Algorithm by Tarleton Gillespie

In his essay Algorithm, Tarleton Gillespie elegantly articulates the etymology of the word “Algorithm”, and how the different ways the term is understood make it difficult to use in an efficient manner.

At its most basic, Gillespie explains that an algorithm is “the insertion of procedure into human knowledge and experience”. This procedure is not necessarily digital; however it is something that can be codified and accelerated through computation.

The complexity arises in how the term algorithm is used by different audiences.

The first audience, programmers and technical people, regard the specific term as representing “a technical solution to a technical problem”. It is a definition which is highly specific scope: a procedure that solves a problem, defined in computational terms.

Defining those terms, and broadening the definition of what an algorithm is, are the Designers. They frame the problem, define the variables and completion state of the procedure. It is at this stage of development “where complex social activity.. and values held about it are translated.”

“Software is never politically neutral,” Snelting said in last week’s reading (A fish can’t judge the water), and the same is true for algorithms. During the design phase they may inherit any implicit biases or assumptions from their creators. Oversights in the design phase can be imperceptible at first, but as an algorithm is then utilised by greater and greater numbers of people, these small ripples can be amplified dramatically, with serious consequences. 

A stage in the development of an algorithm that I previously hadn’t considered is the “training” phase. Prior to reading this essay, I had always assumed that training data was part of a machine learning process, rather than a standard algorithm, but Gillespie includes it a key point in an algorithm’s development. During this process the designers may be able to tweak the algorithm’s results to be more satisfactory by omitting different data, and thereby baking in a level of distortion into the process.

The third group gillespie identifies as having a different understanding of the term “algorithm” are the general public. For the general public, an algorithm is a much more nebulous concept, with a heavier focus on the social implications. The general public are much less likely to be aware of the internal design and procedures of an algorithm than they are to feel the effects of those design changes in their everyday lives.

As a result the term algorithm can take on an almost mythic quality. It can be used by businesses as a “talisman”, an inscrutable artefact separate to the corporate structure that created it, a useful step of removal for corporate responsibility. Facebook’s algorithm may have put fake news all over the place during the last few years, but it’s not Facebook’s fault, it’s a fault in the algorithm.

Gillespie offers valuable perspectives and insight into how an algorithm is understood. In particular it’s clear that he wants us to keep in mind that these systems are created by groups of individuals and they are responsible for their creations. It’s clear that we are living in a world that will only get more dominated by algorithmic systems, and if we as a society are to submit to this “provisional tyranny” it’s important that we have a clear way of holding the creators to account.