing data can help, Paul Garner, CTO with British Telecom, predicts that context-sensitive computing will really take off with the introduction of physical sensors. BT is already experimenting with battery-powered radio sensors in the home that sense heat and vibration.
“All of that data from 40 sensors in each home is transmitted over a broadband link back to an intelligent data analysis platform that lets us infer high-level human activity from low-level data,” he says. Placed in consumer appliances, such sensors could collect all types of data, meaning that your refrigerator could end up turning on more than just the light when you open the door.
Why is that important? The application of such systems could be broad, but home healthcare could be one. Understanding the behavioural patterns of the elderly, or someone in the early stages of Alzheimer’s disease, could alert a computer when they deviate from the norm. The result: perhaps a computer that can switch on a sound system at the right time to remind a person to take their pills.
But as computers grow smarter, they will need ways of communicating. Researchers are now looking at this. A voice over a stereo system could be one method. BT, meanwhile, has been experimenting with furniture-like devices that fit into the home or office that have a communicative function.
“Our view is that what you want is completely stripped down. You don’t want a line of text in your face,” says Mr Garner. “You want something that just changes the ambient environment.”
Prototypes in BT’s labs include orbs that glow different colours according to preset conditions, and an electronic flower with petals that fold or unfold to tell you different things. Perhaps an orb would glow red if someone important was trying to reach you, or if your traffic route to work was congested. A petal on BT’s electronic flower representing a loved one could begin to fold inwards if you hadn’t answered their emails or called them in a while.
It’s a nice idea, but BT found the same problem that Nielsen Norman Group’s Ms Coyne identifies: “Some of the ambient interfaces are fun but many of them are pretty binary. It’s a light, but that’s it,” she says. Unlike computer screens, which can display lots of complex information, a glowing orb can only tell you so much.
Physical sensors and ambient orbs could help soften our relationships with computers in the future, but before this happens we have to overcome some problems. Privacy will be a huge barrier, as employees and consumers worry about the big brother aspect of computer systems that make decisions based on our movements.
Getting all vendors to develop and adopt standards that enable physical sensors to talk to each other will be far from easy. And then, companies will have to develop context computing technology that doesn’t make mistakes.
An unexpected reboot every so often is one thing, but blocking an emergency call from a loved one because the computer thinks you’re too busy is a system bug that we all could do without.