REGALIA

Understanding the behavior of AI algorithms and, in particular, machine learning algorithms has become an important challenge
 for regulation authorities but also for industry players wishing to develop algorithms that comply with legal regulations or
 ethics requirements. On the one hand, the presence of bias and discrimination is well acknowledged in machine learning and
 may lead to unwanted behavior and predictions. Rather than providing decisions that appear as sharp and accurate, 
algorithms may perpetuate or even exacerbate biases in the training data, resulting in decisions that may discriminate part
 of the users. On the other hand, the opacity and the lack of explainability of such algorithms may lead to behaviors that
 may be different from the original claims on the purpose of the algorithm, leading to harmful behavior.
Our objective is to evaluate AI algorithms and provide new tools to decide whether a machine learning algorithm is compliant with norms, of legal nature or diven by economical regulations. 