If a Gen Alpha tween said, “Let him cook,” would you know what that meant? No? AI doesn’t either.
A research paper written by soon-to-be ninth grader Manisha Mehta was presented this week at the ACM Conference on Fairness, Accountability, and Transparency in Athens. The paper details how four leading AI models—GPT-4, Claude, Gemini, and Llama 3—all struggled to fully understand slang from Gen Alpha, defined as those born between 2010 and 2024.
Mehta, along with 24 of her friends (ranging in age from 11 to 14), created a dataset of 100 Gen Alpha phrases. These included expressions that can mean totally different things depending on context—for example: “Fr fr let him cook” (encouraging) and “Let him cook lmaoo” (mocking).
According to the researchers, the LLMs had trouble discerning the difference. In particular, AI struggled with identifying “masked harassment,” which is concerning given the increasing reliance on AI-powered content moderation systems.
“The findings highlight an urgent need for improved AI safety systems to better protect young users, especially given Gen Alpha’s tendency to avoid seeking help due to perceived adult incomprehension of their digital world,” the study reads.
It wasn’t just the AI models that performed poorly; parents didn’t do much better. The parent group scored 68% in basic understanding of Gen Alpha slang, nearly identical to the top-performing LLM, Claude (68.1%). While the LLMs did slightly better at identifying content and safety risks in the language, only Gen Alpha members themselves scored highly in understanding the slang, its context, and potential risks.
It’s nothing new for young people to feel misunderstood by their parents, but now the gap is widening. Members of Gen Alpha, born post-iPhone and known as the iPad generation, have grown up online. Their native language, often sourced from online spaces (most notably gaming), evolves so quickly that what’s popular today may disappear within a month.
Mehta’s research shows that parents—and even professional moderators—are likely to miss context shifts in comment sections. For example: “OMGG you ate that up fr,” versus “You ate that up ig [skull].” The implications of the study suggest that parents might recognize only a third of the times their child is being bullied in comments, even if they’re closely monitoring their online activity.
Simply put, the systems meant to keep kids safe online don’t speak their language.
Melden Sie sich an, um einen Kommentar hinzuzufügen
Andere Beiträge in dieser Gruppe
Apple has revamped its app store policies in the
“An opportunity to choose chance.”
That’s what social platform startup 222 claims to offer its members. It isn’t a dating app—there’s no swiping, and, mo
Five years ago, I bought an e-bike. At the time, the motor-equipp
Not all emojis are created equal.
The sparkle emoji or red heart emoji are staples of t
As Zohran Mamdani declared victory in the New York C
A first public test of robotaxis by Tesla in Austin, Texas led to multiple traffic problems and dri
