• An agent-based model of user interactions in a simulated polarized Reddit environment

    Abstract: This study constructs an agent-based model in a Reddit-like social media environment. Agents and their posts are oriented around a single, polarizing issue, with polarities scored in [-1,1]. Agent behavior and various system metrics are analyzed with an interest in homophilic behavior: i.e. the tendency for agents of similar polarity to dominante the main interaction unit, a thread or reply tree. The model is successful in replicating a prominent result from the literature: that homophily and system-wide happiness are correlated. We also find that (a) high levels of post curation can have systemic negative effects, reflected in numerous metrics, and (b) permitting agent churn, coupled with sufficient agent policing-downvoting, produces marked majority-polarity dominance effects, with eventual near-complete homophily due to minority-polarity attrition. As a corollary, when downvoting is disallowed the runaway majority-dominance effect disappears, but this comes at the expense of substantially higher churn of polar-neutral agents.
  • Classifying raisins: Kecimen or Bensi?

    Processing the results of machine vision measurements of the humble raisin.
  • Rodent sightings under the New York City 311 complaint system

    An analysis of rodent sightings in Manhattan for the year 2019.
  • Automobile MPG predictions with linear regression

    A 1970s-1980s database on fuel efficiency for global car models is used to make MPG predictions.
  • Triangular billiards: plotting bounce sequences

    Even something as simple as a billiard ball bouncing on a triangular table can get complicated. In this post, some symbolic sequence measures create some (hopefully) interesting visualizations.
  • The noisiest places in New York City

    In this post, publicly available 311 data is analyzed to find some of the noisiest areas in New York City in 2019.
  • Assessing the complexity of entities: the knowledge build complexity

    How complicated is a paperclip? or a microchip? We build a measure and use some natural language processing to find out.