TL;DR
This paper investigates how new words and phrases (linguistic innovations) spread through online communities. By analyzing data from Reddit, the authors found that while central users often introduce new terms, it's the "strong-tie" users within smaller social circles who are crucial for their sustained adoption and spread. This work offers a data-driven validation of sociolinguistic theories in online environments and has implications for natural language processing systems.
What Problem Does This Paper Address?
Language is constantly evolving, with new words and expressions emerging all the time. However, not all innovations become widely adopted; many fade away. Sociolinguistics has long been interested in why some linguistic innovations succeed while others fail, and who the key players are in this process. Traditionally, theories have focused on the role of "weak ties" (acquaintances) in bridging social groups and introducing novelty, while "strong ties" (close friends/family) are thought to solidify new norms within tight-knit communities. However, testing these theories on a large scale, especially in the context of modern online social networks, has been challenging. This paper aims to bridge this gap by examining the birth and diffusion of lexical innovations in large online communities, specifically investigating the relationship between a user's social role (defined by tie strength) and the success of the innovations they introduce.
Methods
The researchers developed a novel approach to define and measure "tie strength" between users in online social networks. Instead of relying solely on traditional centrality measures (like how many connections a user has), they used a measure based on the overlap of users' immediate social neighborhoods. This allowed them to distinguish between users with many weak connections and those with fewer, but stronger, connections within cliques.
They then applied this to a large dataset of Reddit discussions, treating each subreddit as an independent social network. For each network, they identified linguistic innovations (like acronyms and slang) that appeared after the community's initial phase. They tracked the "dissemination trajectory" of these innovations – how widely they were used over time. Finally, they used this data to:
- Characterize Innovators: Compare the tie-strength and centrality of users who introduced new terms versus the general user population.
- Analyze Strong-Tie Users: Investigate the role of users with high tie-strength in the spread of innovations, examining how sustained adoption by these users correlates with future dissemination.
- Predict Innovation Success: Build a predictive model using tie-strength information from early adoption stages to forecast whether an innovation will become successful.
Datasets
The study utilized data scraped from 20 different subreddits on Reddit, covering a period of 4 to 8 years, resulting in approximately 10 million users and around 8,000 linguistic innovations. This dataset is publicly available.
Key Insights and Ideas
- Innovators are Central, but Weakly Connected: Users who introduce new terms are often central figures in the community (well-connected), but they tend to have lower tie-strength. This aligns with the idea that individuals bridging different social circles are more likely to introduce novelty.
- Strong-Ties Drive Sustained Spread: While weak-tie users might introduce innovations, it's the sustained adoption by "strong-tie" users (those within tight-knit cliques) that significantly increases the likelihood of an innovation spreading widely and becoming a norm. This provides empirical support for the role of strong ties in solidifying linguistic changes within communities.
- Tie-Strength is a Powerful Predictor: Even with limited early-stage data, a user's tie-strength is a surprisingly good predictor of whether a linguistic innovation will succeed. The longer the observation period for tie-strength, the more accurate the prediction.
- Online Communities Mirror Sociolinguistic Theories: The study demonstrates that established sociolinguistic theories, particularly Milroy's "Strength of Weak Ties," can be effectively applied and validated in large-scale online social networks, despite the different nature of social connections compared to offline communities.
Key Results
- Innovators were found to be more central (higher degree, betweenness, and eigenvector centrality) and more active (higher posting frequency) than non-innovators.
- Strong-tie users (tie-strength ≥ 0.5) constituted a minority (15-20%) but were more central and active than weak-tie users.
- The probability of an innovation's dissemination increasing after adoption by a strong-tie user grew significantly with sustained adoption (k consecutive months), a pattern not observed for weak-tie users.
- A prediction model using tie-strength information from the first 6 months of usage achieved an average F1-score of 0.68 for predicting successful innovations, significantly above the baseline. This improved to an F1-score of 0.76 when using data from the first year.
Limitations and Future Work
The authors acknowledge that their analysis focuses on specific types of linguistic innovations (Internet slang) and that other factors beyond tie-strength likely influence diffusion. Future work could explore:
- The impact of users belonging to multiple forums and how they transfer innovations between communities.
- The interplay of different types of linguistic innovations.
- Incorporating a wider range of factors into prediction models to improve accuracy.
Summary updated