The freedom to share and access news online comes with the risk of falling into the trap of so-called fake news, and the development of neural networks and machine learning has only exacerbated the problem of the spread of misinformation. Can AI help distinguish false articles from fake ones, and what tools and applications can be used to detect fake news?
What is fake news and what is its danger?
Fake news is news material that is not true and is widely disseminated through online media and social networks. This type of disinformation can be used to promote certain ideas, create public opinion, and undermine reputation and trust. The massive spread of fake news through entire information campaigns has been recognized as a serious risk factor for democracies and stability in the world. A particularly dangerous feature of fake news is that it often consists of manipulation of real facts, creating phenomena that are partially or completely untrue.
Modern media are increasingly using automated fact-checking systems, but they cannot prevent the spread of fake news in the vast field of social networks and instant messaging apps. It is the ability to quickly share sensational or simply liked news that makes fake news especially dangerous. Today, disinformation is becoming increasingly sophisticated, as modern technologies make it possible not only to create fake news that looks as true as possible, but also to target it to specific audiences. Moreover, such a rapidly developing technology as AI has become a tool with which you can both create and detect false news.
How is fake news created and spread?
An example of a well-known technology that blocks dubious information is a spam filter in your email, thanks to which you are saved from hundreds of intrusive or malicious messages. But surely your email spam filter sometimes fails to cope with some individual phishing emails, and they penetrate your Inbox? Or another example: sometimes users complain that their accounts were blocked by services because AI algorithms suspected something illegal or not in compliance with the Therms of Use in their completely harmless posts. Even the most avant-garde neural network algorithms are not perfect.
The disadvantage of systems using machine learning is that they use control markers which are created by training neural networks with different examples of fake content. Based on such numerous examples, AI determines patterns and signs of false information. However, at the same speed that AI technologies are learning to detect false content, its creators are finding ways to identify and bypass misinformation markers. Additionally, biases can creep into AI systems themselves and lead to erroneous conclusions and results.
Сan I use AI to detect fake news?
Yes, although not completely perfect, such tools exist. Today, AI tools are constantly being developed and improved, which include both text and visual algorithms to identify misinformation. Here we bring you some of the most trustworthy and advanced information verification systems:
The Factual online service operates on the basis of an algorithm that scans and evaluates the reliability of more than 10 thousand news materials on the Internet every day. This tool uses the reputation of sites, the experience of the author of the publication, the quality of sources and even the tone in which articles are written as evaluation criteria. To check an article, you just need to enter the URL into the appropriate field at the top of the site. The Factual is also a digest of the most current news, publishing materials daily from various publications with ratings of their reliability. You can search here for news by topic, for example, USA, World, Business, Culture, etc. It is available as a website, app and Chrome extension.
Full Fact AI
Full Fact AI is a charitable project funded by donations from individuals and companies, including major technology corporations and trusts. It is designed to fact-check statements made by politicians, government agencies and the media, as well as viral content on the Internet. This tool can be especially useful for independent journalists gathering information. To check any publication, you need to submit a fact check request. At the same time, the service does not guarantee verification of each request and reserves the right to work with those requests whose importance is justified by their senders. Therefore, when submitting your request, it is important to indicate what consequences you believe the claim will have if left unverified.
Grover is an artificial intelligence service for detecting fake news, so named because of the underlying computing algorithm of the same name. This project was created by researchers at the University of Washington as an experiment in both creating and detecting fake news. Grover works as a demo prototype for now, but you can already try to check any text for veracity by entering it into the verification field.
Deepfakes, that is, images or videos that fake people, are also a branch of misinformation created with the help of modern technology. Sensity AI service is designed to detect this type of deceptive content. It can detect the widest range of deepfakes and perform identity document verification, facial recognition and age verification. This is a paid service, but you can get a free 7-day trial.
One of the most common cases of massive spread of fake news is disinformation campaigns using bots on social networks and instant messaging applications. The Bot Sentinel tool uses artificial intelligence to identify fake X (formerly Twitter) accounts. You can block trolls and bots based on their Bot Sentinel rating, when the account was created, how many people follow it, keywords in their profile, keywords in their tweets, and more. Bot Sentinel is available as a website, as well as mobile apps for iOS and Android and browser extensions.
No matter how advanced AI detectors are for detecting fake news, it is important to combine human oversight with computer verification. Even the most advanced neural network is not yet able to capture the small nuances and intonations that are accessible to an alert person. Therefore, it is important to check the news you find yourself. Here are some simple but very effective tips for detecting fake news:
- Use only trusted sources of information: find the original source of the news and check if it is real.
- Be wary if you receive a link or screenshot of news articles: make sure the article has a real URL.
- Think about what's behind the generalizations in the headlines: read the entire article to make sure the headline is accurate.