Mitigating Racial Bias in Machine Learning.

When applied in the health sector, AI-based applications raise not only ethical but legal and safety concerns, where algorithms trained on data from majority populations can generate less accurate or reliable results for minorities and other disadvantaged groups.

A Bayesian network approach to determine environmental factors controlling Karenia selliformis occurrences and blooms in the Gulf of Gabès, Tunisia.

A Bayesian Network modeling framework is introduced to explore the effect of physical and meteorological factors on the dinoflagellate red tide forming Karenia selliformis in various sampling sites of the national phytoplankton monitoring program. The proposed models took into account the physical environment effects (salinity, temperature and tide amplitude), meteorological constraints (evaporation, air temperature, insolation, rainfall, atmospheric pressure and humidity), sampling months and sites on both Karenia selliformis occurrences and blooms. The models produced plausible results and enabled the identification of the factors that directly impacted on the species occurrences and concentration levels. The sampling sites dominated the species occurrences. The models show that the relationship between salinity and Karenia selliformis is more apparent when the species concentrations are focused on and that the bloom occurrences can be predicted based on salinity. Concentrations up to 105 cells L-1 were recorded when salinity exceeded 42.5 and dominated the shallow and weak water renewal areas.