Fairness is a complex and multifaceted concept that has garnered increasing attention in many computer science research areas. Defining and measuring fairness is often context-dependent, as different applications may require tailored approaches to address specific ethical concerns.
Understanding the relationships between involved groups can be facilitated by defining an ontology or identifying key properties. However, achieving fairness demands ongoing evaluations, transparent decision-making, and adaptability to evolving contexts.
In this effort, another crucial challenge is managing and analyzing heterogeneous data types - data from diverse sources, formats, and structures. To achieve fairness, there are several possible approaches: (i) Establishing standardized definitions for similar data fields promotes consistency; (ii) Identifying potential biases and designing mechanisms to reduce them; (iii) Handling missing data with appropriate imputation techniques and augmenting data to guarantee that representation of different groups; (iv) Promoting diverse representation to support an equal decision-making system.
Ultimately, guaranteeing fairness in computational systems depends not only on data management but also on the careful design and evaluation of algorithms at various stages, including pre-processing, modeling, and post-processing. Incorporating fairness into these stages is critical for mitigating biases and fostering equity in data-driven systems.
This workshop aims to serve as a small step forward in a field as vast as it is significant to the scientific community. We welcome articles with a particular focus on descriptive ontologies, fairness metrics, and properties, as well as frameworks for managing data heterogeneity to ensure fairness, and fairness-aware algorithms. We are also open to discussing other fair-related topics that are not explicitly mentioned.
This issue welcomes, but is not limited to, submissions on the following topics:
Fairness definition
Definition of ontologies and properties
Fairness metrics, verifying and measuring the level of fairness in results
Challenges in Managing Heterogeneous Data Types to Ensure Fairness
Semantic Inconsistencies: Standardization of similar data sources
Standardization of ontologies and properties across different datasets
Identifying and reducing bias in data and algorithms
Pre-existing inequalities embedded in the data.
Over- or under-representation of certain groups.
Errors or inconsistencies in data collection methods.
The impact of data combination on fairness and bias
Missing data imputation techniques and data augmentation to ensure fairness
Diversity preservation: Techniques to balance data representation and ensure fairness in diverse contexts
Addressing overfitting issues in fairness-aware systems
Fairness-aware algorithms and methodologies
Evaluating existing fairness-aware algorithms (e.g., recommender systems, machine learning, and deep learning models)
Ensuring fairness during the various stages of algorithms (e.g. pre-processing, modeling, post-processing)
Integrating fairness components into existing algorithms