Public Policy

The exponential rise of Machine Learning and Deep Learning technologies has transformed the economy and the society in the 21st century, driving a digital transformation and adoption at an unprecedented scale. While this has resulted in overall boost in efficiency and rapid innovation, there have been significant concerns regarding the world being created.

The regulatory response till now has been mixed at best. While legal tools and remedies are still catching up, there is a widespread concern that “privacy is dead” that has sparked fears of a dystopian future. Governments have started looking at the business models of the tech firms, but this approach threatens of a regulatory hammer that looks at every problem like a nail, which in turn has people concerned of government takeover of a free digital space.

This project looked at a subset of the wider digital ecosystems, the machine learning systems, and identified the risks and harms involved. The harms are classified based on the market failure that they exemplify, and based on this, it is found that the most prevalent form of market failure experienced is Information Asymmetry.

Essentially, in training and deployment of these Machine Learning (or Deep Learning) models, the users are unaware of how their data is being used and once deployed how the decisions made by these models are shaping their behavior as well as the world around them.

Finally, it is recommended for the purpose of mitigating the harms identified, any future regulations aimed at Machine Learning systems should look primarily at solving the problem of Information Asymmetry and what could be potential consequences, both desired as well as undesired. Implementation of GDPR in Europe has some interesting insights in this regard.