Have you ever tried Googling controversial topics like the 2020 presidential election or the effectiveness of COVID-19 vaccines? Unfortunately, the internet is awash in conspiracy theories and other misinformation related to subjects like these.
Google (GOOG, GOOGL) said on Thursday that it’s trying to cut down on phony content by improving the quality of results in its so-called featured snippets and making it easier for users to find the sources of information they find on Google. To do that, the company says it’s rolling out a new artificial intelligence model called Multitask Unified Model, or MUM.
Featured snippets are the short answers you get to frequently asked questions in your Google search results. Ask a question like “How tall is the Empire State Building?” and Google spits out a quick answer above the list of site results with an answer.
Sometimes, however, those snippets will provide an answer that's technically accurate, but doesn't fit a user’s query. Google Search VP Pandu Nayak says a recent example was when a user asked how long it took light from the Sun to reach Earth, and the featured snippet pulled in data from a site about the solar system and how long it took light from Pluto to reach the Earth.
Google's new artificial intelligence model aims to prevent those snippets from pulling in erroneous information. With this update, Google grabs a consensus answer from sites that demonstrate expertise, authoritativeness, and trustworthiness for a related topic and spits out the featured snippet.
Nayak says an example of a site that demonstrates these qualities would be a health site produced by a medical institution.
What if the top results come from sham sites? Nayak says Google can determine which sites are low quality or use exaggerated titles, spread lies, or preach hate, and won’t use them for featured snippets.
To prevent featured snippets from pulling in information when they shouldn’t, Nayak explained, the company is ensuring the feature can recognize false premises. Nayak used the example of a person searching for the year Snoopy assassinated Lincoln. Previously, a featured snippet would provide the date of Lincoln’s death, but not tell the user that a cartoon dog wasn’t involved.
Now Nayak says a featured snippet will no longer appear when someone searches for something that isn’t true.
“We've trained our systems now to get better at detecting these sorts of false premises and show fewer featured snippets that might feel inaccurate or unhelpful,” Nayak explained. “This improvement has reduced the snippets from appearing for such false premises by about 40%.”
In addition to improving the accuracy of its featured snippets, Google is also expanding its content advisories option. Content advisories are notifications that pop up when someone is searching for a breaking news story or an item that involves a rapidly evolving topic.
In such instances, Google will provide you with a note at the top of the screen saying the results you’re looking for are changing quickly. The idea is to help stop the spread of false information related to major news stories.
Now the company is also adding alerts for when it detects that the only information available on a given topic comes from sites that don’t have reliable information. You can still see the search results, but the idea is to ensure that users don’t just click on a link to a site about, say, UFOs landing in New York, and take it as the truth.
Of course, as with anything that relies on algorithms, there’s no guarantee that Google’s featured snippets or content advisories will be accurate 100% of the time. There’s also no guarantee that you’ll see it every time you search for something that’s inaccurate or quickly evolving.
What’s more, misinformation and disinformation isn’t just spread via Google search; YouTube is also a vector for a good amount of both forms of fake information. Google says that it goes to great lengths to cut down on that kind of content by surfacing high-quality videos, but there are still plenty of instances where misinformation spreads across the video sharing platform.
And with 95% of teens surveyed in a recent Pew Research poll saying they use YouTube, it’s more important than ever for Google to not only crack down on phony information on its search platform, but on YouTube, as well.
More from Dan
Got a tip? Email Daniel Howley at firstname.lastname@example.org. Follow him on Twitter at @DanielHowley.