The internet's noisy echo chamber of "facts", claims and assertions may soon be able to be verified -- by a squadron of robot fact-checkers.
In a study published in Computational Fact Checking From Knowledge Networks, researchers from Indiana University revealed they have developed an algorithm designed to determine the accuracy of simple facts.
The team, led by Giovanni Luca Ciampaglia, of the Centre for Complex Networks and Systems Research (CNetS), applied the algorithm to answer a range of questions, including matching states with their capitals and Oscar-winning film directors with their movies -- and found the majority of results were highly accurate.
Moreover, the code was also able to search beyond the initial "infoboxes" set up in the tests, finding a broader range of links and correlations between sources.
Ciampaglia said: "If we prevented our fact-checker from traversing multiple nodes on the graph, it performed poorly since it could not discover relevant indirect connections. "But because it's free to explore beyond the information provided in one infobox, our method leverages the power of the full knowledge graph."
Although the fact-checking algorithm isn't the first of its kind to be created, its accuracy means it could pave the way for similar automated systems in the future. However, these robot fact-checkers are by no means foolproof. Like their human counterparts, a fact-checker is only as good as the sources it uses -- and as we know, sites like Wikipedia can get things wrong when it uses secondary sources, such as blog posts and websites, which can offer inaccurate information.
Despite its high success rate, the fact-checking algorithm isn't quite ready to be used outside of the laboratory just yet.
This article was originally published by WIRED UK
