Cecilia Vazquez, from the chequeado.com team, reflected on the current challenges to dismantle fake content.
Disinformation is one of the main problems in the field of communication today, a situation that has been deepened by the emergence of artificial intelligence. In this context, Cecilia Vazquez, a specialist from the chequeado.com team, was invited by Marcela Rosales, a professor of the Writing I course of the Bachelor's Degree in Social Communication, to give a talk at the Faculty of Political Science and International Relations of the Universidad Nacional de Rosario, addressing the relationship between disinformation and artificial intelligence (AI).
Cecilia, a graduate in Education Sciences, has been part of the education team at chequeado.com for more than 8 years, a non-profit organization that has been working since 2010 to improve public debate and fight misinformation.
Verification journalism is becoming an emerging branch of communication, and it already has a very particular narrative and style. With the emergence of social media in our daily lives, our consumption of information has turned to them, making it the perfect space for false news or information to slip through.
When defining what disinformation is, the specialist explained that “it is content in which facts that did not happen are invented, or are published out of context or are distorted to change their meaning.” This can be presented in different formats, for example: audio, video, image, text, and circulate on social networks, media, messaging systems, among others.
Vazquez stressed that it is correct to talk about disinformation and not fake news because disinformation is often not news, but false content that goes viral, and these come in different formats. “That is the challenge, disinformation always appeals to emotions, beliefs, or some fact of reality so that we believe in that false information. They can even be images out of context, it happens a lot that they go viral, for example, photos of marches that happened even in other countries but that are attributed to another event that has nothing to do with it. The challenge of formats is a whole issue, it is not the same to check an image, a post on a social network, or an audio.”
In general, misinformation is not black or white, but there are shades of grey. Sometimes the information is true but the context is false, or on the contrary, the context is true but the information is false. Social networks are managed with the logic of immediacy, which is why information is fragmented in order to attract the user's attention. This often means that, in the effort to explain everything in a short, attractive and eye-catching headline, reality ends up being distorted. The goal is to generate clickbait to get more visits, but the intention is not always to focus on misinformation, but rather it ends up being a consequence of implementing a communication strategy.
In this way, he stated that it is impossible to check everything, but that at Chequeados they developed a method that helps them identify those contents that can be verified. “We have selection criteria, related to virality and the potential damage that this possible misinformation can cause. Misinformation increases its circulation in vulnerable times, such as electoral periods or the pandemic. Also when there is a topic that is anchored to our beliefs and our biases. Many times it is believed that thought is constructed by what is consumed, but in reality one goes looking for what one really believes to confirm one's assumptions,” he indicated.
In recent years, artificial intelligence has been incorporated into this, which represents a very big challenge when it comes to corroborating whether content is authentic or has been modified. In this sense, the specialist explained that although false content can be created with artificial intelligence, which then goes viral mainly on social networks, there are also tools that help detect them. “There is currently an ethical concern about who made a note or an illustration, whether it is a person or was it directly an AI.”
Similarly, Vazquez described that there are tools to check the veracity of content. “There are programs to check if specific content is produced with AI, in any format: text, audio, image or video (which are usually called deep fakes). However, audios are the most difficult to identify and create a real complication.”
There are online text verifiers that can check to see how much of the text could have been created with AI. As for images, although there are also verification applications, there are certain key factors that can be taken into account when checking the essence of the image: such as details, the context of the image, or the watermark. “There are elements that AI cannot yet recreate perfectly, such as hands or skin textures. Although they are complex and have a very fast level of creation, they are still in development and have errors. Another thing to keep in mind is that every image created by AI has a watermark, although many times it is so small that we cannot see it, it is there. That is also an indicator.”
On the other hand, the specialist also referred to deep fakes, which are much more complex to create. “This type of content generated with AI is false but often very realistic. A common example is videos where face swapping is performed. However, you have to pay attention to inconsistencies in the image and audio of the video. You often see errors in dubbing, random shadows and misalignments in parts of the face, such as the eyes or mouth. These are clear signs of manipulation.”
The member of Chequeado stressed that it is very important to develop one's own critical thinking and that before sharing something that we are not sure about, we must corroborate the sources. “When we doubt information, we must look for some data in that content that can be verified to verify whether it is true or not. It often happens to us that we read a post in which a specialist recommends something, and in reality that person or the institution to which he supposedly belongs does not exist, so from the start we are at least faced with misleading content that may later be totally false. In the Google image search engine we can track if an image has already been published and who did it or if that content was ever verified. Another recommendation is that when in doubt, do not share that information because it can generate much more viralization.”
Can AI be poisoned by false information?
Artificial intelligence has become a powerful and transformative tool that has come to revolutionize and reconfigure our ways of doing things, significantly influencing not only our daily practices, but also the processes and approaches in multiple areas of knowledge and production.
Like any new tool, it is adored by some people and demonized by others. The arguments for and against are varied, and it is one of the current debates worldwide. For Cecilia Vazquez, the key is not to think of it in a black or white duality, but to position oneself in the middle. “When the web appeared, there were also detractors, however, now most people have incorporated searching for something on Google or Wikipedia. I think the important thing is to understand that it is already among us and that does not mean that it is always good to use it for everything or to go to the other extreme of thinking that it is bad.”
For the professional, there are two key criteria regarding AI: understanding how to use it and how it works. “You have to be able to understand this in order to put it into practice in a way that is useful to us and contributes to us, in a responsible way.”
However, not even these complex systems are immune to being influenced by misinformation. “Unfortunately, no artificial intelligence has any system that prevents them from being fed false content. False information can contaminate the data extracted from the Internet to train the AI systems that we usually use, such as GPT Chat, Gemini, Copilot, etc. This is an issue that worries us and is linked to the possibility of generating more false information,” Vazquez certified, adding: “There is no filter, and that is why it is so important that we can develop our own critical thinking. We have the challenge of not staying only with what a certain artificial intelligence tells us and being able to search multiple sources, doing manual research.”
Journalist: Gonzalo J. García