The Computing Community Consortium’s (CCC) Fairness and Accountability Task Force held a visioning workshop on Economics and Fairness, May 22-23, 2019 in Cambridge, Massachusetts. This workshop brought together computer science researchers with backgrounds in algorithmic decision making, machine learning, and data science with policy makers, legal experts, economists, and business leaders to discuss methods to ensure economic fairness in a data-driven world.

The workshop addressed five main areas:

Data Concentration

• Machine learning (ML), regression, etc. are essentially commodities; as a result, the data is the primary source of differentiation. Should we be concerned with the concentration of data in a small set of hands? If so, how do we guard against this?

Algorithmic Decision Making

• With many consequential decisions being delegated to algorithms, how do we ensure that such decisions comply with both the spirit and letter of the law, as well as with ethical/societal fairness considerations?

Algorithmic Recommendation

• In many settings algorithms make recommendations rather than decisions. How do humans interpret and make use of these recommendations? In particular, can the combination be worse than either one on its own? Particular focus will be paid to the following three domains: 1) Interaction between crime prediction and judicial decision making, 2) news feeds and misinformation, 3) marketing and consumer decisions (e.g, search by voice).

Implications for Platforms

• New technologies have disrupted traditional industries (taxi, hotel) by reducing barriers to entry and lowering the costs of search and coordination. The traditional structure made regulating them easier and ensuring that they complied with extant norms. The new forms disperse decision making (to various degrees). Who is responsible for ensuring compliance? If it is the platform, how can algorithms be used to ensure compliance with the law and/or norms. For example, how does AirBnB ensure compliance with fair housing? How does Uber ensure equality of earning opportunities for drivers?

Equality of Opportunity

• Much of the fairness discussion in the context of ML has been on fairness of outcomes with less attention paid to measuring equality of opportunity and/or ensuring access to opportunity. However, all three areas warrant consideration when discussing fairness. What constitutes fair equality of opportunity, and what role do algorithms play in ensuring economic equality through data driven decisions?

The goal of the workshop was to produce a report or white paper that articulates best practices and research challenges with regards to fairness and economics, as well as provides a sense of direction for the field. A workshop report is in progress.