Friends or contacts might be unwittingly giving away information about the likes or dislikes of people who have shunned or abandoned digital social media networks, a team of scientists said on Monday.

The scientists, based in Australia and the US, who have analysed over 30 million tweets from 13,905 Twitter accounts say their study has shown that online posts and the words of friends can be used to predict a person’s future activities.

They said their findings suggest that computer software could be used to extract — with an expectation of 95 per cent accuracy — information about individuals from text exclusively posted by their friends.

“Our results question the traditional view of privacy,” Lewis Mitchell, a senior lecturer in the school of mathematics at the University of Adelaide who led the research, told The Telegraph. “It has been generally assumed that you can protect your privacy by simply not joining social media. But we find that this is not necessarily true.”

The study by Mitchell and his colleagues, published in the journal Nature Human Behavior, has found that information from the tweets of eight or nine contacts of a person can be used to predict a person’s future posts.

They said an entity — whether a company, government or other organisation — with access to all social media data has only slightly more potential predictive accuracy (64 per cent versus 61 per cent) with access to posts by a person than with access to posts only from the person’s friends or contacts.

Their findings have implications for privacy. “Information is so strongly embedded in a social network that one can profile an individual from their available (digital) social ties even when the individual forgoes the platform completely,” the researchers wrote.

The scientists examined 30,852,700 public posts from 13,905 account holders on the Twitter social media platform, analysing the content from within 927 so-called “ego-networks” — each network consisting of one user (the ego) and their 15 most frequently mentioned Twitter contacts (the alters).

They say their study suggests that a platform provider can use information from a user’s social ties as a substitute for missing information about that user.

However, the scientists have acknowledged that in reality, this substituted predictive information may become dated over time as people change their social ties or behaviour. Such changes would challenge prediction.

Earlier research on digital social media networks had pointed to what scientists call the “echo chamber” effect, which results from people making friends with others who are likely to say the same things as they would.

“This study does a good job in actually measuring similarity and proving it happens in a quantitative way,” said Dheeraj Sanghi, professor of computer science at IIT Kanpur.