Could Google have caught the Norway killer?

Norwegian daily Dagbladet reported that terrorist Anders Breivik spent 200 days searching Google for terms such as "how to make a bomb" before he went on his rampage.

Already Google uses data from searches to improve its targeted advertising and search rankings - and even to track the spread of the flu - leading academic ideas site The Conversation to ask the question "Why didn't Google catch the Norway killer?"

Speaking to the website, Charles Sturt University computer science lecturer Craig Wright said he expected people in European countries to call for search engines to monitor user behaviour as a result of the attacks.

He said it "would be possible, but extremely difficult" for Google to monitor searches and raise red flags for users who frequently and repeatedly perform suspicious searches.

Wright's comments were prescient as in recent days politicians and police across Europe have called for expanded internet monitoring powers as a possible preventative measure.

"There may be moves towards greater monitoring, but less privacy for users," he said.

Already, Wright said, the Australian Federal Police and the FBI could obtain court orders asking Google to monitor and flag certain search terms.

Google can monitor or trace specific individuals via its Google accounts, IP addresses or even browser cookies.

Nigel Phair, a former Australian Federal Police officer who is now a private security consultant, said real-time monitoring of search queries would be a welcomed "additional intelligence tool for national security agencies". Google could theoretically be a "good citizen" and report users it has identified as potentially high risk.

However "you have to get over the privacy hurdles" first, which will be a significant barrier.

Futurist, inventor and judge on ABC's The New Inventors, Mark Pesce, said: "I suspect Chinese search engine Baidu does something like this for the Chinese government."

Professor Sanjay Chawla, head of the school of information technology at the University of Sydney, is researching how to make machines and algorithms learn and behave like humans do. For instance, he is helping the insurance industry design technology to distinguish between legitimate and fraudulent claims.

Professor Chawla said using Google to spot terrorists was a tough technical challenge as "it's very hard to train computers or design algorithms which can distinguish between a terrorist and a non-terrorist".

"It could work but there will be lots of false positives and that can actually hurt Google more," he said.

"The number of terrorists is actually very small to create a meaningful profile of what the average terrorist looks like."

In the case of targeted advertising, if Google's algorithm got it wrong the worst outcome would be slightly reduced profits. But when picking out terrorists there is a far lower tolerance for errors.

And the idea of Google regularly monitoring search terms without a court order would raise significant privacy issues.

"Yes, some kinds of automated monitoring would occasionally produce a lead. But the leads would be swamped by the massive false positives," said Roger Clarke, chairman of the Australian Privacy Foundation.

"And the collateral damage to freedoms would be massive."

Danny Sullivan, editor-in-chief of Search Engine Land, raised the point that searching for something - for example bomb making instructions - wasn't the same as actually committing the criminal act.

And he also said that searching for mass murderers could be the thin edge of the wedge.

"Would it only look at things it believes are related to mass murder? If you were searching for recreational drugs or pirated content, should that be picked up and reported?" said Sullivan.

Of course, any real-time monitoring would be useless without law enforcement using it to act. It has been reported that Breivik was flagged for illegally purchasing a large amount of chemicals online but Norway's intelligence service decided he was likely to be using them in his farming business.

Mr Clarke said people with serious intent would find plenty of ways to get around the monitoring, while the suggestion that widespread real-time monitoring could enable real-time interventions by law enforcement agencies was "fantasy land".

But Mr Clarke said he would support tightly targeted monitoring of people based on reasonable grounds for suspicion and judicial warrants.

Google Australia declined to comment.

This reporter is on Twitter: @ashermoses