Saeideh Bakhshi, David A. Shamma, Eric Gilbert
Analyzed 1.1 million Instagram photos and found that images containing human faces receive significantly more likes and comments than those without, regardless of subject matter or filter use.
Saeideh Bakhshi, Phuong Mai Nguyen, Robert Schiller, Tiantian Xu, Pawan Kodandapani, Andrew Levine, Cayman Simpson, Qifan Wang
Introduces a survey-based feedback measure for recommendation systems that captures users' intent to return for similar content, targeting long-term user value rather than immediate satisfaction signals.
Saeideh Bakhshi, David A. Shamma, Lyndon Kennedy, Eric Gilbert
Studied Instagram photo filters at scale to understand why people apply them and whether filtered photos receive more engagement — they do: filtered photos are more likely to be viewed and commented on.
Saeideh Bakhshi, David A. Shamma, Lyndon Kennedy, Yale Song, Paloma de Juan, Joseph Kaye
Combined qualitative interviews and large-scale analysis of Tumblr GIFs to understand what makes animated GIFs engaging — motion energy, faces, and emotional expressiveness are key drivers.
Saeideh Bakhshi, Partha Kanuparthy, Eric Gilbert
Examined millions of Yelp reviews and found that local demographics and weather conditions significantly predict review sentiment and star ratings, independent of food quality.
Eric Gilbert, Saeideh Bakhshi, Shuo Chang, Loren Terveen
Large-scale statistical study of Pinterest revealing that aspirational content — things people want to do or try — is a dominant motivation for pinning and re-pinning behavior.
Saeideh Bakhshi, Eric Gilbert
Studied how image color properties influence content diffusion on Pinterest, finding that images dominated by reds, pinks, and purples spread significantly more than blues and greens.
Saeideh Bakhshi, Partha Kanuparthy, David A. Shamma
Analyzed Yelp's social vote signals to understand what distinguishes high-quality reviews, revealing distinct behavioral and linguistic patterns for each vote type.