Measuring Trust in Social Networks

A common assumption about social networks is that linked people trust one another's recommendations and opinions. University of Minnesota researchers test such assumptions in large-scale experiments conducted in live social networks in order to shed light on fundamental—and complex—human dynamics like trust. In the physical world, more shared social connections is a good predictor of trust between individuals. Because of this, it's often assumed to be the same in online networks. In this paper, Ravi Bapna and Alok Gupta, together with their co-authors, examine Facebook activity to test the strength of online social bonds. Using actual Facebook friends as test subjects, the researchers custom-designed a Facebook app for a well-known investment game that uses real money to measure trust between participants. Combing game data with user characteristics and behaviors culled from the Facebook API, the researchers developed a predictive model for trust based on empirical evidence. The results showed: » Number of common friends did not predict trust, whereas other kinds of Facebook activity did—especially wall posts and number of photos together. » Results differed for some groups of people. For users who are more selective about connecting on a social network, and thus have fewer friends on average, there's a stronger relationship between Facebook activity and trust. For users who are less selective, and thus have a greater number of friends overall, only the photographs predicted trust. Randomized experiment using a previously tested game technique to ensure validity Switching regression model used to examine difference in behavior between groups of people with different characteristics TO O L S php STATA MySQL Facebook API I MP A C T Trust is a crucial element of many new business models in the sharing economy. Online services like Airbnb, GetAround car rental, and TaskRabbit connect individual buyers and sellers to unlock economic potential. For such models to work, the parties need to trust one another. Broader applications include better digital marketing strategies for any brand trying to tap into the word of mouth potential of social networks. Traditional thinking about spreading marketing messages and promotions through people with high friend counts do not appear to be as effective as believed. Unbiased evidence like that found in this study can help marketers build more sophisticated approaches.