DPO Presenation - Fairness
There is a plethora of metrics. At Code4Thought, we use the disparate impact ratio, as it is an industry standard derived by the four-fifths rule of the EEOC
DPO Presentation - Transparency
How “transparent” is transparency under the GDPR? Many scholars argue over whether there is a right to demand an explanation for an automatic decision. Even in the best case scenario there are some “loopholes”.
IEEE BigData 2020 Presentation
Presentation of our paper at the 2020 IEEE International Conference on Big Data. We compared our ML Interpretability method called MASHAP with the most well-known counterpart, i.e. LIME. Our experiments showed that MASHAP, while achieves similar consistency, is much faster than LIME. (Paper title: “Evaluating MASHAP as a faster alternative to LIME for model agnostic machine learning interpretability”).
DPO Presentation
Challenges we face in order to build trust
Although the trustworthiness of an algorithmic system is undoubtedly a desired property, there are several challenges that demote it to a nice-to-have feature instead of a priority one.