Eticas Foundation has developed the Observatory of Algorithms with Social Impact, or OASI. The Observatory’s objective is to better understand the risks and challenges posed by algorithms and machine learning processes. After publishing the Guide to Algorithmic Auditing, its methodology to audit algorithms, now Eticas brings together the algorithms used by governments and companies all around the world to shed light on their use and impact.

The rise and spread of mass data, statistical models, machine learning and algorithms have led to advances in fields like early cancer detection and anticipation of natural disasters. And because digital technology and algorithmic systems are often easily scalable, both public entities and private organisations are increasingly relying on algorithms and Artificial Intelligence (AI) to provide services to people.

As algorithmic advances transform public life, we have seen that the impact of algorithmic systems is not always positive but may also have a darker side. AI is systematically biased and causes race, gender and age discrimination, unfair and inequitable treatment, and also fails to deliver expected results.

At first glance, algorithms appear to be neutral: after all, they simply are series of logical steps and mathematical operations. However, from the data they are trained on and fed, and to the way they are sometimes designed, algorithms may work poorly and be unaccountable. Moreover, this may result in negative consequences for the rights and welfare of people, regardless of equality, fairness, transparency, or redress.

The Observatory of Algorithms with Social Impact aims to clarify the basic concepts as definitions of algorithm, machine learning and other relevant concepts. But OASI also contains a Register of algorithms being developed and implemented by different institutions and organisations across the world. The OASI Register gathers, classifies and makes algorithms searchable, and is regularly updated with new content.

See below part of the OASI Register: