ALTER aims to advance research in the intersection of explainability, fairness, and user interaction in the domain of automated decision systems. ALTER will design methods to (1) provide explanations capable of conveying fairness aspects, (2) audit the system to examine fairness and discover latent source of bias via explanations, (3) semi-automatically intervene to ensure system fairness, and (4) interact with users to facilitate and guide the aforementioned scenarios. ALTER will cover the needs of different stakeholders, such as consumers of recommendations, providers/producers of data items to be recommended, and system owners, who may have a varying level of familiarity and expertise with the system and the underlying AI technologies. The proposed research will be materialized via open source libraries and tools within the ALTER framework, supporting the execution of user interactive explanation, fairness auditing and fairness correction pipelines on recommendation results.

2024

2028