AEQUITAS aims to widen the array of stakeholders usually involved in the development of AI systems, to ensure a more balanced mix of relevant actors.
To address and tackle the multiple manifestations of bias and unfairness in AI, the AEQUITAS project will use a participatory design process. This way, not only developers, but also (private or public) providers and users, as well as groups sensitive to AI bias will be engaged in this participatory process with the final goal of designing, developing and deploying fairer AI systems.
The AEQUITAS framework will be validated through use cases covering distinct domains representative of industrial and digital sectors (hiring), public bodies (health) and sensitive social issues. These use cases were selected in order to cover both a wide range of data typologies and broad sector coverage as well as different types of bias (on data, algorithm and interpretation) and different framework components (assessment/repair service and design of new systems service).
Fair tool supporting the diagnosis phase in pediatric dermatology diseases
Fair classification of ECG traces as symptomatic or normal
Fair AI system to detect and assess risk for child abuse and neglect within hospital settings
Fair AI system to detect educational disadvantages
Fair AI assisted recruiting system to target the cognitive and structural bias associated with the recruiting process
Assess of existing algorithm exploited for selection of candidates
Through the AEQUITAS use cases, enterprises, public bodies, associations, representative bodies and citizens will provide input on the individual dimension with existing tools and datasets that are currently exploited by companies or available on the market. The combination of case studies from Adecco, Philips, IRCCS Bologna Hospital, and the University of La Laguna, as well as of the association partners of the Consortium – Women in AI, Arcigay, Period Think Tank, Eurocadres and Asociacion Rayuela – will provide indicators that will then be used to develop the AEQUITAS technology.
