[A Google fact box singled out
Kannada, a language spoken in the country’s south. The faux pas highlights the
algorithm’s fallibility.]
By Mike Ives and Paul Mozur
It was an odd, unanswerable question. Still, it was on the mind of at least one Google user in India.
What is the country’s “ugliest”
language?
For anyone who typed the question
into the platform’s search bar recently, its algorithm produced a fact box confident
of the answer: a tongue called Kannada, spoken by tens of millions of people in
India’s south.
Informed of that result, many of
them weren’t happy.
Several politicians in the state of
Karnataka, where most Kannada speakers live, went on social media this week to
register their outrage.
“Legal action will be taken
against @Google for
maligning the image of our beautiful language!” Aravind Limbavali, Karnataka’s
forestry minister and a member of Prime Minister Narendra Modi’s political
party, said in a tweet on Wednesday.
Google apologized on Thursday for
“the misunderstanding and hurting any sentiments.” It also deleted the fact box
about Kannada.
But its faux pas — and the response
from Mr. Limbavali and other members of the state’s conservative political
brass — had already been picked up by major Indian news outlets. By Friday, the top results for
the search “What’s the world’s ugliest language?” were articles about Google’s
apology for having answered it.
The episode illustrates the
fallibility of the fact boxes, a function that Google created seven years ago.
The boxes, known as “featured snippets,” contain information that the company’s
algorithms pull from third-party sources. They appear above the links that
usually pop up in Google search results.
The company has said that featured snippets work well, based on usage statistics and
evaluations from people paid to evaluate the quality of its search engine’s
results. But it also admits that they sometimes get the facts wrong — or stray
into the realm of opinion.
“Search isn’t always perfect,”
Google India said in its apology on Thursday. “Sometimes, the way
content is described on the internet can yield surprising results to specific
queries.”
That’s putting it mildly.
Earlier this year, a search for why
Google was banned from China returned a fact box — garnered from a
nationalist state-run tabloid, The Global Times — noting that Google
had left the country of its own accord after deciding that Chinese laws did not
“conform with its so-called democratic values.”
The box made no mention of a
cyberattack that the company had cited as an immediate reason to stop running
its search engine in China. Nor did it mention that most Google services are
widely blocked from China’s internet.
Google is also unreliable on the
question of whether it is a reliable source of information.
The search “Does Google lie to
you?” produces a fact box with this answer: “Google does not give answers (sic)
to questions and therefore it does not lie.”
That is from an article in the newspaper The Australian that quoted
a businessman who accused the company of stealing content and putting it up
directly on its site. The quote was used in the article as a sarcastic
reference to the first result for the search query “Does Google ever lie?”
Kannada, the language that Google’s
fact box said was India’s ugliest, is part of a family of Dravidian languages
that are native to southern India and go back thousands of years.
The snafu this week was not the
first time that Kannada speakers have said that their language was
disrespected.
Karnataka inspired many of the
novels and short stories by RK
Narayan, one of India’s most famous novelists. A popular 1980s television adaptation
of his work was made in Hindi, the country’s most common language, with Kannada
subtitles. Even though Mr. Narayan wrote in English, some critics said the adaptation should have been made in Kannada,
or at least dubbed into it.
“It could very well have been
dubbed when it was made,” the critic Prathibha Nandakumar wrote in 2012. “Why
was that not thought of?”
Google has no fact box for that.
Paul Mozur is a technology
correspondent focused on the intersection of technology and geopolitics in
Asia. He has been twice named a Pulitzer Prize finalist. @paulmozur