Book review

Review of Omri Asscher. Machine Translation and Translation Theory

New York and Abingdon: Routledge, 2026. vi + 172 pp.

The last decade has witnessed a profound transformation in the practice and theorization of translation. The emergence of Neural Machine Translation (NMT) and, more recently, the proliferation of generative AI tools have altered not only the workflow of translators but also the very ontology of translation as a human-centred act. Omri Asscher’s Machine Translation and Translation Theory addresses this paradigm shift with admirable clarity and scholarly depth. His book situates the technological evolution of translation within a long intellectual lineage, asking to what extent traditional theories of translation remain adequate — or are rendered obsolete — when confronted with machine translation (MT) as both a tool and an autonomous agent.

Asscher’s central concern lies in exploring, as he puts it, “the subtleties of the intersection between the two: the places where MT corresponds well with the ideas that have been developed on human translation throughout the years, and the places where MT seems to challenge translation theory” (2). The volume thus offers not merely a survey of MT developments, but a theoretical inquiry into what it means to translate in an age where language mediation is increasingly performed by machines. Asscher draws on an impressive command of both the theoretical and ethical discourses of translation, blending conceptual sophistication with empirical insight.

From the outset, Asscher situates his work within a wider socio-technological context in which artificial intelligence is reshaping professional, social, and personal life. The book takes seriously the disquieting proposition that translation theories — developed for, and predicated upon, human cognition and agency — may no longer suffice in the post-humanist age. Without sensationalism, Asscher pushes his readers into intellectually uncomfortable terrain, probing how far humanistic frameworks can stretch before they collapse under the weight of automation.

Following the European Commission’s (2018European Commission 2018 “Communication from the Commission : Artificial Intelligence for Europe — COM(2018) 237 final”. Brussels: European Commission. https://​eur​-lex​.europa​.eu​/legal​-content​/EN​/TXT​/?uri​=CELEX%3A52018DC0237European Commission 2018 “Communication from the Commission : Artificial Intelligence for Europe — COM(2018) 237 final”. Brussels: European Commission. https://​eur​-lex​.europa​.eu​/legal​-content​/EN​/TXT​/?uri​=CELEX%3A52018DC0237, 1) definition of AI as systems that “display intelligent behaviour by analysing their environment and taking actions — with some degree of autonomy — to achieve specific goals,” Asscher treats MT as an instance of artificial intelligence rather than a mere technological aid. This framing allows him to explore not just functional parallels between human and MT, but also ontological questions about autonomy, decision-making, and ethics.

Empirical realities underscore his theoretical inquiry. Already by 2024, more than ninety-nine percent of all translated texts were produced by machines rather than humans (Pym and Hao 2024Pym, Anthony, and Yu Hao 2024 How to Augment Language Skills: Generative AI and Machine Translation in Language Learning and Translator Training. New York: Routledge. Pym, Anthony, and Yu Hao 2024 How to Augment Language Skills: Generative AI and Machine Translation in Language Learning and Translator Training. New York: Routledge. , xxi). By 2025, the global information landscape had tipped toward a new era in which machine-generated texts, whether translations or original compositions, outnumbered those authored by humans. In this environment, language itself, long considered the quintessence of human interaction, increasingly serves as a medium of machine-to-machine communication. For Asscher, this displacement of human linguistic agency necessitates a renewed engagement with what makes translation a distinctively human practice: the capacity for aesthetic judgment, ethical discernment, and contextual fitness. These capacities, he argues, are precisely what differentiate human translation from its algorithmic counterpart and thus warrant theoretical re-examination.

The book is organized into five core chapters, each structured around a key concept in Translation Studies: definitions of translation, equivalence, aesthetics, ethics, and translation as cross-cultural communication. Within each thematic frame, Asscher juxtaposes traditional theoretical constructs with the defining features of contemporary MT systems, examining where the two converge or diverge. This comparative method not only maps conceptual continuities but also exposes deep fractures between anthropocentric translation theories and the computational logics of NMT and large language models (LLMs).

Each chapter moves from a descriptive account of current translation paradigms to a speculative exploration of what translation was, is becoming, and might yet be. Particularly striking is Asscher’s proposal to reconceptualize MT as a historical agent. Rather than viewing AI translation systems as merely contemporary or emergent technologies, he invites readers to consider them as participants in an evolving historical narrative of mediation and meaning-making. This inversion of perspective — treating MT not as an external disruption but as an integral force within translation history — is both provocative and generative, opening avenues for future theoretical inquiry.

Asscher’s discussion primarily focuses on high-resource language pairs, acknowledging the dominance of tools such as Google Translate and generative systems like ChatGPT, Gemini, and Claude. His analysis distinguishes between deterministic systems (e.g., Google Translate), whose outputs exhibit a degree of structural consistency, and generative systems, whose outputs are non-deterministic, variable, and context-sensitive even for identical source inputs. This distinction leads to what Asscher terms a “one-to-several formula for translation equivalence” (52), which challenges long-held notions of stable correspondence between source and target texts.

For practitioners and theorists alike, this unpredictability raises fundamental questions. If translation equivalence is no longer reproducible, can the very notion of equivalence retain theoretical value? Moreover, as Yamada (2019)Yamada, Masaru 2019 “The Impact of Google Neural Machine Translation on Post-Editing by Student Translators.” Journal of Specialised Translation 31: 87–106. Yamada, Masaru 2019 “The Impact of Google Neural Machine Translation on Post-Editing by Student Translators.” Journal of Specialised Translation 311: 87–106. has shown, post-editors struggle to identify errors in MT outputs that are fluent yet semantically misleading — a phenomenon Asscher revisits to underscore how fluency itself has become an unreliable marker of quality. Beyond these technical issues, Asscher also highlights the profound implications for creative writing and authorship: AI systems now perform not only translation but also paraphrasing, summarization, localization, and even stylistic imitation — something he evocatively terms “Shakespearisation”– whereby these “interactive refractions of MT” (37) blur the boundaries between translation, adaptation, and creative production, demanding a redefinition of the translator’s role.

In response, Asscher calls for Translation Studies to reintroduce conscious agency as a constitutive element of translation. This proposal resonates strongly with post-structuralist and deconstructivist approaches that emphasize indeterminacy, transformation, and the dialogic nature of meaning-making. For scholars aligned with such perspectives, the instability of semantic relations between source and target texts, long celebrated as a site of creativity and resistance, finds an unexpected technological analogue in the probabilistic outputs of LLMs.

The book’s discussion of technological evolution — from rule-based and statistical models to contemporary neural and generative systems — is historically astute and conceptually grounded. Asscher acknowledges that while performance remains uneven for low-resource languages, AI-driven translation continues to advance at a pace that exceeds most users’ comprehension. This technological opacity, he argues, underscores the historicity of meaning: MT systems not only reproduce linguistic conventions but also continually reshape them as their training data evolve. Thus, translation becomes both a mirror and a motor of shifting social and linguistic norms.

A key insight emerges from Asscher’s analysis of MT training and evaluation practices: “Taken together, training, evaluation and corpora-building practices emphasize how a target-oriented approach to equivalence is built into data-driven MT systems’ very structure and worldview” (60). This observation aligns MT’s inner workings with long-standing debates in translation theory regarding domestication and foreignization. By privileging fluency and user comprehensibility, MT systems inherently favour domestication, thereby marginalizing foreignizing strategies that foreground linguistic or cultural difference. In this sense, the architecture of MT embodies a tacit ideological stance, one that prioritizes assimilation over alterity.

Asscher largely ignores another complicating factor, the issue of ‘dirty data’ — the vast corpus of machine-generated texts that feed back into the systems that produced them. This recursive loop amplifies biases, inaccuracies, and homogenization effects, particularly when English serves as the pivot language between low-resource pairs. The computational logic of ‘rubbish in, rubbish out’ thus acquires an ethical dimension, implicating questions of representation, transparency, and trust. Notwithstanding such blind spots, the chapter on ethics is one of the book’s strongest and most original contributions. Recognizing that MT’s datasets and algorithms are effectively opaque to most users, Asscher critiques the ‘black box’ nature of AI translation for its lack of transparency and accountability — qualities enshrined in professional codes of ethics across the translation industry. He advances instead a model of situational ethics, attentive to the text-specific and context-dependent parameters that shape translation decisions. Crucially, Asscher proposes that MT systems themselves function as autonomous ethical agents, capable, however imperfectly, of decision-making processes that bear moral implications (101). While this anthropomorphic framing is provocative, it serves a critical heuristic purpose: it forces the field to rethink the boundaries of ethical responsibility in a hybrid human–machine translation ecology.

Nevertheless, not all aspects of Asscher’s study are equally convincing. Structurally, the book’s final two chapters — devoted to non-human cross-cultural mediation and the history of MT, respectively — would have benefited from earlier placement. They provide foundational definitions and theoretical touchstones (such as distinctions between overt and covert translation, or issues of communicative trust) that would have oriented readers more effectively at the outset. As it stands, key conceptual clarifications arrive belatedly, giving the book an uneven rhythm.

The same can be said about Chapters 2 and 3, which are reproductions of previously published articles. Chapter 2 (“Definitions: Demarcating the Field”) incorporates material from Asscher (2023)Asscher, Omri 2023 “The Position of Machine Translation in Translation Studies: A Definitional Perspective.” Translation Spaces 12 (1): 1–20. Asscher, Omri 2023 “The Position of Machine Translation in Translation Studies: A Definitional Perspective.” Translation Spaces 12 (1): 1–20. , and Chapter 3 (“Equivalence: Target-Orientedness to the Rescue”) does so from Asscher (2024)Asscher, Omri 2024 “The Explanatory Power of Descriptive Translation Studies in the Machine Translation Era.” Perspectives: Studies in Translation Theory and Practice 32 (2): 261–277. Asscher, Omri 2024 “The Explanatory Power of Descriptive Translation Studies in the Machine Translation Era.” Perspectives: Studies in Translation Theory and Practice 32 (2): 261–277. . While they flow nicely within themselves, they suffer from minor issues of repetition in this new context as chapters of a book.

On a stylistic level, the frequent use of acronyms occasionally hampers readability. Terms like ‘DTS’ (Descriptive Translation Studies), though familiar to specialists, are not consistently unpacked, and the use of ‘S’ to denote ‘Studies’ may confuse readers accustomed to its more common associations with ‘source’ or ‘statistical’. Moreover, minor typographical errors — such as the German idiom “Der Preis ist heiß” (misspelled “heis,” 75) and “Dietsch” instead of “Ditsch” (131) — detract slightly from the text’s polish. While these are minor editorial oversights, they ironically echo one of the book’s central themes: the tension between human oversight and automated production in language work.

Such shortcomings, however, do little to diminish the intellectual force of Asscher’s argument. Machine Translation and Translation Theory succeeds in bridging disciplinary divides between linguistics, computer science, communication, and Translation Studies. It offers both a lucid introduction for students and a sophisticated theoretical provocation for early-stage and established researchers. Its implications reach far beyond Translation Studies narrowly conceived, engaging with larger philosophical questions about language, cognition, creativity, and the ethics of automation.

Ultimately, Asscher compels us to confront an unsettling but necessary question: if translation is no longer the exclusive province of human beings, what remains of its humanistic core? His answer is not to lament the loss of control but to reimagine the field’s conceptual foundations in light of technological reality. By situating machine translation within the intellectual genealogy of translation theory, rather than outside or against it, Asscher offers a framework that is both critical and forward-looking.

In a world where machines increasingly write, translate, and interpret for us, Machine Translation and Translation Theory stands as a timely and incisive contribution. It challenges scholars and practitioners alike to rethink what translation means, who or what performs it, and how we might continue to theorize it without recourse to comforting humanist certainties. Herein Asscher’s book differs from a plethora of other publications concerning AI and MT. He has opted for a conceptual approach. Not the inner workings of technology are his focus but the interface between machine and human/ities. Asscher presents a coherent argument across these themes and ethical conundrums, tackling theoretical implications well worth reading and thinking about further.