Google wants to solve the misinformation problem of its search engine

In 2014, if you were to Google “King of the United States,” you would have seen a photo of Barack Obama, shouting passionately into a microphone while on a podium.

Advertising

That’s not true, of course. But Google proclaimed it with authority after publishing an article by Breitbart titled “All Hail King Barack Obama, Emperor Of The United States of America!” »

The episode highlights how clumsy Google can be with snippets, search results that the company elevates to the top of the page as a definitive answer to a query. The problem has persisted over the years, with Google clearly stating that some presidents were in the KKK, or that women are bad.

Google on Thursday unveiled a handful of new features aimed at combating such lies on its search engine, one of the most widely used information tools on the planet. Far from its origins as a simple website that listed 10 links as search results, Google is now a sprawling, cluttered site that highlights news, tweets, maps, hotel reservations, and more. As the site has grown – and disinformation peddlers have become more sophisticated – the search engine has become more vulnerable to spreading lies and misinformation.

“In recent years, the growth of misinformation has become an even more pressing challenge for us as a society,” Pandu Nayak, vice president of research at Google, told reporters on Wednesday. “We can only fulfill our mission if we can achieve high quality results. »

Google said it would use its artificial intelligence systems to improve search snippets. The company will use machine learning software, called MUM, or Multitask Unified Model, to verify information from multiple trusted sources that agree on the same facts. The process will allow the system to reach a general consensus, Google said, even if sources don’t phrase the information the same way.

The company is also expanding its “About This Result” feature, originally released last year, to include more context about search results. In addition to seeing a brief description of the website or business and when the result was indexed, people will now see more granular information about the result. For example, it will tell you if a business is owned by another entity. On the other hand, if Google can’t find much information about a result, it will disclose that as well. The company is also launching “About This Result” in more languages, including Spanish, German and Indonesian.

Google also updates its “Content Advisories,” which it typically displays during last-minute news situations, such as a mass shooting or natural disaster, when the situation is changing rapidly and little information is available. . Now, in addition to telling people when information is scarce, it will also notify people when information is available but may not be reliable, based on Google’s ranking system for search results.

The new features underscore the ongoing and growing battle tech giants are waging against disinformation providers. The problem will only get worse as the United States looks to midterm elections later this year. Google, along with Facebook and Twitter, have come under fire for years for exploiting their platforms for conspiracy theories, Covid-19 news and extremism.

None of the updates, however, apply to YouTube, which Google also owns and has long been a major culprit in spreading misinformation. “Their problem is a bit different than ours in search,” Pandu said, noting that YouTube hosts content and uses a custom feed. “We don’t work directly on YouTube, and YouTube doesn’t work directly on us. »