Paper 1:

Abstract:

As artificial intelligence increasingly reshapes the ways in which texts, images, audiovisual

materials and interactive experiences are created, translated, and analysed, the humanities

face both unprecedented opportunities and challenges. This paper interrogates the intersection

of AI-driven translation and the practices of distant reading, viewing, and perceiving, a triad

of methodological lenses proposed by Franco Moretti (2013), Lauren Tilton, and Taylor

Arnold (2019). The availability of large-scale linguistic and multimodal datasets, coupled

with powerful generative models, enables researchers to explore materials across languages,

media, and geographies at scales previously unimaginable (Arnold & Tilton, 2023). Such

methods may offer new insights into how different audiovisual materials travel translingually

and transmedially.

Yet this expansion of scale is accompanied by pressing epistemological and ethical

questions. The reliance on algorithmic systems for translation and research foregrounds long-

standing debates over authorship, authenticity, and interpretation (Baker, 2025), while also

introducing new anxieties about ethics, bias and opacity (Li, 2023; Saunders & Byrne, 2020;

Prates et al., 2019). Moreover, as AI translation systems increasingly mediate media

circulation between languages and audiences, questions of fidelity, creativity, culture and

agency come sharply into focus (Cao et al., 2025; Kirk et al., 2025; Tao et al., 2024).

Positioned within the emerging discourse of digital humanities, particularly in Asia

and transnationally (Arnold & Tilton, 2024), this paper takes a kaleidoscopic view of distant

reading, viewing, and perceiving in the age of AI. Rather than treating AI solely as a tool, it is

becoming a collaborator, a mediator, and at times, a constraint that forces renewed reflection

on the limits of quantification and the necessity of human judgment (Smits & Wevers, 2023).

Drawing on case studies, the paper foregrounds both the transformative potential of AI for

enabling more plural, comparative, and accessible forms of linguistic and cultural transfer

study, and the critical need for ongoing ethical vigilance.

By examining how AI complicates and extends the practices of distant engagement

with audiovisual and multimodal creations, the paper contributes to debates at the intersection

of translation studies, media theory, and digital humanities (Frontoni et al., 2024). In tracing

both the possibilities and the limits of AI-assisted translation research, it aims to demonstrate

how scholarly practice in the age of AI must negotiate the double demand of scale and

critique, mapping broader fields while maintaining reflexive awareness of the interpretive

stakes at hand.

Paper 2:

From Characters to Calligraphy: Multimodal Distant Reading of Chinese Poetry in the Age of AI

This paper examines how the convergence of large-scale cultural databases and multimodal AI analysis reshapes our understanding of Chinese poetry across history. Building on Franco Moretti’s concept of distant reading and extending it toward visual and performative dimensions, I propose a framework for multimodal distant reading that allows poetry to be perceived not only as text, but also as sound, image, and embodied practice.

Drawing on the China Biographical Database (CBDB) and related digital corpora, I situate poetic texts within their broader social, cultural, and historical networks. At the same time, I employ natural language processing, image recognition, and handwriting analysis to examine how poems circulated through calligraphic manuscripts, anthologies, and performance. By integrating semantic, phonological, and visual dimensions, this approach attends to meaning, sound, and form together, bridging computational analysis with long-standing interpretive traditions in Chinese literary scholarship.

Case studies include stylistic shifts across dynastic transitions and the impact of political upheavals such as the An Lushan Rebellion. I also explore the visual features of poetic calligraphy and their relationship to textual meaning, showing how handwriting analysis and image recognition technologies can illuminate the material and aesthetic dimensions of poetic transmission.

In highlighting both the insights and the limitations of multimodal AI methods, this paper argues that large-scale cultural analytics must remain accountable to the ethical and epistemological challenges posed by cultural memory and aesthetic heritage. By combining methodological innovation with close engagement in Chinese literary culture, it demonstrates how multimodal distant reading can expand the scope of literary history while sustaining the critical reflexivity central to humanistic inquiry.

Paper 3:

Augmented Reading: Retrieval-Augmented Workflows for Literary Historiography in the Age of AI

In digital humanities, the metaphor of distance has shaped how scholars imagine the relationship between scale, evidence, and interpretation. Close reading privileges attention to singular texts; distant reading (Moretti) emphasizes patterns across large corpora; distant viewing (Tilton and Arnold) extends this to visual and media archives. Each has functioned not only as a method but as a conceptual touchstone, a way of asking what becomes visible — and what is lost — when mediation is scaled and reframed. Yet as artificial intelligence and large language models increasingly structure our engagement with cultural data, we are entering a new interpretive terrain that neither close, distant, nor viewing fully captures. This paper proposes Augmented Reading as a third mode of literary interpretation: a practice in which human analysis is scaffolded by retrieval-augmented generation (RAG) workflows that reshape the evidentiary surface of cultural materials.

Augmented Reading differs from earlier metaphors in three ways. First, rather than a question of proximity (close vs. distant), it names a relationship of collaboration: the critic reads alongside and against machine-mediated retrieval, using AI systems not as authorities but as scaffolds that surface connections, contexts, and resonances. Second, it highlights the workflow dimension of contemporary AI research. In contrast to the one-off distant reading chart or visualization, augmented reading depends on a repeatable pipeline: retrieval of relevant archival fragments, generation of contextual interpretations, error audits to expose bias, and iterative refinement. Third, it insists on methodological accountability. Because retrieval systems are subject to selection bias, data imbalance, and opacity, augmented reading requires scholars to document how corpora are prepared, how retrieval is configured, and how interpretive claims are constrained by machine mediation.

To demonstrate the potential and limits of augmented reading, the paper presents a case study of Irish Troubles poetry (1960s–1990s). This corpus is uniquely situated for AI-assisted interpretation: its poems register deep layers of cultural trauma, memory, and political conflict, yet they are often entangled with paratexts, testimonies, and archival events that exceed the boundaries of the poems themselves. Using a prototype workflow (AR-TRACE: Augmented Reading via TRACE), the project retrieves materials from the CAIN archive of Northern Irish conflict, pairs them with poetic texts, and prompts LLMs to generate contextual framings of themes such as mourning, resistance, and community fracture. Rather than treating the model’s outputs as interpretive truths, the workflow treats them as evidentiary scaffolds — points of entry for humanistic analysis. For example, a retrieved cluster linking a Heaney elegy to contemporaneous accounts of Bloody Sunday does not substitute for close reading of the poem, but it augments the reading by situating affective registers within a broader temporal and cultural frame.

The case study also foregrounds the risks of augmented reading. Retrieval bias may privilege certain events over others; generative summaries may flatten or distort affective nuance; and over-reliance on machine scaffolding can obscure the singularity of literary expression. These risks are not incidental but constitutive, raising urgent ethical questions about authority, authenticity, and the politics of cultural memory. Accordingly, the paper argues for embedding error audits and protocol documentation into augmented reading workflows, so that interpretive claims remain accountable and reproducible.

By formalizing augmented reading, the paper contributes to ongoing debates in digital humanities about the interpretive limits of machine-assisted analysis. It situates the proposal in dialogue with Moretti’s call for distant reading, Tilton and Arnold’s framework for distant viewing, and emerging discussions of “distant perceiving” in the age of AI. In doing so, it makes three claims. First, that augmented reading provides a vocabulary for understanding how scholars might responsibly collaborate with AI without collapsing interpretive authority into the machine. Second, that retrieval-augmented workflows offer a method for bridging cultural analytics with literary historiography, allowing trauma narratives and cultural memory to be studied at scale while retaining contextual nuance. And third, that by aligning itself with the concordance tradition of Josephine Miles, augmented reading extends the genealogy of DH methodologies into the present AI moment, showing how longstanding practices of scaffolding human reading with machines can be reimagined today.

Ultimately, Augmented Reading is less a replacement for close or distant reading than a complement to both. It offers scholars a way to inhabit the new epistemological space opened by AI: a space where perception is mediated, where interpretation is scaffolded, and where the task of the critic is not to surrender judgment to machines but to document, contest, and reframe what machine-assisted perception makes visible.