Participant selection for workshops and conferences made easy.
Selection of participants for meetings as a discrete optimization problem.
Given a list of participants with various "attributes" (e.g., gender, career stage, subfield, geography, years since PhD), the code finds the distribution of values within each attribute (e.g, male, female) and generate a subset that approximates target value distributions (e.g, 50% male/female, 30% junior, 30% non-US). The attributes and values are user determined and based on the data present in uploaded CSV file.
Important: While a flask app exists in this repo, it is only a very rudimentary reference implementation of what we'd like an actual app to look like and not fully functional. Please use the Python code instead, as demonstrated in the tutorial Jupyter notebook, or contact us for advice.
This tool was born out of a very practical need: given that there are more applicants for a workshop than spaces, and given that some constraints on a target mixture of participants, and perhaps some pre-selected participants (e.g. invited speakers), how can the set of participants be identified such that it meets the target mixture as much as possible?
Instead of letting human organisers decide heuristically (including all the various biases that this entails [refs?]), this tool treats it as a (fairly complex) optimization problem, where the computer finds a subset of applicants that most closely matches the target mixture.
The targets could be based on various objectives that depend on the workshop or conference: for example, one could optimize for a certain mixture of abstract scores, for a breadth of talk topics, for a certain ratio of (academic) seniority, or for measures like gender or ethnicity.
Note that this is explicitly not the same as a quota: the underlying algorithim optimizes for the subset of participants that overall matches up with the targets, taking into account all targets simultaneously. It is also possible to include relative weights between targets, depending on how important they are to you.
Note: The app is under active development and requires updates before running reliably. As of now, it's not really usable. Use the Python library instead (see below).
Download the app/ directory
pip install -r requirements.txt
python server.pythen go to http://localhost:5000/ in your web browser
Below are some practical considerations. You can find a Jupyter notebook in this repository that will show how to run the code itself in more detail. For a recent example of how Entrofy was used in practice, see the PyAstro 2017 participant selection repo.
First, collect data about the acceptable participants. This requires you to know which criteria you actually care about when selecting participants. This will very strongly depend on the scope, the objective and format of your workshop or conference.