Testing Propositions Derived from Twitter Studies: Generalization and Replication in Computational Social Science

Abstract

Replication is an essential requirement for scientific discovery. The current study aims to generalize and replicate 10 propositions made in previous Twitter studies using a representative dataset. Our findings suggest 6 out of 10 propositions could not be replicated due to the variations of data collection, analytic strategies employed, and inconsistent measurements. The study’s contributions are twofold: First, it systematically summarized and assessed some important claims in the field, which can inform future studies. Second, it proposed a feasible approach to generating a random sample of Twitter users and its associated ego networks, which might serve as a solution for answering social-scientific questions at the individual level without accessing the complete data archive.

Collective aspects of privacy in the Twitter social network

“…To generate an initial unbiased random sample of users, we applied the Random Digit Search method [24,25]: We generated random Twitter user ids in the range between 1 and 30 Billion, looked them up through the Twitter REST API, b and saved the basic user information of the valid sampled users. To avoid celebrities and spammers, we filtered out users with a ratio of followers to friends below 0.1 or above 10, as well as users with less than 50 friends or followers.…”

Trace of pace, place, and space in personal relationships: The chronogeometrics of studying relationships at scale

“…Consistent with Dunbar's number, social networks are constrained by the number of social relationships that can be functionally sustained, typically between 80 and 300 nodes (Arnaboldi et al, ; LaRose et al, ; Liang & Fu, ; Mac Carron et al, ; Miritello, Lara, et al, ; Sandel et al, ; M. Stephens & Poorthuis, ; cf., Manago, Taylor, & Greenfield, ; Wellman, ). In addition to support from research on ordinary social networks (e.g., Dunbar & Sosis, ), various studies of mediated network topologies have tended to support these ranges of social ties (Arnaboldi et al, ).…”

“…Across three studies analyzing 15,000 users (Rogers‐Pettie & Herrmann, ), 200,000 original tweets, and over 100,000 users (Stattner, Eugenie, & Collard, ) and 400,000 users (Szell, Grauwin, & Ratti, ), most users posted only a single message, had few if any followers, and only a small percentage tweeted 10 or more times. In Liang and Fu's (, p. 4) study of 4.7 million tweets among over 32,000 users with almost 2.5 million unique alters, and almost 2.4 million of those alters' tweets, “2% of users created 80% of tweets, or 20% created nearly 100% of posts,” with over half of the users not posting any tweets. In a study of over 19 million tweets, Ferrara and Yang (, p. 4) found that only 10% of users had produced more than 200 tweets during the time span investigated and that most tweets were never retweeted (79%) or favorited (87%).…”

“…We exist in a complex communication ecology, in which our absolute amount of time (Miritello, Moro, et al, ), cognitive capacity for processing information (Cowan, ; Marois & Ivanoff, ), and ability to establish and maintain relationships (Arnaboldi, Guazzini, & Passarella, ; Dunbar, Arnaboldi, Conti, & Passarella, ; Gonçalves, Perra, & Vespignani, ; LaRose, Connolly, Lee, Li, & Hales, ; Liang & Fu, ; Mac Carron, Kaski, & Dunbar, ; Miritello, Lara, Cebrian, & Moro, , ; Zhao et al, ) are limited, whereas the potential for producing more information is virtually unlimited (Hilbert, , ; Hilbert & López, ). We interact in an attention economy in which every communicative activity competes at multiple levels with alternative activities (Ciampaglia, Flammini, & Menczer, ; Huberman, ; Miritello, Lara, et al, , ; Wagner, ; Weng et al, ), especially as information and connection overload nears (Feng et al, ; LaRose et al, ; K. K. Stephens et al, ).…”

“…In addition to support from research on ordinary social networks (e.g., Dunbar & Sosis, ), various studies of mediated network topologies have tended to support these ranges of social ties (Arnaboldi et al, ). For example, examining 4.7 million tweets among over 32,000 users, almost 2.5 million others, and almost 2.4 million tweets by those others, Liang and Fu (, p. 8) found that “the average number of replies per friend as a function of the number of friends reaches its maximum at around 80 friends, indicating the effect of the cognitive limits of human brain on the ability to maintain social relationships.”…”