[Paper Review] A Genetic Algorithm to Optimize a Tweet for Retweetability

Ronald Hochreiter, Christoph Waldhauser|arXiv (Cornell University)|Jan 20, 2014

Complex Network Analysis Techniques13 references3 citations

TL;DR

This paper proposes a genetic algorithm to optimize tweet composition for maximum retweetability in a simulated Twitter-like network. By evolving message features such as length, polarity, emotionality, hashtags, and URLs, the algorithm significantly increases message reach across diverse node types, with notable improvements even for users with few connections.

ABSTRACT

Twitter is a popular microblogging platform. When users send out messages, other users have the ability to forward these messages to their own subgraph. Most research focuses on increasing retweetability from a node's perspective. Here, we center on improving message style to increase the chance of a message being forwarded. To this end, we simulate an artificial Twitter-like network with nodes deciding deterministically on retweeting a message or not. A genetic algorithm is used to optimize message composition, so that the reach of a message is increased. When analyzing the algorithm's runtime behavior across a set of different node types, we find that the algorithm consistently succeeds in significantly improving the retweetability of a message.

Motivation & Objective

• To investigate how message composition affects retweetability in social networks, independent of sender influence.
• To develop a deterministic simulation framework for studying information diffusion on Twitter-like networks.
• To optimize tweet features (e.g., length, sentiment, hashtags) using evolutionary computation to maximize reach.
• To evaluate the performance and stability of the optimization across different network topologies and node types.

Proposed method

• Simulates a Twitter-like directed network with nodes representing users and edges representing follow relationships.
• Assigns each node a deterministic retweet decision rule based on message features such as polarity, emotionality, length, URL count, hashtag count, and timing.
• Employs a genetic algorithm to evolve tweet content over 250 generations, using fitness defined by total retweet reach.
• Initial population consists of randomly generated tweets; selection, crossover, and mutation operators guide evolution toward higher-fitness messages.
• Each optimization run is replicated 50 times with random initial values to assess stability and convergence.
• Evaluates performance using mean fitness over generations and solution stability across replications.

Experimental results

Research questions

• RQ1 Can a genetic algorithm effectively optimize tweet features to maximize retweet reach in a simulated Twitter network?
• RQ2 How does the number of first-degree connections (followers) affect the convergence speed and performance of the optimization?
• RQ3 What specific tweet features (e.g., length, sentiment, hashtag count) are most influential in increasing retweetability?
• RQ4 How stable are the optimized solutions across multiple independent runs, and does stability vary with network structure?
• RQ5 To what extent can message styling alone improve retweetability, independent of sender popularity or follower count?

Key findings

• The genetic algorithm consistently improves tweet retweetability across all tested node types, with faster convergence for users having more connections.
• For users with 5 first-degree connections, the algorithm required an average of 720 generations to reach a fitness level beyond 200 retweets.
• Users with 10 and 170 connections reached optimal performance in approximately 50 generations, indicating faster convergence with higher connectivity.
• The mean polarity of optimized tweets was 0.38 (SD 0.17) for 10-node users and 0.32 (SD 0.15) for 170-node users, suggesting a positive sentiment preference.
• Optimized tweets averaged 104.28 characters (SD 9.80) for 10-node users and 72.18 characters (SD 14.59) for 170-node users, indicating shorter messages for highly connected nodes.
• The number of URLs and hashtags in optimized tweets varied by node type, with 6.84 hashtags (SD 0.71) for 170-node users, suggesting high engagement with topic-specific content.