The blogosphere as an excitable social medium: Richters and Omoris Law in media coverage

We study the dynamics of public media attention by monitoring the content of online blogs. Social and media events can be traced by the propagation of word frequencies of related keywords. Media events are classified as exogenous - where blogging activity is triggered by an external news item - or endogenous where word frequencies build up within a blogging community without external influences. We show that word occurrences show statistical similarities to earthquakes. The size distribution of media events follows a Gutenberg-Richter law, the dynamics of media attention before and after the media event follows Omori’s law. We present further empirical evidence that for media events of endogenous origin the overall public reception of the event is correlated with the behavior of word frequencies at the beginning of the event, and is to a certain degree predictable. These results may imply that the process of opinion formation in a human society might be related to effects known from excitable media.

💡 Research Summary

The paper investigates the dynamics of public attention by analyzing the content of a large sample of political blogs in the United States over a 670‑day period (July 2008 – May 2010). The authors collect time‑stamped entries from 168 high‑traffic blogs, extract daily frequencies of roughly 4,000 meaningful keywords, and define a “media event” as a sharp rise in the frequency of a keyword followed by a decay back toward baseline. Each event is characterized by its peak time t₀, peak frequency w(t₀), and an event size E, which is the ratio of the peak frequency to the average frequency in a 30‑day window before and after the peak.

Two distinct classes of events are identified. Exogenous events are triggered by external news items (e.g., the nomination of Sarah Palin as vice‑presidential candidate). These events show virtually no pre‑peak growth; the frequency jumps abruptly at t₀ and then decays. The decay follows a power‑law w(t) ∝ (t − t₀)^{‑γ}, with γ distributed similarly to the Omori law exponent for aftershocks in seismology (γ≈1). Endogenous events arise from internal dynamics within the blogosphere (e.g., the word “inauguration”). They display a clear pre‑peak growth phase w(t) ∝ (t₀ − t)^{‑α_g} and a post‑peak decay w(t) ∝ (t − t₀)^{‑α_d}, both obeying power‑law forms. Importantly, the growth exponent α_g and the decay exponent α_d are strongly positively correlated (ρ≈0.67, p≈10⁻²²), indicating that events that rise quickly also fade quickly—a pattern reminiscent of the relationship between foreshocks and aftershocks in earthquakes.

The distribution of event sizes follows a Gutenberg‑Richter law: Pr(E* > E) ∝ E^{‑(β+1)}. For exogenous events β≈1.00 ± 0.03, while for endogenous events β≈0.57 ± 0.04, the latter being close to the canonical seismic exponent β≈2/3. Exogenous events are roughly ten times more frequent than endogenous ones, suggesting that external news generates “earthquake swarms” of many small shocks, whereas internal cascades are rarer but follow the same scaling.

Methodologically, the study proceeds as follows. First, a proprietary Java crawler continuously downloads all posts from the selected blogs, storing timestamps, headlines, and full text in a SQL database. Short high‑frequency stop words and common suffixes are removed. The authors compute frequencies of all 26³ possible three‑letter trigrams for each day, discard those never occurring, and select the day with the highest frequency for each remaining trigram as a candidate peak day t₀. For each candidate, they compile a list of words containing the trigram and extract their daily frequencies in a ±30‑day window. Only words whose maximum frequency occurs at t₀ are retained, yielding about 4,000 keywords.

Event detection involves sliding a 30‑day window across each keyword’s time series, locating local maxima, and fitting the pre‑ and post‑peak portions to the power‑law forms (Eqs. 2 and 3). Fits with a residual sum of squares below 0.15 are accepted; the optimal fitting interval τ (14–30 days) is chosen via the Akaike Information Criterion. This systematic approach ensures that the classification into exogenous versus endogenous is data‑driven rather than subjective.

The results demonstrate that the blogosphere behaves as an excitable social medium: information spreads, amplifies, and then relaxes in a manner statistically indistinguishable from seismic activity. The authors argue that these parallels are not coincidental but reflect underlying mechanisms of threshold‑driven activation and refractory periods common to many complex systems (neuronal networks, chemical oscillators, traffic flow).

From an applied perspective, the strong correlation between early growth (α_g) and later decay (α_d) for endogenous events suggests that monitoring the initial rise of a keyword can provide early warnings about the eventual magnitude and longevity of public attention. This could be valuable for policymakers, marketers, and crisis managers who need to anticipate the spread of narratives or misinformation.

In conclusion, the paper provides robust empirical evidence that media attention dynamics obey the same scaling laws as earthquakes, with exogenous shocks obeying Omori‑type aftershock decay and both exogenous and endogenous events following Gutenberg‑Richter size distributions. It opens avenues for cross‑disciplinary research linking statistical physics, network science, and communication studies, and hints at the possibility of constructing a “Richter scale” for media events to quantify and perhaps predict societal reactions to information cascades.