Here is a curated collection of scholarly sources offering an in-depth academic foundation for the study of Digital Humanities and its most pressing contemporary debates!
This annotated bibliography highlights key perspectives, methodologies, and critical discussions shaping the field today.
For full access to a comprehensive list of sources used in the development of this website, consult the Zotero library linked below.
www.zotero.org/groups/6375579/dhlitstud-/collections/CIW5PJ54/collection
Brennan, Sheila A. ‘Public, First.’ Debates in the Digital Humanities, edited by Matthew K. Gold and Lauren F. Klein, University of Minnesota Press, 2016, pp. 384–90. dhdebates.gc.cuny.edu/read/untitled/section/11b9805a-a8e0-42e3-9a1c-fad46e4b78e5#ch32
Brennan begins by breaking down how public history and digital humanities are defined by remembering their core functions and to whom they are held responsible. They address various challenges the field of public history has faced in its development since the 1970s, citing various theorists who have been influential in its fruition, and emphasise its importance as a ‘democratic process’ in which the entire project of remembering and recording must be done in view of and with the contribution of the public. Because of this key tenet of the public history field, Brennan articulates that simply placing public history work on social media or in another digitally mediated space does not make it accessible to the public, and can increase the danger of viewing the ‘public’ as a monolith, presenting projects to them rather than building those projects with them. Brennan walks through the planning, creation, testing, and publication of the Histories of the National Mall project, identifying the particular strengths of the project in how it was able to engage with the public effectively. These strengths are worked into four concise points concerning public digital literacy, accessibility, approachability, and effective naming for the consideration of digital public historians to refer to when engaging in public history projects.
Brett, Megan R., et al. ‘Reframing the Conversation: Digital Humanists, Disabilities, and Accessibility.’ Debates in the Digital Humanities, edited by Matthew K. Gold and Laura F. Klein, University of Minnesota Press, 2023, pp. 324–43. dhdebates.gc.cuny.edu/read/debates-in-the-digital-humanities-2023/section/c389972a-3d63-4025-b5c2-e4f108e43e0d#ch21
In this article that provides suggestions for how to make digital humanities (DH) projects more accessible, Brett, Otis, and Kelly discuss how disability interacts with DH. DH is painted as a field that lowers barriers to accessibility. However, the truth is that many digital humanities projects are designed in ways that make them inaccessible to many various disabilities. This article discusses solutions to inaccessibility within a DH team that includes a disabled researcher as well as solutions to make projects more accessible to disabled members of the public. Many solutions are generally good practice—commenting code or using microphones even if they are annoying—but some are substantially more innovative, such as using hover text to provide definitions for key specialist terms. The authors acknowledge that it is not possible to make every project accessible to every disability, but they provide general solutions to shorten the gap between researchers and disabled readers.
Clement, Tanya E. ‘The Ground Truth of DH Text Mining.’ Debates in the Digital Humanities, edited by Matthew K. Gold and Lauren F. Klein, University of Minnesota Press, 2016, pp. 534-35. dhdebates.gc.cuny.edu/read/untitled/section/ef78ddc7-4087-4bb3-b192-16724631a172
In this chapter, Clement is providing a critical intervention into how data is treated in the humanities; specifically, she offers a contrast between the ‘certainty’ of text mining and the ‘uncertainty’ of sound mining. In doing so, she argues that within the digital humanities, text mining has become a ‘logocentric practice’; indeed, Clement critiques the logocentric tendency to treat text mining results as objective ‘ground truth,’ where researchers treat ‘The Word’ as an unchanging fact. This reliance on binary logic, one where machine results are accepted as settled evidence of identities like gender or genre, effectively erases the interpretive labour of the humanist. Clement goes on to introduce sound mining as a corrective model. She argues that unlike text, sound is aporetic and is thus inherently subjective. By lacking the perceived stability of a text, this practice instead forces researchers to acknowledge their own interpretations; for instance, when working with an audio file, a researcher must decide what constitutes meaningful signal and what constitutes background noise. By doing so, bias and perspective become overt within the practice of acquiring meaning. In providing this context, Clement is challenging digital humanists to apply this same logic to text by reimagining ‘The Word’ as aporetic rather than apodictic. By claiming that a word is indeterminate, she foregrounds the idea that meaning is not inherent in data but emerges through human interpretation. Thus, the chapter argues that text mining should not be used to produce definitive solutions but rather as a hermeneutic tool through which the contingency of meaning is continually negotiated.
Dhrubo, Abdullah Muhammad. ‘Digital Dasein: A Heideggerian Existential Typology of Digital Human Twins.’ AI & Society, 2025. doi.org/10.1007/s00146-025-02633-y
In this journal article, Dhrubo delineates a framework for assessing and naming the experiences of engaging with ‘data selves’ or personal human twins in terms of their potential to emphasise or undermine our humanity. The author defines digital human twins as ‘sophisticated, real-time digital representations of individuals.’ By drawing on Heidegger’s existential ontology and Boddington’s embodied technology framework, Dhrubo categorises digital twin experiences regarding their authenticity vs. inauthenticity, which results in four possible ‘existential modalities’ – ‘Digital Dasein, Alienated Authenticity, Transparent Inauthenticity, and Existential Datafication.’ This article constitutes a critical engagement and analysis of selfhood in an age when humans are perceived in terms of their data. It further provides an applicable and critical ‘existential typology of digital human twins’ that offers ways to understand and respond to datafication experiences and the emerging relationships of our datafied doubles. Further, it offers means of recognising ‘human-centred values’ or their lack in the technological design and policy frameworks by challenging possibilities for user agency.
Hao, Karen. ‘Dreams of Modernity.’ Empire of AI: Inside the Reckless Race for Total Domination, Penguin, 2025, pp. 46–57.
In Chapter 4, Karen Hao charts the history of AI development. She locates AI within a historical pattern of technological progression that primarily serves the interests of elites and exploits more vulnerable sections of society. She observes how entangled AI research and commercial interests are; AI companies frequently poach academics by offering large salaries. Consequently, these talented individuals are working to fulfill the commercial interests of AI companies, and academia loses valuable resources in its ability to engage critically with this new technology. Hao also questions whether the limitations are inherent due to its ‘connectionistic’ structure (the probabilistic pattern recognition system of training machine learning, and how the majority of contemporary AI programmes operate) and whether this system necessitates a development ceiling. She proposes that no matter how much data connectionist AI models learn from, they can never ‘know’ anything. The ‘black box’ (the AI’s connection centre) becomes more and more opaque as more data is analysed, so the programme’s decision-making process cannot be examined. Contextually, she observes how the scaling of AI development is entangled with the political interests of national governments and state funding; consequently, the race to develop places constraints on opportunities for more diverse research.
Hao, Karen. ‘Plundered Earth.’ Empire of AI: Inside the Reckless Race for Total Domination, Penguin, 2025, pp.121–133.
Chapter 12 exposes the extent and scale to which tech companies from the Global North exploit and extract resources from the Global South. Hao shines a light on the physical spaces that generative AI demands for its construction, functioning, and training while also exposing the narrative misconceptions pushed by the global north and big generative AI companies onto the global south. These company narratives celebrate concepts of progress when associated with technology in order to justify the perpetuation of colonialism embedded in these, thus exposing how the AI industry is rooted in a colonial ideology. The author focuses on two specific case studies: Chile and Uruguay, to additionally showcase the environmental impact of how these countries became new grounds for extractivism. Data centres powering generative AI are built at massive scales, draining important resources such as potable water to cool down the computers (potable, since unfiltered water carries minerals and contamination that might corrode and damage the computers) and huge amounts of energy to power these computers. The scale of these centres not only affects the nearby communities’ land and resources, but it also impacts their aural environments, where citizens have commented on the buzzing sound emitted by these data centers.
Hartzog, Woodrow, and Jessica M. Silbey. ‘How AI Destroys Institutions.’ SSRN Scholarly Paper, no. 5870623, 2025. papers.ssrn.com/sol3/papers.cfm?abstract_id=5870623
In ‘How AI Destroys Institutions,’ Woodrow Hartzdog and Jessica M. Silbey discuss the destructive potential of AI on civic institutions, which, they stress, form the backbone of democratic civilisation. These institutions, the free press, universities, and the rule of law, beyond organisations, have an ability to evolve and adapt within a hierarchy while maintaining legitimacy. Crucially, AI systems erode expertise, decision-making, and promote social isolation; they function in ways that prioritise hyper-efficiency over natural evolution. Created through social construction, institutions’ legitimacy is gained through human behaviour. Legitimacy, for Woodrow and Silbey, is a two-way process between the imposition of institutional norms and how they take shape in human behaviour. AI systems, they argue, rupture this process. Built to reproduce and amplify existing patterns and bias, AI’s illusion of reliability encourages skill atrophy and cognitive offloading. Equally, their reliance on human input and failure to exert intellectual risk disregards the complex, unpredictable nature of human systems. Simply, AI systems can only look backwards; they lack the authority to perform critical decision-making, respect institutional ground truths, or take part in the human connection our institutions depend on.
Ho, Jerlyn Q. H., et al. ‘Potential and Pitfalls of Romantic Artificial Intelligence (AI) Companions: A Systematic Review.’ Computers in Human Behavior Reports, vol. 19, 2025, p. 100715. doi.org/10.1016/j.chbr.2025.100715
This systematic review interrogates how Artificial Intelligence has become more integrated into daily life and how individuals have increasingly turned to AI-driven systems for emotional support, companionship, and even romantic relationships. A total of 23 articles are identified from the following databases: EBSCOhost ERIC, EBSCOhost PsycInfo, PubMed, Scopus, and Web of Science. Results highlighted the key potentials of being in a romantic relationship with AI companions as the facilitation of personal growth and well-being; provision of emotional connection and perceived social support; availability of customisation options; and ability to form a sexual connection, as well as a possible outlet for users to seek entertainment and stress-relieving companionship. However, the phenomenon also raises concerns, as it may lead to users’ over-reliance and susceptibility to manipulation from the chatbot, perceived shame from stigma surrounding romantic-AI companions, risk of personal data misuse, erosion of human relationships, perpetuation of biases, erosion of emotional connection from abrupt system updates and technical glitches, discomfort from uncanny valley effects, or concerns surrounding coercion to respond and early exposure to sexual content. This systematic review synthesises the dual-edged nature of romantic relationships with AI companions and outlines critical avenues for future research.
Klein, Laura F., and Matthew K. Gold. ‘Introduction: Digital Humanities: The Expanded Field.’ 2016. Debates in the Digital Humanities, University of Minnesota Press, 2016, pp. xi–xiv. www.jstor.org/content/oa_chapter_edited/10.5749/j.ctt1cn6thb.3?seq=1
In this introduction to the field of digital humanities (DH), Klein and Gold discuss how the field of DH has evolved from the first volume of Debates in the Digital Humanities. They discuss the concept of the ‘big tent,’ in which DH is viewed as an expansive field that includes an extreme diversity of methods and focuses. They also discuss the responsive concept of the ‘expanded field,’ which is constructed by the relationships amongst key concepts in DH rather than DH functioning as an umbrella term. The expanded field view of DH is one that ‘informs and is informed by allied disciplines’ —as an extension of this, Klein and Gold state that this version of Debates will not separate pedagogy from the topics they are discussing teaching and instead bake the idea of teaching into the learning of a particular topic. Klein and Gold’s introduction to DH presents a field that is best understood through its interactions with other fields and thus introduces a volume that brings interactive concerns to the forefront of DH.
Nisirim, Onyema. ‘Hallucinations in Artificial Intelligence and Human Misinformation: Librarians’ Perspectives on Implications for Scholarly Publication.’ Folia Toruniensia, vol. 25, 2025, pp. 79–98. doi.org/10.12775/FT.2025.004
Onyema Nisirim examines how AI hallucinations, anomalies, and errors that appear in AI-generated text are undermining the validity of scholarly publication. Nisirim is a lecturer and specialist in Library and Information Science, and therefore focuses specifically on the perspective of librarians, and how information professionals can mitigate risks of AI-generated misinformation. The journal article is a study of librarians’ perceptions on the implications of hallucinations. The challenges that are identified in the article include limited training in AI technologies, a lack of institutional guidelines for verifying AI-generated information, and the rapid evolution of AI tools that outpace policy development. However, with these findings Nisirim also provides examples of how skills developed by librarians can combat misinformation. These include promoting AI literacy, implementing training programs, and advocating for clearer guidelines on AI use in academic writing. The article concludes that librarians play a critical role in safeguarding the integrity of scholarly communication and recommends ongoing professional development to help them respond effectively to emerging AI-related challenges. The article, while exploring the threat of AI hallucinations and misinformation, suggests that skills and strategies can be employed to maintain long-term credibility in scholarly publishing.
Prophet, Jane. ‘My More-than-Human Digital Twin: Embodiment, Feminist AI, and the Struggle for Representation.’ AI & Society, 2025. doi.org/10.1007/s00146-025-02659-2
In this article, Prophet critically examines the author’s process of creating an identical-looking self-portrait —a digital twin using a generative AI model, RunwayML, as well as voice cloning and text-to-speech synthesis. The paper is a scholarly reflection on a part of the artist’s process to create the project HerbAIrium. By employing practices of autotheory and feminist technoscience critique, the author studies the process of generating an image intended to reflect her and analyses the misrepresentations embedded in the generated results. This study outlines biases built into AI’s training data and how they are further carried into generated outputs. Prophet emphasises artists’ potential to interrogate AI’s ‘black box’ and to create counter-narratives. The personal experience is intertwined here with critical theory to question the generative materialisations of identities and embodiments and to showcase the limits of the artist/author’s agency. As a result of the demonstrated process, Prophet draws attention to feelings of alienation, unfamiliarity, and misrecognition stemming from the generated images that reinforced binary gender representations, and hyper-femininity while reinforcing prioritisation of youth aesthetics. This essay is a practical and creative engagement with the notion of a personal digital twin, understood as an AI-generated version of a person.
Risam, Roopika. ‘What Passes for Human? Undermining the Universal Subject in Digital Humanities Praxis.’ Bodies of Information: Intersectional Feminism and Digital Humanities, edited by Elizabeth Losh and Jacqueline Wernimont. University of Minnesota Press, 2018, pp. 39–56. dhdebates.gc.cuny.edu/read/untitled-4e08b137-aec5-49a4-83c0-38258425f145/section/34d51cdb-2a89-4e4b-9762-bf6461cf0bb7#ch03
This chapter outlines key areas of debates associated with human interaction with large language models. The chapter cites the Turing test, proposed by Alan Turing as a way of evaluating computer-generated natural language, and uses the example of artists Zach Blas and Jemima Wyman’s video installation ‘im here to learn so :))))))’ in response to Twitter’s built-in AI chatbot. The article stresses a range of issues at stake in the development of machine learning and natural language processing algorithms intended to imitate ‘human’ speech and behavior online. Alongside this critical engagement, the chapter outlines that developments in computing technology have influenced investigations about the nature of humanity as well, considering technology blurring the boundaries between human and machine. It argues computing is increasingly focused on replicating human processes and urges users to understand the assumptions subtending their development. The chapter argues that language processing software has not received the attention it requires to ensure that digital humanities projects are not unthinkingly reproducing the normative white, male, European subjectivity. It concludes that while artificial intelligence wants to represent universal ‘human’ intellectual processes, it is only representative of a fictive ‘universal’ model of human cognition.
Salter, Anastasia, and Mel Stanfill. ‘Game Studies, Endgame?’ Debates in the Digital Humanities, edited by Matthew K. Gold and Lauren F. Klein, University of Minnesota Press, 2023, pp. 261–72. dhdebates.gc.cuny.edu/read/debates-in-the-digital-humanities-2023/section/328947d3-8bf1-4d99-9a51-b5ab3b330af9#ch17
Salter and Stanfill argue that video game studies are the ‘canary in the gold mine’ for broader digital humanities. They argue that the debates ongoing within video game discourse are representative for broader discourse in the field as a whole. They do this in 3 ways. Firstly, they highlight the toxic manosphere of ‘Gamer Gate,’ showcasing the need to understand how we communicate in online spaces (particularly between academic and non-academic) and how we need to consistently be self-critical to promote intersectionality and diverse perspectives. Secondly, they highlight the working conditions of many video game companies, demonstrating how we need to shift our perspectives on understanding fair labour lest it become increasingly exploitative and inaccessible. Finally, they highlight the progress and pledges made by a variety of video game companies, showcasing DH’s interdependent relationship with the degradation of the environment, and attacking that very complicity. Ultimately, the digital humanities field is diverse, but the large scope of the field of video games allows us to understand important insights that every aspect of the digital humanities will have to face.
Scharre, Paul, et al. ‘AI Safety Concerns and Vulnerabilities.’ Artificial Intelligence: What Every Policymaker Needs to Know, Center for a New American Security, 2018, pp. 11–16. www.jstor.org/stable/resrep20447.7
This research report aims to provide a general overview of the major issues and vulnerabilities of AI systems and the reasoning behind them. Various problematic functions and manners of operation are identified, ranging from the inability to operate outside of the environments in their training data; unpredictability and a lack of explainability to a human overseer; and even vulnerabilities to malicious actors. A recurring theme underlying many of the problems identified in AI system operations is that the manner in which an AI may arrive at a solution is impractical for human use and may be exploitable or cause damage. A strong example of this is a collection of ‘reward hacking’ examples, wherein the AI system has exploited loopholes or unintended reasoning to achieve the goal that has been set. This may manifest in low-stakes environments, such as a Tetris-playing AI pausing the game to infinitely prolong it and therefore never lose, or it may lead to damaging consequences, such as a cybersecurity-managing AI locking humans from its system after categorising them as major introducers of malware, or even taking the system offline completely to ensure it remains safe from attack. The report concludes with a warning of the risks of AI performance being a focus while AI safety concerns and vulnerabilities are forgotten and calls for further AI safety research and implementation before practical use.
Shu, Matthew, et al. ‘How Latent and Prompting Biases in AI-Generated Historical Narratives Influence Opinions.’ PNAS Nexus, vol. 5, no. 3, 2026, pg 22. doi.org/10.1093/pnasnexus/pgag022
This study investigates whether AI-generated explanations of historical events can influence people’s political opinions, even when the information provided is factually accurate. The researchers examined two types of bias in large language models (LLMs): latent bias (unintentional bias embedded in the model) and prompting bias (bias introduced intentionally through prompts). In a survey experiment with 1,912 participants, readers were randomly assigned to view summaries of two-twentieth century US historical events — the 1919 Seattle General Strike and the 1968 Third Liberation Front student protests. These summaries were either taken from Wikipedia or generated by GPT-40 with default, liberal, or conservative framing. The results showed that the default AI summaries, even when factually accurate, produced more liberal opinions compared to Wikipedia, demonstrating its latent bias. Yet this default framing did not affect liberals, moderates, or conservatives differently from the Wikipedia summaries. However, AI summaries with liberal framing increased liberal opinions across all ideological groups, while conservative-framed summaries led to more conservative opinions, primarily among participants who already identified as conservative.
Tian, Jinrui, and Ronghua Zhang. ‘Learners’ AI Dependence and Critical Thinking: The Psychological Mechanism of Fatigue and the Social Buffering Role of AI Literacy.’ Acta Psychologica, vol. 260, 2025, p. 105725. doi.org/10.1016/j.actpsy.2025.105725
This study aims to assess the cognitive implications of AI dependence in higher education settings, collecting self-reported data on the critical thinking skills of 580 Chinese university students. The research is informed by Social Cognitive Theory (SCT), which highlights the reciprocal interactions between personal traits, behaviors, and environmental factors, and Cognitive Load Theory (CLT), which explains how excessive information demands can deplete cognitive resources. The results depict a correlation between higher AI dependence and lower critical-thinking skills. Within these findings, cognitive fatigue is understood as a mediating factor, with AI over-reliance leading to mental exhaustion by cognitive overload (being overwhelmed by the volume of AI-generated information) but also cognitive underload (through delegating tasks to AI). Information literacy is presented as a moderating factor, higher literacy buffering the negative impact of AI dependence on critical thinking while simultaneously amplifying cognitive fatigue. Highly literate students are more likely to critically evaluate, cross-reference, and refine AI outputs, with this vigilance leading to increased mental workload and fatigue. Concluding on a practical note, it emphasises the necessity of AI-literacy initiatives and instructional approaches that effectively address cognitive load, particularly as AI becomes more integrated into education settings.
Ward, Megan, and Wisnick, Adriani. ‘The Archive after Theory.’ Debates in the Digital Humanities, edited by Matthew K. Gold and Lauren F. Klein. University of Minnesota Press, 2019, pp. 200-05. dhdebates.gc.cuny.edu/read/4805e692-0823-4073-b431-5a684250a82d/section/a8eccb81-e950-4760-ba93-38e0b1f2b9d0#ch18
In ‘The Archive After Theory,’ Ward and Wisnicki trace the emergence of ‘speculative collections’ in the postcolonial archive, stressing the importance of future-oriented archival practices (where scholars act as ‘co-creators’ rather than ‘producers’) in place of archival practices centred around an ethos of repair. In distinguishing the digital archive, a ‘reactive entity’, that has the theoretical and technological capabilities to account for its own authorial logic to earlier physical archives, Ward and Wisnicki call for a ‘double awareness’ of both the imperial conditions in which it was collected and the impact of the archive at the time of publication. Taking a solely reparative stance, they argue, risks future manifestations of the oppression under which they were created; repairing the damage of the archive requires new technological and institutional infrastructures that acknowledge imperial complicity. In the Archive after theory, an attempt is made to both acknowledge the pain caused by the imperial archive and, through digital mediation and critical theory, create sites of resistance, preserving and decontextualising underrepresented voices in imperial materials. The combination of postcolonial theory, ‘bottom-up’ archival work (created by the communities the archive represents), and digital tools, built with potential colonial complicity in mind, allows humanistic critique to conceive new archival practices. In doing so, speculative collections (revealing the past in a way that serves future communities) reject the understanding that present affairs are logical or inevitable results of the archival past. Applying an understanding of the diverse identities of the present to the post-colonial archival past creates an opportunity to reveal misinterpreted, unseen, or ignored histories embedded within these materials. The archive after theory stretches beyond the lens of retrospect and attempts to conceive an alternative future.
