The chilling effect refers to a phenomenon where individuals or groups refrain from engaging in expression for fear of running afoul of a law or regulation. The Chilling Effect may be caused by legal actions such as the passing of a law, the decision of a court, or the threat of a lawsuit or any such legal action that would cause people to hesitate to exercise a legitimate right for fear of legal repercussions.
Edward Snowden is an ex-agent of the CIA who disclosed classified documents showing evidence of mass online surveillance from the United States. This includes revelations of collaborations with companies (through the PRISM program) or with other states (the Five Eyes countries), in order to accumulate data on every citizen.
Allegedly, these programs are meant to fight terrorist threats and improve global security, but it can be argued that these surveillance techniques go too far and violate people’s privacy.
Penney demonstrated a chilling effect after the Snowden revelations by studying the 2012-2014 period. The following analysis is based on the same methodology, but extends it by using a longer time period and by taking into account external factors such as the growth of the Internet usage, while also considering other events that may have caused a chilling effect.
Penney, Jonathon, Chilling Effects: Online Surveillance and Wikipedia Use (2016). Berkeley Technology Law Journal, Vol. 31, No. 1, p. 117, 2016.
This analysis is based on the English Wikipedia traffic of terrorism-related articles. These articles were chosen because they may be considered privacy-sensitive by a majority of people, which means that someone would be less inclined to visit them under the assumption that this action is being watched.
In order to detect long-lasting trends, the data spans a wide range of time, from 2008 to 2018.
A set of security-related articles was also selected to serve as a control group. We make the assumption that these articles should not show a chilling effect, because they are not privacy-sensitive.
One way to verify the chilling effect hypothesis would be to observe the views’ pattern before and after the event. Here, we address one of the article’s limitations, by observing more data : we have 5 years before and after the revelations. A long term drop in the views would be an interesting finding, and it can be interpreted as a persistent, lasting chilling effect.
The sudden drop in views and trend shift from increasing monthly views over time to decreasing after June 2013 is consistent with a significant and long-term chilling effect.If we look at the security articles, the drop-off is not significant. However, there still seems to be a slight change in the slope. Instead of steadily growing, the data shows a sudden peak followed by lower values than in the first segment. It means that part of the slope variation for the terrorism articles may be explained by other factors, but the result is still significant enought to indicate a chilling effect.
After having disclosed classified documents to Wikileaks, Chelsea Manning faced a sentence of 35 years of imprisonment after his trial.
The leak of the documents Vault 7 sheds light on an impressive amount of cyber weapons developed or detained by the CIA, that can be used for targeted or mass surveillance.
The Cambridge Analytica scandal exposed the mass data collection of millions of Facebook users. This data was used to make targeted advertisement campaigns for the 2016 US elections, by creating psychological profiles of voters.
This event is very similar to the Snowden incident, and such an extreme prison sentence can have a very dissuasive effect on the population. Since it is so close in time with the Snowden revelations, it may have contributed to the chilling effect shown above but it is hard to isolate it.
The nature of this incident makes it quite similar to the Snowden incident, but a potential chilling effect may be diminished since the documents were released in 24 parts (from March to September 2017) and not in a single batch. It turns out that unlike the Snowden event, the data does not show any sign of chilling effect at this date.
Even if this incident is not directly related to government surveillance and terrorism, it may be able to revive and accentuate privacy concerns in the population. This could either decrease the feeling of security and anonymity online, or serve as a reminder of the Snowden scandal. However, it seems that the traffic of terrorism-related articles was not affected by this event.
So far we have seen a persistent Chilling effect, especially after the Snowden revelations. However is this the only factor that governs the users behavior on the internet? What may be the other factors that affected this change?
For example, the internet has been one of our most transformative and fast-growing technologies. Globally the number of Internet users increased from only 413 million in 2000 to over 4.8 billion in mid 2020.
In the following section we will analyse the overall trend using
this dataset containg number of users in the Internet.
To compensate for the effect of the growing number of Internet users, we can plot the Wikipedia article views per a million Internet users. For this we have used the Internet users per month calculated from the regression done above.
This tells a different story than the conclusion we made above. Now the division between the two categories are not so significant. In other words, when we take the number of Internet users in the world, the total traffic for these terrorism related articles have a negative trend.
Interestingly, the security articles seem to be less impacted by the growing number of Internet users.
The first results are a very interesting finding : since the Snowden revelations, the trend of monthly number of views for terrorism-related articles shifted, from increasing to decreasing. We have here demonstrated that the chilling-effect suggested in Penney's article would be persistent, even five years after the revelations.
None of the other events that we considered here seem to have caused another chilling-effect, even though the Chelsea Manning trial may have contributed to the 2013's trend-shift.
Interestingly, when taking into account the growth of Internet users by plotting the monthly number of views per million of Internet users, it kept decreasing even before the revelations. Does that mean that there is no chilling-effect ? That's not an easy question to answer. First of all, most of the new Internet users come from developing countries. They may not have any interest, or even be aware of Snowden's revelations. Secondly, we have only considered here the articles in the English version of Wikipedia, and there's a big chance that the new users prefer to browse the Internet in their native language, or that they don't even speak English.
