Ramsay’s book, Reading Machines: Toward an Algorithmic Criticism, contends that reading, and specifically critical reading, was algorithmic even in the pre-digital age. Now, with rapidly advancing technology and computer software, reading is becoming increasingly algorithmic and the language used to describe reading and criticism is reflecting this change. Despite opposition in the Humanities to this language of mathematically enabled reading, it nonetheless occurs subconsciously. In order to make an effective argument about any text, it is necessary to defend the argument with evidence, much of which is gleaned from the formal analysis of text and the relationships between particular words.
A much less controversial form of text analysis is the analysis of words across texts and the historical relevancies of particular words and concepts. This form of algorithmic analysis, as Stephen Pumfrey et al explain in “Experiments in 17th Century English: Manual versus Automatic Conceptual History,” can allow for expanded research “across more texts and in a shorter time scale” (Pumfrey et al. 396). As a result, digital text analysis, such as that which Pumfrey conducts, can allow for research questions to be asked and answered that could not previously be considered as a result of the large volume of work needing to be analyzed.
Pumfrey’s search of EEBO for the shift in the meaning of the word “experience” (later “experimental”) from a religious to scientific connotation, prompted him to manually sift through 2,700 hits in over 1,000 records (Pumfrey et al. 399). This process, the authors note, took nearly four weeks, in contrast to the minutes it took for the computer program to process the same amount of data and provide equally accurate results. Both of these processes, using computer software to one extent or another, significantly trump the 76 years it took to compile the first edition of the Oxford English Dictionary.
Evidently the algorithmic processes used in the analysis of large volumes of text are significantly more efficient in their use of time than a manual, non-computerized analysis of the same, or lesser amounts of text. As a result, it is possible to explore much larger questions, and text analyses are generally much more accessible. However, it is still important to question the validity of these digital programs: it is not possible for a computer (or a human for that matter) to know more than what they have been taught. This is one of the main reasons why we are consistently asked to type words in a box as part of a security feature for online websites: not only does being able to accurately type the text prove that we “are human,” it also allows for the digitization of texts and fonts that have not yet been digitized in an effort to allow the computer to “know more.”
As Ramsay’s analysis of the ELIZA program shows the reader, computer programs often struggle to develop an appropriate response to the inputs provided by a human, as a direct result of its lack of knowledge (Ramsay 59). Should a computer program be challenged with a request to which it cannot accurately respond, what will its response be? Does it only answer the parts of the request it can effectively answer? Does it attempt to answer unknown aspects of the request? Does it just not work? It is important to find answers to these questions in order to effectively understand what is happening “behind the program”; to discover the process of the program and interpret ourselves the effectiveness and accuracy of the program, and ultimately that of the information the program provides the user. Criticism may or may not already be algorithmic, but the algorithmic process must continue to be questioned, regardless of whether the algorithm is executed by a human or a computer.