Artificial Intelligence and Machine Learning
Artificial intelligence (AI) and machine learning are rapidly evolving fields that will continue to have a high impact on the digital humanities. These advanced computational techniques can potentially provide many benefits to the study of cultural artifacts, text analysis, distant reading, and even to close reading. AI and machine learning have been widely employed in image processing and analysis, and are important in video analysis, medical imaging, geographic information systems, and other areas with a high emphasis on images and imaging technology. Consequently, image-intensive digital humanities areas, such as cultural analytics, are the direct beneficiaries of advances in these computational approaches.
However, some digital humanists caution against applying these techniques in an ad hoc manner, urging instead to use them critically, taking sustainability (i.e. re-usable real-world applications) and ethical implications into account (Hedges et al., 2019). In a working paper “Digital Humanities Foresight. The future impact of digital methods, technologies, and infrastructures”, King’s College London cultural informatics researcher and digital humanist Mark Hedges and colleagues recount the thoughts of an interviewee concerning ethical issues arising from the utilization of algorithmic techniques: “Stylometry [the statistical analysis of variations in the writing style between genres or writers] can even trace whether there is a potential for Alzheimer’s disease. Can we use and disseminate this knowledge? … Can we give information about a person to an eventual employer? Can we use information about a person’s psychological profile or sexual identity?” (Hedges et al., 2019).
Some of these scholars also argue that it not sufficient to simply employ these technologies and to know when to use them as a tool. The user must move beyond the “black box” aspects of these methods. Users must understand what the algorithm is doing and how it works. AI and machine learning techniques are susceptible to biases, and, as a result, digital humanists must have a full understanding of the underlying assumptions of these methods, including their limitations and when their use is warranted. Additionally, many AI and machine learning algorithms are still in the early stages of development, and are considered to be experimental. Because of these issues, some digital humanists argue that AI and machine learning need to be custom-developed for humanities work, instead of “forcing” existing techniques to be applied to humanities scholarship (Hedges et al., 2019).
Another interview, Hedges and colleagues recount the thoughts of an interviewee concerning AI in the digital humanities: “We have to be aware of the dangers [of AI]. The more the infrastructure is doing the scholarship for us, the more it’s applying its own lens, do we really know what we’re doing anymore?” (Hedges et al., 2019).