This week, Conservative leader Andrew Scheer tweeted a screenshot of some curious Google search results. The search for "Canadian soldiers" gave a photo to former Guantánamo Bay prisoner Omar Khadr, who was accused of killing a US soldier in 2002.
Scheer asked Google to act, and the other user does not need a long time to recommend that everything is the work of the Russian troll.
Omar Khadr is a sentenced terrorist who killed the medicine and pardoned another. He is not a victim, and in this way he should not be depicted with real Canadian heroes. @googlecanada: Fix this. pic.twitter.com/qywUGQihVb
Actually? It is much more commonplace.
However, the episode is another reminder of how many algorithms with the best intentions can unintentionally promote the spread of disinformation online. And in just a few months, the Canadian federal elections are even higher when politics are involved.
How did Khadr get to?
The name of Khadr appeared, which Google called the results of his knowledge schedule. Sometimes they appear above Google's usual search engine results if a user asks a question, searches for general knowledge, or searches for a well-known place or public personality.
The Knowledge Graph pulls your data from a variety of sources – one of which is Wikidata, an open information repository maintained by the same organization as Wikipedia. Think of Wikipedia, such as a ready-made report, and Wikidata's raw data used to write it. Like Wikipedia, everyone can contribute to Wikidata's better and worse situation.
Twitter user Stephen Punwasi pointed out it seems that the timetable of data used to place Omar Khadr among Canadian soldiers seems to have been pulled out of the Omar Khadr Wikidata page – and the one who did it was a "Russian troll".
The Google Knowledge Graph collects data from a variety of sources, so I checked the changes that Omar Kadh put under the "Canadian soldier".
Changes from the Russian account a month ago. ?♂️
So it was the work of the Russian troll?
It doesn't seem like that.
The Changes to the Khadr Wikidata page were made by user Ghuron. The user seems to be an active site contributor and, according to their Github account, lives in St. Petersburg, Russia.
But Ghuron does not seem to focus on data related to a particular person, ideology, country, or political topic. Rather, it resembles an automated cleaning job designed to improve the quality of Wikidata at a speed that is much faster than any person who could do it by hand.
In accordance with discussions between Ghuron and other Wikidata members, Ghuron runs a script that uses machine learning to automatically add and modify large volumes of Wikidata data (such as a person's occupation). Essentially it is meant to put the data in the buckets.
His script ensures that Street Fighter is properly classified as a video game, the Faroe Islands are exacerbated under the largest "island" category, or the right Renaissance artists are properly classified as painters. Just see for yourself.
From time to time, his script seems to be getting errors too – and other Wikidata users have not been shy to let him know.
Did it happen to Khadr?
Yes! Using Khadr's The Wikidata page edits history as a guide, here's a short timetable that's even further than Scheer's #:
July 26, 2018, Ghuron script categorizes Omar Khadr's work in Wikidata as a "soldier" – part of a larger, automated effort to assign professions to anyone from the Zodiac Killer to the Danish priests.
September 24, 2018 users start publishing the discussion section of the Omar Khadr Wikipedia page asking why his name describes him as a Canadian soldier in Google search results. However, the phrase "Canadian soldier" has never appeared on his Wikipedia page, which means that the phrase may have been pulled out of his Wikidata page. Google has not yet explicitly approved it.
Later that day data is removed from Khadr Wikidata page. In the discussion section of the Khadr Wikipedia page, the user wrote: "It was a Google error related only to their search engine and it was not processed by Wikipedia. Google tonight eliminated the error and Omar Khadr is no longer displayed as a Canadian soldier."
On September 30, 2018, the Ghuron script classifies Omar Khadr's activity as a "warrior" in Wikidata again. It is not clear whether Google's knowledge schedule this time ignored the changes.
Either way, on December 8, 2018, Ghuron script then categorizes Khadr's military rank in Wikidata as a "warrior" – data that is so different that it is likely to return to the results of the knowledge graph. Before long, people once again saw Khadr between "Canadian soldiers" results.
What did Google do?
Danny Sullivan, a wardrobe that Google is a search engine ombudsman replied to Canadaland's Jesse Brown on Twitter.
I'm not familiar with who says what. As has already been said, we have just seen that the concerns are widespread. We reviewed, and because it was a problem with the knowledge schedule, we acted here (we do not search for search records). And yes, last, because it's accurate.
"We reviewed, and because it was a question of the knowledge timetable, we acted here," wrote Sullivan.
It doesn't seem that the Khadr Wikidata page has changed – only the knowledge graph processes the data in it.
CBC News has reached Google and will update this story if we hear more.
Is it normal?
As Sullivan also points to Twitter, Google does not change search results – at least, unless it is forced to remove information from its index. Rather, Google changes the results of its knowledge schedule.
Whether this difference is obvious to most users, given its placement, especially in situations of high political tension, is less clear.