Tracking disinformation? These AI tools can help.

par Priyal Shah
26 juil 2023 dans Combating Mis- and Disinformation
The letters AI on a black background

This article was originally published by The Fix and is republished here with permission. Learn about the latest from the world of European media by signing up for their newsletter.


Detecting disinformation is a key skill to have for journalists today. Fact-checking remains a human-led endeavor, but the massive volume of disinformation cannot be tackled by manual capacity alone. Here, AI-powered tools can help.

University of Bergen’s Laurence Dierickx, a researcher on AI-driven journalism and fact-checking, says that “several projects are ongoing to provide more AI layers to help fact-checkers speed up a time-consuming process – an inherent characteristic of fact-checking; it requires much time.”

The Fix gathered a range of tools that will help journalists and fact-checkers identify and address disinformation. These tools are categorized into four main categories – tools for text, images, videos, and detecting bot activity. 

AI tools for written content  

Exorde network 

Exorde is a global information processing platform that analyses the spread and virality of information pieces on the internet. The way it works is by spotting URL submissions which is textual content marked to be of interest to Exorde. It then processes the “sentiment” of the claim and what people’s reaction to that claim is globally. In the final stage, Exorde archives this information for public accessibility. 

FactCheck.org 

FactCheck.org is a project hosted by the Annenberg Public Policy Center at the University of Pennsylvania. The center verifies claims made by American politicians at local, state and federal levels. This includes political and scientific remarks. With Ask FactCheck and Ask SciCheck, journalists can get their information fact-checked by the team. 

Factiverse

Factiverse is an AI tool that journalists can leverage to speed up the fact-checking process. Based on the research produced at the University of Stavanger, Factiverse has produced its AI Editor that can spot disinformation in content and point towards credible sources. Along with this, they have a fact-checking database named FactiSearch and a context-providing and jargon-explaining tool called Microfacts.  

ClaimReview

ClaimReview is a tagging system created in partnership with Google that helps identify and promote fact-checked stories to appear prominently on search engines and social media sites. With this tag, the fact-checked statement is presented alongside a note that indicates its validity. 

AI tools for images

Dierickx’s research found that many fact-checkers underestimated the potential of generative AI in creating false audiovisual content. Highly reliable tools of reverse searching would not produce satisfactory results when it comes to AI-generated content. She adds, “The quality of generated images and video is also increasing, and detection of this kind of content will become increasingly difficult.” Here are some tools that can help journalists get to the truth.

EXIF checking tools

Exchangeable image file (EXIF) is a standard format that accompanies all images taken on a digital camera. EXIF annotation carries information about the image like ISO, shutter speed, and exposure alongside the date and place where the image was taken. Any modifications to the original photo are recorded in the image’s metadata. Journalists can benefit by checking the EXIF of any photo to determine whether it has been tampered with or not. Web-based tools for extracting EXIF are EXIF Data Viewer and Metadata2go. A computer application for the same is ExifTool by Phil Harvey.

TinEye

Another tool to help verify images is TinEye. Journalists can search for duplicates of the image, track where the image has appeared before and use the reverse searching technique. 

AI tools for videos 

Verification plugin – WeVerify 

The latest version of InVID-WeVerify verification plugin hosts a variety of tools to fact-check visual content. Along with their new CheckGIF feature, they have tools like Image content analysis, Optical character recognition in Images and the ability to compare content against the Database of Known Fakes. 

The Glorious ContextuBot | Shared Nation

The Glorious ContextuBot is an early-stage tool that helps detect the original source of audio files. With its fingerprinting technology, it aims to find not only the original source of a sound bite but also the context surrounding it. The tool also helps locate where else the audio file was used. 

Truepic

With Truepic, journalists can validate visual content for authenticity. Truepic Display provides the source and history of any media file whereas Truepic Lens provides the metadata of any file.  

YouTube DataViewer 

Amnesty International’s YouTube DataViewer helps extract the metadata of any video uploaded on YouTube. Users can also find the video’s thumbnail and reverse search that image.

AI tools for checking bot activity

Botometer® by OSoMe

Botometer is a project created by the Observatory on Social Media (OSoMe) at Indiana University. This tool provides scores for Twitter accounts, with a low score (0) denoting a human account and a high score (5) meaning a bot account. Further, it distinguishes the type of bot accounts, categorizing the accounts into: Echo-chamber, Fake follower, Financial, Self declared, Spammer, and Other..

Hoaxy 

Hoaxy tracks the spread of information shared by low-quality human-generated websites. With this tool, journalists can track the virtual path articles have taken. It also provides a score for the likelihood of bot-tech and automation that assisted in the spread of those articles. 

Bonus – five more tools 

Media Bias/Fact Check

MBFC is a source to verify fact-checkers and news websites. It rates websites and media organizations on their bias, political affiliation and authenticity of fact-checked information. They have provided detailed reports on the credibility and truth-providing capacity of more than 6300 media sources. 

Hamilton 2.0 Dashboard

Created by the Alliance for Securing Democracy at the German Marshall Fund of the USA, the Hamilton 2.0 dashboard specializes in highlighting the media messages and narratives promoted by the Russian, Chinese and Iranian governments. With their ASD Social Data Search, journalists get greater control over finding Tweets relevant to them for the same. 

The Factual

The Factual algorithm rates news articles for four metrics – site quality, author’s expertise, quality of the sources and the tone of the article. Based on these parameters a grade is given between 1 and 100%, with 50% and above denoting an informative article and below 50% denoting a non-informative article. 

Climate Feedback 

Climate Feedback is a dedicated platform for debunking false information related to climate change. The corrected information is supplemented by a network of global scientists. The website hosts explanations by scientists given for false information found in news articles and claims made on the internet. 

Tank Classifier

Tank Classifier is another good example of an AI-based tool that can help journalists debunk false information. Dierickx says this tool “allow(s) fact-checkers to detect if a tank comes from one country or from another, in the context of the war.” The tool can currently work on classifying only 20 types of military vehicles. 


Photo by Markus Spiske on Unsplash.