Cold, hard facts rarely dictate what people care about most or what they believe. Instead, it’s the power of a well-told story that truly resonates. Whether it's a heartwarming anecdote, a personal testimony, or a meme that reflects shared cultural narratives, stories have a remarkable way of sticking with us, moving us emotionally, and shaping our beliefs. This very aspect of storytelling can be particularly dangerous when misused. Social media platforms introduce new layers of complexity and amplify these narratives. While artificial intelligence (AI) intensifies the problem, it also serves as a robust defense against such manipulations. Researchers are now harnessing machine learning techniques to analyse content that spreads disinformation.
Difference between disinformation and misinformation
Understanding the difference between disinformation and misinformation is crucial. Misinformation refers to false or inaccurate information, simply getting facts wrong. Disinformation, on the other hand, is intentionally fabricated and disseminated with the specific aim to mislead and manipulate.
As humans, we are wired to process information through narratives. From a young age, we hear and share stories, using them to make sense of the world around us. Narratives not only aid our memory but also evoke emotions, fostering connections that influence how we interpret social and political events. This makes them powerful tools for persuasion—and, by extension, for spreading disinformation.
A compelling story can easily overshadow skepticism and sway opinions more effectively than a barrage of statistics. For instance, a tale about rescuing a sea turtle caught up in plastic pollution often instills greater concern than extensive environmental data.
Usernames, cultural context and narrative time
AI tools can piece together details about the narrator, the timeline they follow, and cultural specifics that ground their story, helping to pinpoint when a narrative doesn’t add up. Narratives extend beyond mere content; they encompass the personas users create to convey them. Even something as simple as a social media handle can offer persuasive cues.
Usernames can be analysed to reveal demographic traits—such as name, gender, and location—along with underlying sentiments and personality traits when embedded in the handle.
For example, a user might select a handle like @JamesBurnsNYT, attempting to pass off as a credible journalist, rather than a more casual option like @JimB_NYC. Both could imply a male user from New York, but one carries more weight in terms of perceived credibility.
Disinformation campaigns often manipulate these perceptions by crafting handles that mimic trustworthy voices or affiliations. Although a handle alone can't confirm authenticity, it significantly contributes to the overall assessment of whether an account is genuine or manufactured to gain trust and blend into a target community.
This nuanced interpretation plays a vital role in detecting disinformation—not just considering what is said but who is saying it and why. Additionally, stories often don't follow a straightforward timeline. A social media thread may kick off with a shocking event, flash back to earlier moments, and skip crucial details in between. While humans can navigate this fragmented storytelling effortlessly, it's a significant challenge for AI to determine the sequence of events based on a narrative.
Methods for timeline extraction can help AI recognise events, understand their order, and map their relationships, even in nonlinear storytelling. Moreover, objects and symbols carry different meanings across cultures. Without cultural awareness, AI risks misinterpreting the narratives it analyses. Malicious actors can exploit these nuances to craft messages that resonate more deeply with specific audiences, enhancing the persuasive power of disinformation.
For example, the sentence “The woman in the white dress was filled with joy” evokes a happy image in a Western context. However, in parts of Asia, where white symbolises mourning, it might come across as unsettling or even offensive. To effectively detect disinformation that weaponises such symbols and storytelling techniques, AI must be culturally literate. Research has shown that training AI on diverse cultural narratives boosts its sensitivity to these distinctions.
Who stands to gain from narrative-aware AI?
Narrative-aware AI tools play a crucial role in helping intelligence analysts quickly spot orchestrated influence campaigns or emotionally charged narratives that are spreading at an alarming rate. These analysts can leverage AI to sift through large volumes of social media posts, allowing them to map out persuasive story arcs, identify similar narratives, and flag the coordinated timing of social media activity. This enables intelligence agencies to implement countermeasures in real time.
Moreover, crisis-response agencies can swiftly detect harmful narratives, such as false emergency claims during natural disasters. Social media platforms can also benefit from these tools by efficiently routing high-risk content for human review while avoiding unnecessary censorship.
Researchers and educators can gain from such technologies too, as they enhance the tracking of story evolution across different communities, thus making narrative analysis more rigorous and easily shareable.
Even ordinary users stand to benefit from these advancements. AI tools can identify potential disinformation in real time, prompting readers to approach suspicious stories with skepticism and helping to counteract falsehoods before they take hold.
As AI increasingly takes on the task of monitoring and interpreting online content, its ability to understand storytelling extends far beyond traditional semantic analysis, making it essential for our current digital landscape. The Cognition, Narrative, and Culture Lab at Florida International University is actively developing AI tools designed to detect disinformation campaigns that utilize narrative persuasion techniques.
ALSO READ: Stolen smartphones are coming back to owners via courier: Here's how
Inputs from PTI