Happy coincidences and personalized filter bubbles

Foto: no more lookism / Photocase

What do internet users have in common with intrepid princes of the 18th century and travelers to exotic islands? Both are looking for a happy coincidence, or serendipity. Studies have begun to explore how frequently serendipity occurs on the internet.

The English term “serendipity” is hard to nail down precisely but, put simply, it means something like “happy coincidence”. It was coined by the English writer Horace Walpole in 1754. In a letter to his friend Horace Mann, Walpole described the fantastic adventures of three princes in what is now Sri Lanka, which was then known as Serendip. The princes made many unexpected and revealing discoveries on Serendip—as do many internet users on their daily forays into the web.

But what does serendipity mean? In most languages there is no corresponding term. Wikipedia offers a comprehensive definition: Serendipity, it says, is “a chance observation of something which was not originally sought, which represents a new and surprising discovery.” Serendipity connotes a warm, positive and sentimental feeling. Googleing serendipity, one finds many photo-montages of inspirational quotes against kitschy backgrounds.

For around 15 years, academic researchers have been interested in the phenomenon, in particular in the field of information technology. There have been investigations into serendipity in libraries, blogs, jazz music or in the arrangement of workplaces within large office blocks. In the office, the focus of such research is how to improve relationships between coworkers such that productive and positive experiences can arise when employees from different teams encounter each other more or less by chance.

The current discussion on serendipity in the internet was kicked off by Eli Pariser and his influential book on the “filter bubble”. Pariser advances the thesis that increasing hyper-personalization in the internet creates filter bubbles where users increasingly encounter information and search results which are personalized on the basis of their browsing and search behaviour. Examples for this are Amazon product recommendations, or messages in their newsfeed which Facebook considers relevant. These user experiences and search results largely chime with their current preferences and views.

To take a political example: If a user has previously mainly googled for and selected conservative material, Google will present them with more conservative-leaning material even when they enter more general political search terms. This has to do with the fact that Google algorithms create user profiles on the basis of previous search queries, so that they can be linked to appropriate offers and results.

Pariser fears that in the course of increasing personalization of internet services, other perspectives and contrary opinions will be harder to access, and that the average internet user will increasingly live in a filter bubble. This filter bubble will be algorithmically controlled and reinforced by major internet companies like Google, Facebook, Netflix, Spotify and Amazon.

Miriam Meckel, a specialist in communication and editor-in-chief of Wirtschaftswoche, a German weekly economic magazine, argues in a similar vein. In an essay for the German Federal Agency for Political Education, she notes that chance and serendipity are disappearing from the internet. In another article about Google’s introduction of personalized searches in 2009, she writes: “Everyone receives search results listed in such a way that they correspond to their previous preferences. The result is the creation of an individualized profile for every person, which then becomes a point of contact for the machine. In this way, the unexpected discoveries which can be produced by happy coincidences are being progressively eliminated. Quite simply, coincidence is being calculated out of internet use.”

While this perspective seems plausible, there are experts who take the opposite view. They say that instances of serendipity are in fact being fostered online. In particular they point to social media like Twitter. On Twitter, depending on who you follow, you frequently stumble by chance across valuable information.

In any case, until now these different assessments have not been subject to empirical investigation. In the political context, a study carried out by Facebook in 2015 made waves. This study showed that personal preferences contributed more strongly to avoiding opposing opinions than Facebook’s algorithmic filtering. However, the study was limited to the US and to political content. It was heavily criticized in some quarters.

In order to shed some light on the subject, we conducted our own study at the University of St. Gallen in Switzerland in Autumn 2013, based on an online survey of 1173 internet users. In the survey, we distinguished between three areas: online shopping (Amazon as typical example), social media (Facebook as typical example) and information (Google as typical example). In each of these areas, we measured the perception of serendipity, and tried to explain the findings.

What did we discover?

The respondents experienced the most serendipity in the context of information, followed by social media. Serendipity was experienced least in online shopping. While we did not ask any open questions about specific experiences of serendipity, the results do suggest that many respondents regularly happen upon interesting texts or photos when searching for information, in a way that is surprising and coincidental. The finding that serendipity is most pronounced in the context of information tallies with previous literature on the subject. This locates serendipity firmly in the fields of information science, libraries and archives.

Does serendipity lead to more satisfactory user results? This is only true of social media. For online shopping and information services, there is no significant relationship between chance discoveries and satisfaction. Here, users do not appear to evaluate serendipity positively or negatively.

Different factors serve to explain the different experiences of serendipity in each of the three contexts. In online shopping, trust plays the most important role, whereas on social media divulging personal information most reliably promotes serendipitous experiences. When it comes to information services, self-confidence regarding one’s own efficacy (i.e. perceived user competence) is the most influential factor. In the online shopping context, the important role played by trust can be explained insofar that the risk of abuse is higher than on social media and information services because financial transactions are normally involved.

The positive effect on serendipity created by releasing personal information was a surprise for us, as we had expected the opposite: according to the filter-bubble argument, divulging personal data ought to lead to less serendipity, because more personalization takes place. But our results showed that on social media at least, that is not the case. We assume that supplying personal information enriches the user experience, and covers over the personalization effect. The finding that perceived user competence has the strongest effect on the experience of serendipity in the information context makes clear the central role played by self-confidence and competence on the internet. This result chimes with other studies which show how important reading and writing skills are online.

The results show that experiences of serendipity are most distinct in information-rich environments, but are most sought after in social environments.

Presently, non-personalized search engines like “Duckduckgo” have a negligible share of the market—probably because most users value personalized search results. In social media platforms too, the current trend is toward more personalization and away from serendipity. Within a few months, Twitter and Instagram have both altered their newsfeeds, so that instead of being purely chronological, they are now ordered by personal preference. But the move received an overwhelmingly negative response from users and in the media. Our results indicate that a certain portion of the dissatisfaction is to do with the loss of serendipity, which is associated more with a purely chronological layout. And what does that show? The desire for serendipity is still an important motive for users.

Christoph Lutz

Christoph Lutz

Christoph Lutz is Assistant Professor at the Nordic Centre for Internet & Society and in the Department of Communication and Culture at the BI Norwegian Business School (Oslo). He researches social media and digital communication. As part of his work he investigates new forms of participation, questions of internet privacy, serendipity, the uses of social media in science, the sharing economy and social bots.
Christoph Lutz

Latest posts by Christoph Lutz (see all)