The study delves into the dynamics of human communication within various social settings, from intimate groups to global online platforms, focusing on the reciprocal exchange of information as a cornerstone for social stability, cohesion, and cooperation. My contributions were to develop the research ideas, write the article, and individually develop median-based hypothesis tests, to understand how reciprocity and memory arise in human communication. One of the results is that communication on Twitter seems less reciprocal and bursty, possibly due to the intended use of the platform as a public setting dominated by unidirectional messaging aimed toward wider audiences. Below are the illustrations of measures we used in our research:

For further details, check out these links:

• EPJ-Data Science article link
• arXiv preprint link

The project comprises several Jupyter notebooks, each dedicated to a distinct phase of the research:

1. shuffling_resampling_null_model_functions.ipynb - Scripts to generate resampled networks (graphs) based on different hypotheses.
2. median_based_hypothesis_test.ipynb - Describes how the median-based hypothesis test is performed on the Calls dataset.

Intermediate data used is uploaded in the pickle (.pkl) format. Measures are computed in the shuffled datasets - shuffled_calls_measures.pkl and the original empirical datasets - original_results.pkl.

To access the full data used in the research (SMS, calls, e-mails, Twitter), please refer to this GitHub repository