Algorithmic transparency is openness about the purpose, structure and underlying actions of the algorithms used to search for, process and deliver information. An algorithm is a set of steps that a computer program follows in order to make a decision about a particular course of action.
The question of whether or not algorithms that affect the general public should be made transparent is controversial. Take, for example, a program used to determine credit scores. If someone is given a lower credit rating than they think they deserve, they have the right to appeal the score, but not the right to demand that algorithms used to determine the poor score be made public. This is because the company that determined the person's credit score also has rights -- in this case, the right to protect their intellectual property (IP).Content Continues Below
A common method used to provide transparency and ensure algorithmic accountability is the use of third party audits. This approach is known as qualified transparency. After complaints were made to the Federal Trade Commission (FTC) about the search giant Google, for example, watch-dog algorithms created by FTC staffers found that Google’s search algorithms generally caused its own services to appear ahead of others in search results. To provide transparency, the criteria used in the evaluation, as well as the results, were publicly released and explained. Although the FTC decided Google's actions were not anti-competitive in nature, the negative publicity the investigation created inspired Google to make changes.
Decisions made by algorithms can be opaque because of technical and social reasons, in addition to being made purposely opaque to protect intellectual property. For example, the algorithms may be too complex to explain or efforts to explain the algorithms might require the use of data that violates a country's privacy regulations. Regardless of the reasons, governments, corporations and private organizations throughout the world are exploring ways to address the issue of algorithmic best practices and accountability and provide the general public with as much transparency as possible to build trust.