Gender, Access, and Reading Behavior

This project was largely inspired by a 2014 collaboration between UNESCO and Worldreader, which sought to develop “insights into how mobile technology can be leveraged to better facilitate reading in countries where literacy rates are low.” (UNESCO 2014, 9) The problem of increasing literacy is particularly tricky in many of these contexts, since educational opportunities – including those offered through mobile devices – are often not equally accessible by all segments of the population.

User Behavior - Part 3, Genre Progression

In the last blog post we analyzed user journeys. Another facet of user journeys is not at the same level of granularity as event logs and pages read, but rather at the book level. For readers who read some of at least one book, does the book genre influence whether they read again? For users who read another book, if their first book is in a certain genre, are they more or less likely to read again in that same genre or a different one? Do some genres lead to a mix of books in other genres? Do some genres primarily lead back to themselves?

User Behavior - Part 2, User Journeys

This is the second of three posts on user behavior within the Worldreader app. Where the last post looked at general user patterns by registration status and browser [link to previous post], this post will describe a more granular exploration of user journeys. User journeys is the term we are using for the analysis of user behavior with the added component of time, not just looking at static counts and averages of user activity as before. This necessitates the construction of many additional indicators to capture this extra detail, while at the same time carefully considering the computational cost of each one. Although this section has received a lot of analytical work, results are still mostly elusive. This is partly due to data issues, since these indicators are more specific than basic summaries, any problems with the data crop up quicker and more directly than before. Changing data also means computational challenges, and since these indicators are time-consuming to compute anyways, doing so multiple times compounds the problem.