Responsible Data Management

Incorporating ethics and legal compliance into data-driven algorithmic systems is of interest to the computing research community, most notably under the umbrella of fair and interpretable machine learning. Julia Stoyanovich, co-founder and director of the Center for Responsible AI at NYU Tandon, addresses key insights regarding these issues, using examples from automated hiring systems based on automated decision systems (ADS).

Stoyanovich and co-authors propose a pragmatic definition of ADS. Such systems:

  • Process data about people, some of which may be sensitive or proprietary
  • Help make decisions that are consequential to people's lives and livelihoods
  • Involve a combination of human and automated decision-making
  • Are designed to improve efficiency and, where applicable, promote equitable access to opportunity

While tasks like legal compliance can, in principle, be performed by machine learning systems, accountability for the decisions being made by an ADS should always rests with a person.