We ran our own Microsoft Bing censorship test. This is what we found.

We ran our own Microsoft Bing censorship test. This is what we found.

Last week Greatfire.org, a website that tracks web censorship in China, issued a report claiming that Microsoft Bing was deliberately censoring its Chinese-language search results abroad.

The report generated a wave of media attention. Microsoft issued a statement denying the accusations, adding that signs of censorship should be attributed to technical glitches. Greatfire fought back with two additional pieces.

In an effort gauge the validity of Greatfire.org’s claims, TIA searched ten controversial keywords using Bing’s China version, Bing’s Taiwan version, and Bing’s US version. We also toggled between simplified and traditional Chinese characters (the former are used in mainland China, the latter are used in Taiwan and Hong Kong) because we were curious to see if different character inputs would yield different results.

Before we proceed, we’ll note that it’s difficult to make decisive claims over Bing’s censorship without knowing detailed specifics about the following: Bing’s algorithm within China, how it differs from its international algorithms; and how those algorithms differ from Baidu and Google’s algorithms – AND, how those algorithms interact with the search engines’ censorship mechanisms. Indeed, this gets knotty quite quickly.

For the sake of clarity, we’re going to use some special vocabulary. Hits that direct users towards content that would ordinarily be accessible in China will be described as “harmonious.” Hits that direct users towards content that would be blocked in China will be described as “unharmonious.”

Here’s our thesis in a nutshell: Search results for controversial keywords in simplified Chinese on the US bing servers generate lots of harmonious results. But that doesn’t mean Bing is censoring its search results outside China.

We searched ten key terms and discovered a mixed bag of results. But we’re going to focus on two key search terms here that other media organizations haven’t looked at (to our knowledge). The findings here capture the essence of results for other terms.

Chen Guangcheng, the exiled human rights lawyer

From mainland China, all searches for Chen Guangcheng direct users towards harmonious content – primarily news articles that take a critical view (or condescendingly sympathetic) of Chen’s work.

From Taiwan, simplified Chinese searches for Chen Guangcheng on Bing yield the following results:

  • Simplified chinese Wikipedia page for Chen Guangcheng (blocked in the mainland; unharmonious)

  • Baike (Baidu’s wiki) entry for Chen Guangcheng (harmonious)

  • Links to videos of mainland broadcast news reports on Chen (some from Youku, some from Youtube, which is blocked in China, but all likely harmonious since they’re from mainland media)

  • Link to Chinese-language Financial Times piece (not necessarily harmonious)

Okay, not bad. We get some harmonious results (Wikipedia and FT), and some unharmonious results (state media reports). Christian Bale isn’t showing up, but there’s no need to cry censorship just yet.

Meanwhile, a search for Chen Guangcheng in traditional Chinese yields the following results:

  • Two hits from Taiwan’s Wikipedia (unharmonious),

  • Various news articles from Taiwan news outlets, including pro-DPP paper Liberty Times (also unharmonious).

  • Curiously, a notice at the bottom of the results page reads “some search results have been removed.”

Huh? If a piece from DPP mouthpiece paper Liberty Times – probably the least harmonious newspaper in East Asia – ranks among the top ten hits, what else could there possibly be to remove? There’s good reason to believe that this notice is indeed a technical glitch.

When searching for Chen Guangcheng in simplified Chinese on a US server yields the following results:

  • Simplified Chinese Wikipedia (blocked in the mainland, unharmonious),

  • Baike entry (harmonious), followed by mainland Chinese news articles (harmonious).

  • Various news pieces from the mainland (likely harmonious)

Okay, so aside from Wikipedia, the CCP should be happy here. Let’s try another search term.

Taiwan’s pro-independence political party

Searching for the Taiwan Democratic Progressive Party (台湾民进党, the largely pro-independence political party which the PRC government does not recognize) in mainland China yields a “no data received” page. In other words, “Run along now, kids, nothing to see here.”

When when one searches “Taiwan DPP” in traditional characters (台灣民進黨) in Taiwan, the results are pretty much what one would expect:

  • Links to the DPP homepage (unharmonious)

  • Links to various domestic newspaper articles (some harmonious, some unharmonious)

  • Links to various support organizations (unharmonious)

But search for Taiwan DPP in simplified characters (台湾民进党) and the results are totally different:

  • Link to the DPP homepage (unharmonious).

  • Link to Baike entry (harmonious).

  • The remaining hits are mainland Chinese message board discussions that are extremely critical of the DPP (harmonious). There’s also a link to a mainland Chinese news piece reporting a traffic crisis in Kaohsiung, the unofficial center of DPP activity outside of Taipei (harmonious).

So even searching for the DPP within Taiwan generates harmonious results if one uses simplified characters. What happens if we move to US servers?

Searching for Taiwan DPP in simplified characters in the US yields:

  • 3 hits from Baike (harmonious)

  • An article from BBC China (not necessarily harmonious)

  • 4 hits from Mainland Chinese media outlets (harmonious)

Searching for Taiwan DPP in traditional characters in the US yields:

  • Link to the DPP homepage (not harmonious)

  • Link to Taiwan Wikipedia article (not harmonious)

  • Link to a website for overseas Taiwanese

  • Links to several articles that are critical of the DPP, but not necessarily unfair (by Taiwan media standards)

Conclusion

Censored search results from within China are not the same as results from outside China. At times, Bing will generate results that are indeed consistent with the CCP’s official views on controversial subjects. But sometimes these results will be lumped together with web pages that would undoubtedly be blocked in China.

Meanwhile, alternating between simplified and traditional characters appears to generate different search results, which could push harmonious content either higher up the rankings or further down the rankings, depending on the subject.

Is Bing favoring harmonious results over unharmonious results? Again, without knowing the specifics of its algorithm, it’s hard to say. When it comes to webpages on unharmonious topics, the vast majority of information that’s accessible to Chinese internet users is government propaganda. So that’s naturally where most traffic flows among Chinese users. And by extension, it’s understandable that government propaganda would rank highly in simplified Chinese search results.

In response to this argument, Greatfire.org writes:

But so many foreign media web sites are publishing information in Chinese (including VOA Chinese, FT Chinese, RFA Chinese and independent Tibetan sites publishing in Chinese) that it is hard to believe that they do not appear in search results.

“Hard to believe?” Really? The number of these so-called “foreign media web sites” ought to pale in comparison to the number of domestic Chinese media propaganda machines, especially given Chinese media outlets tendency to re-post content.

It’s also not fair to note the difference between the international Chinese Bing and the international Chinese Google because there’s pretty much no such thing as a domestic Chinese Google. All queries are redirected to the Hong Kong site. Meanwhile, with a market share of just about 1.6 percent, and crippled by maddeningly long load times, the only reason to use Google in mainland China is if one wants to get frustrated. Google users in Hong Kong, meanwhile, can read about Liu Xiaobo as much as they want. So it’s no surprise that Google’s “international” Chinese-language search results will draw up pieces from the New York Times. Its results might be tilted towards the unharmonious as a result of its widespread usage in Hong Kong and elsewhere.

Meanwhile, the Chinese government is likely more concerned with monitoring the web tools that Chinese people actually use. Bing currently has a 0.3 percent market share in China. With the exception of Confucius Institutes, the Chinese government doesn’t have a strong legacy of pushing censorship and propaganda beyond its borders (excluding Hong Kong and Taiwan).

The strong representation of harmonious results on the US version of Bing would understandably raise eyebrows. But allegations that Microsoft is kowtowing to the Chinese government ought to be treated with the same skepticism.

(Editing by Paul Bischoff)


The post We ran our own Microsoft Bing censorship test. This is what we found. appeared first on Tech in Asia.