...My heart's In Accra
Ethan Zuckerman’s online home, since 2003

WeChat: Learning from the Chinese Internet

Four year ago, Hal Roberts and I were researching internet censorship by studying the use of proxy servers around the world. Proxy servers are often used to circumvent internet filtering, letting users access content that’s otherwise blocked by a national government, an internet service provider, or a school or business. We found that the usage of these tools, including both free, ad-supported tools and subscription-based VPN services, was surprisingly small: less than 3% of all internet users in countries that aggressively censored the internet.

This surprised us, as there’s been a great deal of dialog in the US about “internet freedom”, a US government policy that supports providing uncensored internet access to people in China, Iran and other nations whose governments filter the internet. Our research suggested that the efforts already underway by projects like Tor and Freegate weren’t being used nearly as much as their authors hoped. This might be due to weaknesses in the tools, but we suspected another motivation: lack of user interest. Hal designed a study of bloggers in countries that censored the internet, asking whether they used circumvention tools. Almost all were aware of the existence of proxy servers, but few used them regularly. When we asked why they weren’t using these tools, most told us that they weren’t very interested in accessing blocked content.

A new paper from Harsh Taneja and Angela Xiao Wu at Northwestern University offers some insights into why this is the case. Taneja and Wu use internet usage data from Comscore to analyze the thousand most globally popular websites, examining their overlapping audiences. Audiences for websites tend to cluster together – people who visit CNN.com are more likely to also visit ESPN.com than people who visit Chinese search engine Baidu. The authors identify a set of 37 “culturally defined markets”, collections of popular websites that seem to be visited predominantly by people who share languages or cultures. Language, while a powerful factor in explaining this clustering, isn’t the sole factor: a cluster of French and Arabic-language sites represents a bilingual North African culture, for example, and an Indian cluster containing mostly English-language sites is distinct from an English-language North American/UK/Australian cluster. And there’s a cluster of football sites that appears to have a truly transnational audience.Read More »WeChat: Learning from the Chinese Internet