AI tackles social media’s content moderation challenges

AI technology could revolutionize social media by shifting away from engagement-driven algorithms to platforms that respond to users’ stated preferences rather than their clicking behaviors. This fundamental shift might reshape how we understand human behavior online, moving beyond the assumption that our worst impulses drive our digital interactions to a more nuanced view of what people actually want from technology.

The original sin: Social media platforms have long operated like diners that serve whatever catches your eye, regardless of what you say you want.

• For 15 years, internet platforms have prioritized revealed preferences—what users click on and engage with—rather than what users explicitly state they want.
• This approach led to both positive innovations like educational videos and BookTok communities, but also created systems optimized for capturing attention rather than serving user intentions.

The AI difference: Large language models introduce a paradigm shift by prioritizing stated preferences over reflexive clicking behaviors.

• Unlike traditional algorithms that track engagement metrics, LLMs respond primarily to what users explicitly communicate rather than just monitoring their clicking patterns.
• This constraint fundamentally changes the dynamics of content delivery, making users’ direct requests like “help me stop doom-scrolling” actionable metrics rather than just measuring how long they continue scrolling.

Why this matters: This technological transition could transform our understanding of human behavior in digital environments.

• In a world where algorithms could only measure clicks and watch time, users appeared to be driven primarily by base impulses and attention-grabbing content.
• AI systems capable of conversation reveal a more complex picture of human desires and intentions, potentially offering a more flattering and accurate reflection of what people actually want from technology.