The anatomy of a scientific rumor

Abstract

The announcement of the discovery of a Higgs boson-like particle at CERN will be remembered as one of the milestones of the scientific endeavor of the 21(st) century. In this paper we present a study of information spreading processes on Twitter before, during and after the announcement of the discovery of a new particle with the features of the elusive Higgs boson on 4(th) July 2012. We report evidence for non-trivial spatio-temporal patterns in user activities at individual and global level, such as tweeting, re-tweeting and replying to existing tweets. We provide a possible explanation for the observed time-varying dynamics of user activities during the spreading of this scientific "rumor". We model the information spreading in the corresponding network of individuals who posted a tweet related to the Higgs boson discovery. Finally, we show that we are able to reproduce the global behavior of about 500,000 individuals with remarkable accuracy.

Figure 1. Probability density of in-degree, out-degree and total degree of the nodes that tweeted about the Higgs boson.

The corresponding distributions have been shifted along the y axis to put in evidence their structure. Dashed lines are shown for guidance only.

Figure 2. Number of tweets per second as a function of time during the period of data collection.

The curves correspond to tweets containing only the CERN, Higgs, LHC keywords and at least one of them, respectively.

Figure 3

Top: heatmap for the density of tweets before (left panel), during (middle panel) and after (right panel) the main event on 4^th July 2012. Bottom: corresponding networks of re-tweets between users. During the announcement, the Twitter activity is truly global, whereas before and after the announcement, the most active countries were European and American, due to the large presence of scientists in these geographic areas. The map in this figure was generated using TileMill and data from OpenStreetMap contributors, available under the Open Database License (see http://www.openstreetmap.org/copyright).

Figure 4. First and second panels: Global spatio-temporal activities of any user in the social network.

Number of entries for inter-tweets times (first panel) and inter-tweets spaces (second panel) between consecutive tweets, before, during and after the main event on 4^th July. The dashed line indicates a power law ~ τ⁻² and is for guidance only. Third, fourth and fifth panels: Joint probability density of inter-tweets times and inter-tweets spaces between consecutive tweets before (third panel), during (fourth panel) and after (fifth panel) the main event. In both cases, only the sub-set of geo-located tweets is considered.

Figure 5. Activity inter-tweets times of users in the social network.

First panel: Number of entries for inter-arrival times between consecutive tweets, before and after the main event on 4^th July. Power scaling behavior is visible for certain ragnges of values, dashed lines are for guidance only. Second panel: Number of entries for inter-arrival times between consecutive tweets after the main event. The dashed curve corresponds to a lognormal fit. Third panel: Number of entries for inter-arrival times between replies. Fourth panel: Number of entries for inter-arrival times between re-tweets.

Figure 6. Visualisation of the social network of active users, based on k-core decomposition and components analysis.

The size of each vertex is proportional to its degree, whereas color codes the k-coreness. A sample of 10% of the whole network has been used for this visualisation.

Figure 7. Points indicate the fraction of users who are active at least once (see the text for more detail) with respect to the total number of users in the dataset at the end of the period taken into consideration, i.e., (t = 8 July 2012), as a function of time.

Lines indicate the fitting results obtained separately for each temporal range by adopting the model given by Eq. (3). The rate of activation for each period is reported at the bottom of the figure.

Figure 8

First panel: Evolution of the density of active users versus time obtained from simulations of spreading dynamics. The de-activation rate is β = 1 and the different curves correspond to different values of the activation rate λ. Each curve corresponds to the ensemble average of 200 random independent realizations. Second panel: Average value of the density of active users in the stationary state, as a function of λ. Third panel: Observed evolution of the density of active users versus time (points) in Period IV, i.e., during and after the main event, from 03:00 AM, 4^th July. Curves indicate the predictions obtained from the model defined by Eq. 8 coupled to Eq. 9, where the values of the corresponding parameters are reported in the figure for different sub-periods. The reported λ refers to the initial value of the activation rate.