Internet Balkanisation and Increasingly Automated Personalisation

TREND: Behavioural advertising and personalised search optimisation contribute to the creation of balkanised “echo chamber communities” insulated from unfamiliar or alternative cultures and perspectives

In 1995 Nicholas Negroponte, founder of the Massachusetts Institute of Technology Media Lab, published his book “Being Digital” in which, alongside multiple predictions on the future applications of technology, he introduced the concept of the “Daily Me”, a virtual newspaper which was customised for each individual subscriber (Making the Daily Me, Neil Thurman, page 2). Today it is manifestly evident that Mr Negroponte’s vision was a prescient one. 

A 2012 paper from the University of Pennsylvania (see page 2) claims that web personalisation (using statistical techniques to infer a customer’s preferences and recommend content suitable to them) has now become ubiquitous across multiple providers of online news, media and services. A 2011 study from City University (see page 5) surveyed the use of online personalisation across eleven national US and UK news websites and reported an increasing range of approaches including contextual recommendations (links to external content), geo-targeted editions (based on user location), aggregated targeted filtering (selections of news stories filtered by general user popularity) and profile based recommendations (based on data on user behaviour from registration or imported from social media sites). A 2012 article, Existing Trends and Techniques for Web Personalization, in the International Journal of Computer Science Issues (see page 433) reports that web personalisation has become an indispensable tool for both web-based organisations and end users to deal with content overload, and that most major Internet companies are implementing personalisation systems.

In Republic.com 2.0 (2007) Cass Sunstein argues that the Internet has a propensity to foster social fragmentation by encouraging individuals to sort themselves into deliberate enclaves of like-minded people and assisting them in filtering out unwanted or opposing opinions (referenced in Comparative Research in Law & Political Economy, Book Review by Peter S. Jenkins, page 1). In studying hyperlinking patterns across 1,400 political blogs, Sunstein reported that 91% of links were to like-minded sites which indicated a trend towards rarely highlighting or drawing attention to opposing opinions (interview with Salon Magazine 2007). In a 2010 article for Scientific American, founder of the World Wide Web Tim Berners-Lee argued that the effects of social networking companies such as Facebook and LinkedIn shutting their users into online walled gardens could cause the Internet to be “broken up in to fragmented islands”.

In his 2011 book The Filter Bubble Eli Pariser contended that Facebook’s decision to change its filtering algorithm for status updates and newsfeeds (so that by default users would only see material from friends they recently interacted with) had the unintended consequence of suppressing updates from friends who did not share that user’s political and social values (Wall Street Journal article, 2011). Pariser also argued that increasing trends towards personalised search optimization (Google uses 57 different metrics to predict which search results will be displayed for different users) can also produce unexpected outcomes (2011 Slate Magazine article).  In the aftermath of the Deep Water Horizon oil spill in 2010 one user’s Google  search for “BP” yielded a set of links on investment opportunities with BP – while another generated links providing information on the oil spill (2010 interview with Salon Magazine). Pariser suggested that this invisible and automated customisation of our web experience risks trapping individuals in personalised information bubbles which insulate us from uncomfortable or unfamiliar views, cultures and perspectives.

Alongside Google, online retailers and media providers such as Amazon and Netflix have well established personalised recommender systems which direct users to content that is likely to interest them based on previous choices and search histories. In a 2011 interview with the New York Times, computer scientist Jaron Larnier claimed that this trend has a tendency to cocoon users within a personalised echo chamber where more and more of what they experience online conforms to an image of themselves generated by software.

However, it can also be argued that these trends merely reinforce or reflect traditional and long-standing human tendencies to engage with people, ideas and content which strike a chord with their existing values and interests – in terms of adopting a positive test strategy (Confirmation, Disconfirmation and Information in Hypothesis Testing, page 211, American Psychological Association 1987). Furthermore, research by the University of Michigan (commissioned by Facebook) revealed that while Facebook users are more likely to look at links or pictures shared by close friends, in reality they tend to get far more information from distant contacts, many of whom tend to share items users wouldn’t otherwise be aware of. This conforms to previous influential sociological research by Mark Granovetter (American Journal of Sociology: The Strength of Weak Ties 1973) that stipulates that individuals tend to form clusters of a few close friends, alongside larger numbers of disparate social acquaintances.

The International Journal of Computer Science (page 430) suggests that the tremendous growth in the number, size and complexity of information resources available online make it increasingly difficult for users to access relevant information in a context where an individual’s capacity to absorb, read and digest information is essentially fixed. In this context, personalisation is a necessary evil to prevent the information universe becoming progressively unintelligible to human enquiry. It is worth noting that Facebook’s 2010 decision to suppress certain types of newsfeeds was based on complaints from Facebook users that they were being inundated with updates from friends they barely knew (Wall Street Journal article 2010).