Why is this course important?
Suggested time: 5 minutes
In recent years, there has been a rapid proliferation of algorithmic tools into the public and private sectors to improve processes and increase efficiency. Algorithmic decision-making systems (ADS) are systems that use algorithms for making decisions in a specific context, such as finance, employment, healthcare, and education.
While these algorithmic tools have the ability to greatly improve society, they also have the potential to cause great harm. As a case study, Amazon once built and implemented an automated resume screening and hiring system—only to later find out that the system was biased against hiring women. In another instance, algorithmic systems threatened global economic stability by causing the 2010 Flash Crash, wherein erroneous decisions made by complex algorithmic trading systems caused the Dow Jones to lose $1 trillion in value in 36 minutes. Both of these issues were caused, at least in part, by a lack of transparency into the underlying algorithmic system.
In this course, discuss instructions, best-practices, and recommendations on algorithmic transparency to avoid potential risks and harm. Our approach involves making the stakeholders of algorithmic decision-making the primary focus—that is, the individuals or groups (both internal and external to an organization) that are impacted by an algorithmic system.
The content of this course is based on the Algorithmic Transparency Playbook, which is a free-guide to algorithmic transparency published by the New York University Center for Responsible AI.
What will this course teach you?
The main objective of this course is to teach you how to create transparency for the algorithms in your organization.
Along the way, you will learn about many different elements of algorithmic transparency, and by the end of this course you should be comfortable answering the following questions:
- What is algorithmic transparency and how is it defined?
- Who are the stakeholders of algorithmic transparency?
- What are the goals of algorithmic transparency?
- What are the existing methods for algorithmic transparency?
- What are the best practices for designing and implementing algorithmic transparency into existing and future algorithmic systems?
- How can algorithmic transparency be maintained into the future?
- How can I help influence a culture shift in my organization towards adopting algorithmic transparency?
Who should take this course?
This course is intended for a range of audiences:
- C-suite executives and managers. In organizations using algorithmic decision-making systems, it’s critical that leadership understands how important algorithmic transparency is, both from a business and ethical perspective. Executives and managers should understand what algorithmic transparency is, why it is useful for different stakeholders, and how it can be implemented within their organization.
- Policymakers and regulators. In recent years, governments around the world have begun to regulate algorithmic systems, often requiring some degree of transparency. Unfortunately, all existing and emerging legislation on algorithmic transparency to date share a common weaknesses: it has focused on what to do (or what not to do), but has left the brunt of the work to data scientists to figure out how. By reading this playbook, policymakers can better understand concrete ideas about algorithmic transparency, and ergo better understand how it can be regulated.
- Data scientists and data/software engineers. This playbook is useful for learning state-of-the-art and industry-standard practices for algorithmic transparency. By reading this playbook you will learn new methods for transparency and have an understanding of how to implement transparency in the most effective way possible in your organization.