RIP Social Media. What Comes Next Is Messy.

Why It Matters

The findings expose fundamental design flaws that fuel polarization and limit genuine discourse, urging platforms and policymakers to rethink architecture before societal harms deepen.

Key Takeaways

• Echo chambers arise from platform architecture, even without filter bubbles
• Small filter bubbles can stabilize diversity by anchoring agreeable users
• Human posting drops 50%; AI bots now dominate legacy platforms
• Twitter/X engagement shifted 72 points rightward since 2020
• New media split into private chats, protected communities, algorithmic broadcasting, AI bots

Pulse Analysis

Törnberg’s recent papers combine agent‑based modeling with large‑language‑model personas to simulate online interactions, confirming that the structural rules of social‑media platforms inherently generate echo chambers. Even when users are programmed to seek diverse viewpoints, the system’s feedback loops push dissenters out, creating self‑reinforcing homogeneity. Paradoxically, modest filter‑bubble effects—where users see a small proportion of like‑minded content—can act as a stabilizing force, preventing runaway polarization.

Survey data from the 2020 and 2024 American National Election Studies reveal a stark shift in platform dynamics. Human‑generated posts on legacy networks have fallen roughly 50%, while AI‑driven bots now dominate the content stream, especially on X, where engagement has moved 72 percentage points to the right. Meanwhile, TikTok and Reddit buck the trend, showing modest growth and signaling a migration toward short‑form, algorithmic video content that blurs the line between social networking and broadcast media.

The research identifies four emerging digital ecosystems: private group chats (e.g., WhatsApp), protected subscriber communities (e.g., Substack), algorithmic broadcasting platforms (e.g., TikTok, Instagram Reels), and direct AI‑chatbot interactions. These shifts challenge traditional notions of social media and raise regulatory questions about content moderation, data privacy, and democratic discourse. Early experiments—such as BlueSky’s granular blocking tools and X’s community‑notes—illustrate how nuanced design tweaks can mitigate extreme feedback loops, offering a roadmap for more resilient, inclusive online spaces.