How to spot an algorithm’s bias
An algorithm can be a powerful tool in the quest to understand what’s happening in the world, but sometimes it can be used to manipulate its own results.
That’s exactly what researchers at Google have discovered in a study that found the search giant’s algorithm has a tendency to favour certain news sources over others.
The researchers, led by John Boulton, a computational biologist, found that in some cases, Google’s algorithms favour certain articles over others, when they’re presented as a question mark next to articles they already favour.
This can be seen in an example of how Google’s algorithm can skew the results of the BBC article titled “Google’s AI uses its expertise to create fake news”, in which a question is asked on the BBC website.
The algorithm, called GEOfavor, is meant to tell people whether or not the articles are relevant.
But it can also be used as a way to find articles that it’s more likely to favour over others in a search result, the researchers said.
The Google algorithm does this by making a “prediction” of the article’s relevance.
The algorithm then uses this information to make its own decisions, such as which articles to show and which to omit.
Boulton and colleagues created a test case to investigate how Google would use this to favour news articles over articles that were not relevant.
In the test case, the algorithm presented two questions that asked whether the article was about a specific country or a particular type of country.
One of these questions, in this case, was about the number of women in the UK.
The other question was a question about the cost of living in England.
The question that was most relevant to Google was “how many women live in England?”, and Google’s results were very similar to the ones it presented.
The only difference was that Google removed a question that said whether or how many women were in the English national football team.
But the question about “cost of living” was more relevant to the Google search engine.
Google showed the results for the question “how much would you pay for a house in England?” in a question marker next to a question asking whether the price was higher than $1,500.
When the test subject saw that question in the question mark, the answer Google presented was more accurate than the other questions, suggesting that Google was choosing the more relevant question over the more irrelevant one.
This led to the conclusion that Google’s “predictions” about the answer were biased.
“In the past, Google has been criticised for bias in the selection of articles for its search results,” the researchers wrote.
Google’s algorithm, they added, is capable of detecting biases in its results.
“While there are still areas where Google’s biases might be present, the nature of its algorithms and the way in which they operate suggest that these biases are very likely to be low, if not zero, for a given question,” they said.