Research & Papers

Kazuki Nakayashiki's Study Reveals Highlighting's Social Nature

New research shows highlighting reflects social trends more than personal choice.

Deep Dive

Kazuki Nakayashiki and Keisuke Watanabe's paper, 'Personal Salience', reveals a striking asymmetry: which sentences you highlight is predicted far better by the crowd than by your own history, with the personal signal barely a whisper (own-vs-other gap of just +0.017). Yet when it comes to selecting which already-salient passages are yours, individuality roars—your own history predicts strongly (gap of +0.14). That personal signature largely reflects stable thematic preference, shrinking about 6–8x against a topically-matched peer. So highlights carry a genuine individual layer, but it's thin over a strong shared one—far more visible in what you pick than in what stands out.

Key Points
  • Crowd behavior predicts highlighting decisions 6-8x better than personal history.
  • Personal salience is weaker, averaging a +0.017 gap in predictions.
  • Individuality is more evident in selected highlights than in general salience.

Why It Matters

Understanding highlighting behavior can enhance personalization in information retrieval systems.

📬 Get the top 10 AI stories daily