STSE
Through the CyCAT project, Short-Term Staff Exchanges (STSE) between the partners and others can be arranged. The STSE participants (visitor and host) must define some goals to achieve during the exchange. Furthermore, the STSE is expected to result in ideas for future collaboration and/or other collaboration activities. In addition, depending on the goals of the STSE, the visitor may agree to give a talk at a seminar or a related class at the host institution.
*Please send your proposal for a visit to info@cycat.io using the following form:“Discrimination Detection in Benchmarking Datasets for IR and NLP”
The STSE “Discrimination Detection in Benchmarking Datasets for IR and NLP” hosted by Dr. Jahna Otterbacher, took place at the Open University of Cyprus between 11 – 17 of November 2019 with two visitors, Professor Paolo Rosso and Dr. Anastasia Giachanou from Natural Language Engineering Lab, Universitat Politecnica de Valencia. All member of the CyCAT OUC team, as well as the Transparency in Algorithmic Group (TAG) at the RISE research center in Nicosia, participated in the first day of the exchange (Monday 11/11/2019). The local colleagues J. Otterbacher, S. Kleanthous Loizou, K. Orphanou, M. Kasinidou, P. Barlas, K. Kyriakou and P. Kleanthous participated in all activities during the week. This visit had four goals: i) Outline a plan for guiding discrimination detection in datasets, to be used in systematic evaluations, and a plan for detecting and analyzing discrimination biases of gender detection systems; ii) Examine whether gender stereotypes can actually be found in texts written by men and women. In addition, the participants planned to examine if gender detection systems tend to reproduce the gender stereotypes in their predictions; iii) Create a draft of a white-paper report detailing good practices in terms of mitigating social discrimination, when constructing these types of datasets; iv) Organize/deliver a public lecture from each of the two visitors and v) Facilitate meeting with colleagues at OUC and RISE. The last two goals were completed and the first three goals had good progress. The team reviewed existing literature on the types of biases that can occur in social data. An in-depth interview was conducted with Prof. Rosso in order to understand how this affects the process of constructing shared tasks and benchmarking datasets. A draft is also in place for a white-paper report, once additional interviews and analysis are conducted.
“Untangling Concepts and Definitions Surrounding Algorithmic Bias and Transparency”
The STSE entitled “Untangling Concepts and Definitions Surrounding Algorithmic Bias and Transparency” took place at the University of Trento in Italy between 4 – 8 of February 2019, hosted by Prof. Fausto Giunchiglia. The participants in this STSE included K. Batsuren from the University of Trento, V. Bogina, A. Hartman, T. Kuflik and A. Shulner Tal from the University of Haifa, S. Kleanthous Loizou and J. Otterbacher from the Open University of Cyprus. The goals of the meeting were to identify and define the key concepts related to the scope of the CyCAT project and to write an outline of a tutorial paper that the partners could use in various ways; for example, as a session at a conference/ workshop, or published as a position paper in a relevant outlet. The goals were broadly achieved during the visit to Trento. Partners agreed upon the framework for examining the interrelated concepts of Fairness, Accountability, and Transparency (FAT) and the mechanisms proposed in the literature to promote FAT. Also, two collaborative papers were outlined and one of them was published as a conference paper at IEEE SMAP 2019.
