The effect of wording on message propagation: Topic- and author-controlled natural experiments on Twitter

By Chenhao Tan, Lillian Lee, Bo Pang. 2014

209 citations recorded by the source corpus.

Consider a person trying to spread an important message on a social network. He/she can spend hours trying to craft the message. Does it actually matter? While there has been extensive prior work looking into predicting popularity of social-media content, the effect of wording per se has rarely been studied since it is often confounded with the popularity of the author and the topic. To control for these confounding factors, we take advantage of the surprising fact that there are many pairs of tweets containing the same url and written by the same user but employing different wording. Given such pairs, we ask: which version attracts more retweets? This turns out to be a more difficult task than predicting popular topics. Still, humans can answer this question better than chance (but far from perfectly), and the computational methods we develop can do better than both an average human and a strong competing method trained on non-controlled data.

Subfields: Artificial Intelligence, Statistical and Nonlinear Physics, Natural Language Processing, Social Sciences & Humanities

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistical and Nonlinear Physics

In Natural Language Processing

In Social Sciences & Humanities

Citations in the corpus

DeepCas: An End-to-end Predictor of Information Cascades (2016, background): “For example, whether a photo was posted with a caption is shown to be predictive of how widely it spread on Facebook [10]; specific wording on Tweets is shown to help them gain more retweets [33].”
Integrating explanation and prediction in computational social science (2021, background): “Conversely, in quadrant 3 is research that uses as much information as possible to passively forecast content popularity 48,59,60 .”
Creative Writing with a Machine in the Loop: Case Studies on Slogans and Stories (2018, background): “[31] that it is hard for humans to evaluate the quality of writing.”
A Survey on Causal Inference (2020, methodology): “Moreover, a series of methods have been proposed [44, 156, 169, 173] to handle the case when the treatment, confounder, or outcome contains textual data.”
Can Large Language Models Transform Computational Social Science? (2023, background): “…are computational analyses of social norms and conventions (Centola et al., 2018; Bicchieri, 2005), information diffusion (Leskovec et al., 2009; Tan et al., 2014; Vosoughi et al., 2018; Cheng et al., 2016), emotional contagion (Bail, 2016), collective behaviors (Barberá et al., 2015), and…”
Topics to Avoid: Demoting Latent Confounds in Text Classification (2019, background): “Tan et al. (2014) built models to estimate the number of retweets of Twitter messages and addressed confounding factors by matching tweets of the same author and topic.”
On the Diversity and Limits of Human Explanations (2021, background): “movie quotes to predicting the popularity of messages (Danescu-Niculescu-Mizil et al., 2012; Tan et al., 2014).”
Text and Causal Inference: A Review of Using Text to Remove Confounding from Causal Estimates (2020, background): “, 2019), text as treatment (Fong and Grimmer, 2016; Egami et al.; Wood-Doughty et al., 2018; Tan et al., 2014), text as outcome (Egami et al.”
Winning Arguments: Interaction Dynamics and Persuasion Strategies in Good-faith Online Discussions (2016, background): “In an ideal world, we would control for both length [13] and topic [25, 52], but we don’t have the luxury of having enough data to do so.”