Skip to content

Group Allocation

We implemented an algorithm to partition study participants on comparable groups. The algorithm is a heuristic that aims at minimizing skill differences between study groups. Although not needed to replicate our results, we release the implementation, for further transparency and easier extension of our work.

Source Code

The documented source code of the MiniMax implementation and documentation is available on GitHub.

Explanations

We implemented a custom MiniMax Heuristic, to assess the participant's skill profiles and create balanced experiment groups.

  • To maintain transparency we provide source code and documentation of the heuristic used to create the participant allocations.
  • The program parses all provided self assessment forms (see corresponding entry of replication bundle) which they provided during recruitment, and then performs a MiniMax search for the fairest group allocation.
  • A side product of this software is the generation of personalized emails for following communication with the individual participants.

Since the purpose of this program is to select of all applicants and provide a mapping from their legal names to group-specific pseudonyms, it can be only executed with access to the sensitive participant sheets. The latter we do not release, to protect participant privacy.