In the past years, scientific research in Data Science (DS) and Artificial Intelligence (AI) has witnessed vast progress. The number of published papers and digital objects (e.g., data, code, models) is growing exponentially^{1 2}. However, not all research artefacts fulfill the criteria of being findable, accessible, interoperable and reusable (FAIR)³, contributing to a rather low level of reproducibility of experimental findings reported in scholarly publications and to the reproducibility crisis.

The Open Science Best Practices found in this website were collected to encourage the FAIR management and development of digital artefacts in DS and AI. It is worth noting that some of the recommendations should be followed directly in a paper, while the others (e.g., data preprocessing steps) could be addressed in a separate (meta)data/code documentation.

The recommendations are aligned with the typical research timeline associated with the development of scientific articles and split into the four phases:

1. Before starting the research
2. During the research
3. Paper submission
4. Paper publication

Authors

Ekaterina Borisova (Deutsches Forschungszentrum für Künstliche Intelligenz, Germany), ekaterina.borisova@dfki.de

Raia Abu Ahmad (Deutsches Forschungszentrum für Künstliche Intelligenz, Germany), raia.abu_ahmad@dfki.de

Georg Rehm (Deutsches Forschungszentrum für Künstliche Intelligenz, Germany), georg.rehm@dfki.de

To cite

@inproceedings{borisova2023a
  title = {Open Science Best Practices in Data Science and Artificial Intelligence},
  author = {Ekaterina Borisova and Raia Abu Ahmad and Georg Rehm},
  booktitle = {Proceedings of the 1st Conference on Research Data Infrastructure (CoRDI)},	
  year = 2023,
  keywords = {FAIR, Reproducible Research, Open Science},	
  editor = {York Sure-Vetter and Carole Goble},
  publisher = {TIP Open Publishing},	
  address = {Karlsruhe, Germany},
  url = {https://doi.org/10.52825/cordi.v1i.299},
  month = 9
}

Feedback

We consistently maintain this webpage to ensure accuracy and completeness. If you notice any errors or missing information, please do not hesitate to contact one of the authors mentioned above. We will update the site content based on community feedback.

Acknowledgements

This work was supported by the consortium NFDI for Data Science and Artificial Intelligence (NFDI4DS) as part of the non-profit association National Research Data Infrastructure (NFDI e.V.). The NFDI is funded by the Federal Republic of Germany and its states. The paper received funding through the German Research Foundation (DFG) project NFDI4DS (no. 460234259). The authors wish to thank both for funding and support. A special thanks goes to all institutions and actors engaging for the association and its goals.