Balancing Specialization and Adaptation in a Transforming Scientific Landscape

This repository contains the source of the paper "Balancing Specialization and Adaptation in a Transforming Scientific Landscape".

Installation

The repository relies on DataLad, a git and git-annexed based software for reproducible research and data sharing. DataLad installation instructions can be found here.

If you do not want to use DataLad and just want to take a look at the code and data from your browser, please go to https://gin.g-node.org/lucasgautheron/adaptation_specialization_material.

To install the repository and all the dependencies (code and data):

datalad install -r git@github.com:lucasgautheron/specialization-adaptation.git

Please note that the repository contains nested submodules. The above command will install all of them recursively.

For certain analyses, you will need to download the source data:

datalad get specialization_adaptation_material/inspire-harvest/database

Some analyses will require intermediate analyses outputs to be downloaded as well. They can be downloaded individually with datalad get, e.g.:

datalad get specialization_adaptation_material/output/etm_20_pretrained/etm_instance.pickle

Manuscript compilation

The manuscript can be re-compiled by issuing the following commands:

make clean
make

Code organization

The code that performs the main analyses can be found in specialization_adaptation_material/code. The table below describes its organization:

Script	Function	Dependencies
`code/etm.py`	Learns embeddings and topics.
`code/etm_compile.py`	Counts keywords part of each topic within each article using Naives Bayes and the previously trained topic model.	`output/<>/dataset.pickle.py`, `output/<>/<br>`output/<>/etm_instance.pickle.py```
`code/etm_transfers.py`	Evaluates scientists' research portfolios.	`output/<>/topics_counts.py`
`code/authors_sociality.py`	Calculates social capital.	`output/<>/aggregates.csv`
`code/etm_ei.py`	Runs the ecological inference model with Stan.	`output/<>/aggregates.csv`, `output/<>/pooled_resources.parquet`
`code/etm_map.py`	Evaluates the MAP performance of the ecological inference model using K fold cross-validation.	`output/<>/aggregates.csv`, `output/<>/pooled_resources.parquet`
`code/topic_distance.py`	Evaluates "distances" between topics using different metrics.	All of the above
`code/optimal_transport.py`	Recover migration cost matrix using probabilistic inverse optimal transport.	All of the above and MCMC samples from the ecological inference model.
`code/comparative_analysis.py`	Performs a comparative analysis of the effect of capital on different metrics of change in research interests.	All of the above and MCMC samples from the ecological inference model.

Each of the above script should be run from specialization_adaptation_material. Input parameters can be listed by doing python code/<script.py> --help.

The code available under specialization_adaptation_material/plots produces plots using the output of analyses performed by the above scripts.