Call for Papers (closed)

Important Dates

7 February 2022: Abstract submission (in-depth and lightning round track)

25 February 2022: Community voting for in-depth papers

3 March 2022: Acceptance notifications (in-depth and lightning round track)

21 March 2022: Full papers (in-depth track)

8-9 June 2022: European Workshop on Algorithmic Fairness

Goal of the workshop


Algorithmic fairness is a rapidly growing research area, and the importance of this field is only going to increase over the next years. The discussion predominantly involves US-based researchers, but more and more researchers in Europe are getting active in this domain.


The goal of this workshop is to foster the dialogue between European researchers working on algorithmic fairness in the context of the legal and societal framework of Europe, especially in light of the EU’s attempts to promote ethical AI. While this workshop is primarily focused on researchers based in Europe, anyone with an interest in the topic is equally welcome!


The workshop is intended to promote the field by taking an interdisciplinary approach (including mathematical, philosophical, legal, psychological, sociological, and management perspectives). We want to combine conceptual and empirical scholarship with original perspectives on how to bridge the gap between current theoretical concepts and concrete applications.


The workshop is organized by a group of researchers based in Zurich and Milan who have been working on different research projects on the topic of algorithmic fairness since 2017.


Workshop topics


The topic of the workshop is algorithmic fairness for machine learning (ML) applications involving data-based prediction models and prediction-based decision algorithms.


The focus of the workshop is on so-called fairness metrics [1, 2]: Normative fairness goals have to be expressed in such metrics, and the fairness of prediction or decision algorithms are measured using these metrics. Among the major challenges of using such metrics are questions such as:

  • Which fairness metric(s) should be used in a given context? The scholarly community has not yet agreed on a method to make this choice.

  • Which are the moral arguments for preferring one metric over another one?

  • To what extent can and should there be trade-offs between fairness metrics?

  • How can fairness requirements be combined with the original goal of data-based decision-making, e.g., maximizing profit? How should the trade-off between business targets and social fairness expectations be managed?

  • What is the relationship of the many proposed fairness metrics with current European legislation? Which ones are compatible with legislation, which not?

  • How can the different fairness metrics be taught (and especially their differences explained) to data scientists or to management?

  • What are the challenges and possible solutions when implementing fairness-sensitive algorithms in practice?

  • What are the limitations of such fairness metrics? What risks are associated with, e.g., using fairness metrics as benchmarks?


During the conference, we want to discuss ideas and suggestions to address these questions.


As we are convinced that the field of algorithmic fairness needs to be addressed in an interdisciplinary way, we encourage participants with different backgrounds to participate. We expect every participant to contribute with their domain-specific expertise in a way that allows people with other backgrounds to learn and integrate these new perspectives into their own work.


Format


As the goal of this event is to foster the exchange within the European algorithmic fairness community, this workshop will be focused on exchanging and discussing ideas and concepts, and a large part of the time will be devoted to discussion. We think that a physical presence supports this goal, so we are planning an in-person event in Zurich (if the pandemic situation should prohibit physical events, we will make it an online event).


We are envisioning a workshop that allows participants to deeply engage in discussions while learning about the work of as many community members as possible. Therefore, we will have a mixture of in-depth discussion rounds and short lightning round presentations:


The in-depth discussion rounds are focused on contributed full papers. The selection process will include a community voting process: After the abstract submission phase closes, we will distribute the abstracts among all the people who expressed interest in EWAF for gauging the interest in the topics. The organizers will make the final selection considering the voting results, but also other factors such as the diversity of speakers and topics.


Papers for the lightning round will get about 5 minutes of presentation time, based on a submitted extended abstract of max 1000 words. There will not be any time for questions, but there will of course be time to discuss these ideas during the breaks. This is a good fit if your contribution is still in its early stages and if you want to find potential collaborators in the community.


We encourage all participants to submit a paper. Both the full papers and the extended abstracts will be distributed prior to the workshop. Both previously published and new work can be submitted to both tracks.


Submission guidelines


You may submit your abstract either for the in-depth discussion or for the lightning round.


  • Abstracts submitted for in-depth discussions: Submit your abstract by January 31 February 7. If your paper is selected, you should submit the full paper by March 21. If it is not selected, we will automatically consider the abstract for the lighting round (unless otherwise indicated by you).

  • Abstracts submitted for lighting round: Submit extended abstract by January 31 February 7.


Full papers for the in-depth track should be 5-15 pages long. You can use your own formatting, but please make sure to use 10pt Times roman font and a single-column format.


Abstracts can be submitted here: https://docs.google.com/forms/d/e/1FAIpQLSce-B6CDmqQFbUCetkVu6-o3MJeswXeSamSPZXPuzJP--JgCg/viewform?usp=sf_link



Publication options


We plan to put together a special issue in an interdisciplinary journal after the workshop. We are currently contacting different journals to gauge interest and negotiate the conditions. Updates will follow.


For faster publication, you may submit your work to the Swiss Conference of Data Science 2022 (SDS2022), taking place June 22-23. This conference is the most important conference for data scientists in Switzerland, having a special emphasis on socially acceptable AI applications this year. Accepted papers will be submitted for inclusion in the IEEE Xplore digital library. Further information can be found here.


References


[1] S. Mitchell, E. Potash, S. Barocas, A. D’Amour, and K. Lum, “Algorithmic Fairness: Choices, Assumptions, and Definitions,” Annual Review of Statistics and Its Application, vol. 8, no. 1, pp. 141–163, 2021, DOI: 10.1146/annurev-statistics-042720-125902.


[2] S. Verma and J. Rubin, “Fairness definitions explained,” in Proceedings of the International Workshop on Software Fairness - FairWare ’18, Gothenburg, Sweden, 2018, pp. 1–7. DOI: 10.1145/3194770.3194776.