mouse track

Monday, 10 September, 2001, 07:33 GMT 08:33 UK

Reading your mouse movements

Some mouse movements are common to all

By BBC News Online's Alfred Hermida

A website that can read your body language and know what you want before you have even clicked on anything may sound like science fiction.

But this is what researchers in the US are working on.

A team at the Massachusetts Institute of Technology (MIT) in Boston, US, say they have developed a way to record mouse movements on a page and learn how people behave when they are on the internet.

They found that certain movements of the mouse are common, allowing them to predict how someone uses the web.

This sort of information would be invaluable to content providers who are looking for ways of increasing the effectiveness of their websites.

"Just by looking at the way the mouse moves, I can tell whether you are reading a web page," says Ted Selker, an MIT professor focusing on context-aware computing.

"I can tell because when you read a webpage, you do one of a couple of things. You either shovel the mouse off to the right so that it is out of the way, or you will walk down the page with your mouse," he told the BBC's Go Digital programme.

Follow the Cheese

The system developed by the team at MIT is called Cheese, since they are following the mouse, like a mouse follows cheese.

This is incredible. Can you imagine if I can actually tell that you wanted to press a link but didn't

Ted Selker, MIT

It provides the means to find out exactly how people navigate a webpage.

The researchers say that if you could analyse in real time how someone goes through a website, the content and navigation could be adapted to create a more personal experience.

"This is incredible," says Mr Selker. "Can you imagine if I can actually tell that you wanted to press a link but didn't.

"And 75% of the time, I can tell that you were looking at a website but you didn't click to buy a vacation but were thinking of taking a vacation, while doing your travel plans."

"I can change the way the travel site prepares material for you based on what intentions and interests you've demonstrated through your actions that aren't even recorded in the links you followed."

Current technology

The system developed at MIT works by including mouse movement data automatically with embedded scripting. The information is analysed and stored on a server.

This collection technique is implemented using current technology and does not require any additional software on the user's browser.

For their study, the researchers took a group of 17 people familiar with computers and web browsing, but from diverse backgrounds.

They recorded the mouse activity as people performed a list of tasks common in web browsing, such as ordering a CD.

The data was evaluated by redrawing the mouse movements on each page for each user and then visually comparing the patterns on mouse behaviour.

Predicting choice

In one case, people were asked to buy a CD or DVD of their choice.

We're working very hard to make those kind of natural simple communications that people make with their body through computer interfaces

Ted Selker, MIT

By studying the mouse movements, the researchers were able to predict what their second choice would have been.

This was done by determining the link on which they hesitated longest before clicking their first choice.

Some people occasionally moved the mouse straight to the link of interest without hesitation. The MIT team believe this behaviour shows that a user has visited the page before and is familiar with its layout.

"People are extremely good at remembering graphic design," says Ted Selker.

"So when you act like you know where you are going on a place where you have no reason to know, then we know you have been there before."

Computer scenarios

The researchers now plan to put together a website with content that would change according to mouse behaviour.

The research by the team at MIT is part of their efforts to create a world where desires and intentions are enough to get computers to act on our behalf.

They aim to do this by developing environments that use sensors and artificial intelligence to create so-called "virtual sensors" - adaptive models of users to create keyboardless computer scenarios.

"We're working very hard to make those kind of natural simple communications that people make with their body through computer interfaces so that people spend less time and effort trying to laboriously remember what command to type," says Mr Selker.