Digitalisation, neo-Taylorism and translation in the 2020s

ABSTRACT

This paper aims to provide a cursory yet critical overview of digital transcultural communication in the third decade of the new millennium. Today, digital entrepreneurs create lucrative business models based on user data whilst simultaneously enticing individuals to get used to working with these digital models. We will relate these two commercial objectives to modern translation practices, interweaving our discussion with the growing debate on the computer-assisted exploitation of the workforce. Digital neo-Taylorism is especially rife in the global translation industry, which today is overwhelmingly located on globally interconnected virtual online platforms, where largely underpaid professionals tend to be exploited by means of non-creative yet laborious post editing tasks or crowdsourcing translation activities. By linking the three values speed, efficiency and quantity to neural machine translation, postediting translation and crowdsourcing translation, we embark on a cursory critical discourse analysis, partly to show digital neo-Taylorism at work in the language used to describe new digital translation processes. In doing so, we endeavour to account for the underlying sociotechnical relationships across wider systemic processes such as digitalisation, alienation and capitalist social relations, and their effects on language and translation professionals.

KEYWORDS:

• Digital neo-Taylorism
• neoliberalism and translation
• neural machine translation
• postediting translation
• collaborative translation
• crowdsourcing translation

1. Introduction

This article aims to reflect upon the social and cultural effects of new technologies on the translation profession within the context of neoliberal capitalism. Rapid technological development deeply influences the language and translation industries, but critical–philosophical reflections on wide-ranging ethical and societal repercussions remain far in between (Cronin, 2020; O’Thomas, 2017). In this article, admittedly, we aim to brush with a broad stroke, in an attempt to sketch a rough outline concerning the effects of digitalisation, AI technologies and computer-based translation tools on transcultural mediators and on the wider professional and industrial landscape. AI technologies are primarily expected to decrease cost and increase productivity, with demands for speed, efficiency and the throughput of large amounts of data being the dominant considerations amongst most LSP-providers. Our aim is less to provide a nuanced empirical-descriptive overview, but rather a critical exploration as to the potential long-term consequences concerning the use and abuse of language and translation data in the pursuit of commercial profit. The key question to be addressed concerns the general impact of digitalisation on industrial processes – i.e., translation flows – within the modern translation workplace. We endeavour to account for the underlying sociotechnical relationships across wider systemic processes such as digitalisation and neo-Taylorism, and their effects on language and translation professionals. We will initially describe the critical concepts of digitalisation and neo-Taylorism, which will be setting the scene for a deliberation on, first, the inherent connection between the pioneering AI-based technology of neural machine translation and the notion of speed as a philosophical construct; second, the activity of postediting translation plays an increasingly important role in the modern translation workflow, in particular in regards to the growing necessity for maintaining efficient workflows; and finally, we will relate the emerging phenomenon of crowdsourcing translation to the associated production value of quantity, given that modern workflows are increasingly dictated by the need for generating ever larger amounts of translation output in the form of linguistic data assets.

2. Digitalisation and neo-Taylorism

History progresses, as some might argue, along the lines of successive economic revolutions, each one ostensibly offering a brighter, more prosperous future for humankind. Let us briefly recall these apparent revolutions. If revolution 1.0 was based on the invention of the coal and steam engine, revolution 2.0 took shape through capitalist mass production, the electrification of society, chemical engineering and the industrial refinement of oil, then revolution 3.0 emerged through the onset of computing technology in connection with the automation of production and distribution cycles (Popkova Elena et al., 2019). Lastly, what is to be expected from revolution 4.0, the apparent catalyst of digitalisation, cyber-physical systems, artificial intelligence, and eventually quantum technology? Klaus Schwab (2016, pp. 11–13), founder and executive chairman of the World Economic Forum, claims that the fourth industrial revolution would be completely different from previous socioeconomic transformations because it is characterised by unprecedented developments in genetics, artificial intelligence, robotics, nanotechnology and biotechnology, affecting all disciplines, economies and industries, and even the ways in which we see ourselves as human beings. In particular, AI technologies reach far beyond human problem-solving capabilities, but they are primarily geared towards rationalising and optimising a capitalist society built on neoliberal principles, where values such as speed, precision and profit outweigh the urgent need for socioeconomic equality and spiritual wellbeing. The bulk of neoliberalism’s proponents, however, expect artificial intelligence and machine learning technologies as heralding golden times for economic growth, prosperity and trade (Harribey, 2022). It is in this sense ironic that global expenditure on the most disruptive technologies, on those lethal ‘mean machines’ falling under the guise of AI-based military technology, is reaching record heights year after year (Tian et al., 2023). The dominant narrative, of course, remains technology’s promise of lifting humanity out of misery, suffering and toil. Techno-medical advancements might have mitigated physical suffering caused by sickness, disease and accident, yet the crucial philosophical question remains as to whether all these brave new technologies really lead us to the land of everlasting progress and freedom.

In this paper, we will take the global translation industry as a case in point. Since the dawn of the digital age, the translation industry and its associated professions have undergone extraordinary transformations and ruptures. Peter Sandrini (2017) speaks in this context of the onset of Translation 4.0, a phenomenon in a way synonymous with the concept of Industry 4.0 which also relates to the ongoing debates surrounding the positive ideals of a knowledge or information society, but also to the more critical conception of a network society (Castells, 2000). Mainstream computerisation and digital automation bear a multitude of ethical downsides. In the realm of communication technologies, for instance, digital tools create a transparency of communication that causes stress, more surveillance and lower wages for employees (Bartoletti, 2020, pp. 22–25). In her influential work The Age of Surveillance Capitalism, Shoshana Zuboff (2019) identifies two main entrepreneurial objectives by the champions of digital capitalism: (1) create lucrative business models based on user data; (2) get people used to working with these ever-evolving digital models. The second objective rests on the idea of habituation, a new type of power which Zuboff calls ‘instrumentarianism’ and which ‘knows and shapes human behaviour toward others’ ends’ (ibid., p. 8). It is interesting to relate such claims to the translation industry and its professions. As shown in various studies, striving for rationalisation, productivity and efficiency are part and parcel of modern translatorial activity (e.g., do Carmo, 2020; Moorkens, 2020; Sakamoto, 2019). A significant argument in this debate concerns the critical concept of Taylorism, a well-known science-led approach to workplace arrangement, where workers and employees are to be organised alongside rigid principles of labour efficiency and cost-saving measures (Moorkens, 2020). The concept is named after the American engineer and academic Frederick Winslow Taylor, whose major work The Principles of Scientific Management (1911) even today remains the cornerstone of an entrepreneurial ideology that fosters both worker exploitation and capitalist profit generation. A liberal doctrine in theory, its practical components have now widely been adapted to the exigencies of the twenty-first century digital workplace, indeed in the form of ‘neo-tailoring’ the need for efficient and friction-free worker organisation to the dictates of digitalised and AI-enhanced economic competition. In short, the Taylorian model is constantly being adapted to new technologies and practices and thus accounts for a novel digital form of exploitation nowadays being variously described as ‘neo-Taylorism’, ‘computer-assisted neo-Taylorism’ (Gadrey, 2003, p. 51, 127) or as ‘digital Taylorism’ (Günsel & Yamen, 2020).

Modern communication technologies such as the Internet and CAT-Tools have revolutionised professional translation practice. However, in particular since the advent of NMT around 10 years ago, professionals in the language and translation industries are becoming exposed to entirely new ways of management, office organisation and text work. Apart from the automatic and thus non-human transfer processes inherent in machine translation technologies, practices such as postediting translation and crowdsourcing translation are on the verge of transforming the language and translation professions. Neural machine translation systems, for instance, need to be ‘trained’ through human labour (Bartoletti, 2020, p. 68), and they are based on huge corpora of linguistic data that grow with every new translation being generated by the system which, in turn, improves the quality of any successive machine-translated output. A growing legion of translators today, instead of translating from scratch, earn their living by postediting machine-generated translations (Sakamoto, 2019). In addition, online-based ‘digital labour platforms for translation’ (Fırat, 2021, p. 62), with translators from all over the world simultaneously ‘crowd-labouring’ on translation commissions, are gaining in popularity (Schmitt, 2017). According to Joss Moorkens (2020, p. 12), industry 4.0 in the translation business means that ‘large translation companies … break tasks down into smaller chunks and … rigidly define and monitor translation processes’. In the era of digital translation, ‘new technologies enable more varied and invasive monitoring and surveillance of workers to ensure that their role is carried out as expected’ (ibid., p. 16). It is, in this context, fitting to discuss the new realities of translation labour under the umbrella concept of ‘digital neo-Taylorism’. Not surprisingly, this digital form of exploitation increasingly harnesses artificial intelligence in order to boost the efficiency, effectiveness and productivity of industrial processes. Especially relevant here are those problem-solving tasks and work procedures that humans can manage only up to certain degrees and levels of difficulty. Digital neo-Taylorism represents the professional undercurrent of the language and translation industries’ transformation into an AI-based digital economy. Owing to growing international competitiveness, the translation industry is forcing translators to work ever faster as well as more and more efficiently and productively (do Carmo, 2020, p. 36). In the following, we will link the above mentioned three – let’s say neoliberal – values of speed, efficiency and quantity to automatised neural machine translation, postediting and crowdsourcing translation. In doing so, we embark on a cursory critical discourse analysis, partly to show digital neo-Taylorism at work in the language used to describe new digital translation processes.

3. The speed of machine translation

One of the few things in life which are predictable and consistent are societal change and cultural transformation. Predictability and consistency, however, come in different shapes and forms, with today’s globalised societies being confronted with inevitable technological growth and diversification and accelerated societal and psychological transformations (Rosa, 2013). For some, the phenomenon of sociopsychological acceleration has morphed into a ‘general social norm’ (Korunka & Kubicek, 2013, p. 19), for others it even constitutes the driving force of digital modernity (Tomlinson, 2007). Societal acceleration is accompanied by a host of contingent phenomena: if structural change is speeding up, then the individuals making up society also – consciously and subconsciously – accelerate their cognitions and behaviours. People develop new coping strategies, if only to survive the pace of change, by speeding up their actions, resulting in a steady ‘acceleration of the pace of life’ (Rosa, 2013, p. 223). Societal acceleration can be linked to the idea of ‘time–space compression’ (Harvey, 1989, p. 147) in that capitalist globalisation significantly scales down the dimensions of space and time in comparison to earlier epochs, one simple case in point being today’s availability and sheer speed of air travel. In the labour market, societal acceleration and time–space compression are, naturally, reflected in an increasing pace of work as well as a gradual overcoming of procedural and spatial boundaries and temporal constraints. Pietrzak and Kornacki (2020, p. 10, 15), for instance, state that ‘[t]he demand for translation and the speed of delivery grows daily’ and that interactive translation memory tools ‘may hamper the creative potential of an individual since they promote speed and efficiency of translation, which is achievable only in the case of repetitive content that leaves scarce (if any) room for creativity’. Automatised NMT and their AI-based offspring such as ChatGPT, however, represent a truly epochal breakthrough in the acceleration of translation work.

Automatised NMT-processes are firmly established since around 2015 and they allow for a flexible and fast turnaround from delivering linguistic source data to linguistic target – i.e., translation – data (Alonso & Nunes Vieira, 2020). Fast turnaround times are achieved with the help of algorithmic probability calculations scouring through massive linguistic databanks, in connection with concordancing tools, integrated dictionaries and glossaries. NMT-systems rely on large language models (LLMs) consisting of complex Big Datasets, and data are the decisive factor for achieving high quality machine translation. Neural machine translation can thus be described as a model for automatic translation geared towards the speedy translation of large amounts of linguistic data. The widespread availability of NMT enables the language industry to translate significantly more texts than would have been imaginable only a few years ago. Human translation efforts simply would never suffice to translate the ever-growing mountains of textual material in today’s global political economy. To give an example, depending on text type and difficulty, an experienced human translator is capable of translating around 2,000–3,000 words per day without the help of an NMT-system, yet by using postediting techniques, a translator can reach around 7,000 words daily. The combination of NMT with other computer-assisted translation tools allows productivity increases sometimes in the order of more than 150%, with some translators achieving throughputs of more than 7,000 words per day (Stasimioti, 2022, online).

The demand for acceleration and speed is inextricably intertwined with a mythical belief in progress. Professional technology is inseparable from the ideals of an ever-thriving modernity and its illusory march towards steady progress. Standing in the way of these values effectively means opposing a dominant progressivist ideology. It is, of course, all-too-human to strive to circumvent such negative projections by priming oneself to accept all kinds of inconveniences, even though the usual short-notice commissions and high work targets lead to risks of burnout and health problems, while the occasional simple joy of translating barely comes to fruition (cf. Robinson, 2003, p. 30). Normative societal acceleration, as outlined in detail by Rosa (2013, pp. 35–93), is fuelled by the continuous creation of novel time-saving technologies which in turn entice individuals to keep pacing and speeding up (cf. Korunka & Kubicek, 2013, p. 19). Far from simply reducing human suffering and toil, new technologies also create new expectations where everything appears to us as restless and urgent, where there is constant need for quick turnaround times (ibid., p. 21). The dominant neoliberal ideology entices people to believe that constant innovation, competition and development are the cornerstones for technical, economic and even ethical progress (Brune, 2003). Not accepting the dominant ideology of capitalist progressivism literally leaves you in the mud, which is why translators – just like the industrial precariat throughout the world – are both perpetrators and victims at the same time: by improving the translation results of NMT-systems with their own world knowledge, translation professionals (and also the general public at large) tend to pass on valuable data to NMT-companies and to other big industry players. And all this entirely free of charge!

If we relate these practical and ideological considerations to Zuboff’s claim that the biggest industry players create lucrative business models based on user data, then today’s advanced machine translation technology represents more than a gold mine for big capital. Without Big Data, there would be no NMT. Big Data refer to massive sets of information which are too complex, too weakly structured or too fast-moving for being analysable with conventional data processing methods. The emergence of Big Data is, no surprise, largely due to the success of digital tech-giants like Microsoft, Google or Amazon, who have been accumulating vast amounts of data over the years (Fasel & Meier, 2016, p. 5). Big Data is calculated in zettabytes and corresponds to the ‘5Vs’ of volume, variety, velocity, value and veracity. To get an inkling of the sheer amount of data that is being processed in large language models, one zettabyte represents 10²¹ bytes, in other words this is a number one followed by twenty-one times the number zero, representing one trillion or 1,000,000,000,000,000,000,000 bytes. Such volumes of data are barely imaginable for most people, which is why the 5Vs are destined as methods for providing processing infrastructure in order to extract and analyse huge volumes of data (Fasel & Meier, 2016, pp. 5–6). Let us now, apart from these digital facts, look at the societal reality of technological progress. As a catalyst for wealth accumulation, Big Data constitutes a societal fact par excellence, it is sought after by high-tech companies like gold dust, given that big information, as Zuboff (2015, p. 75) maintains, constitutes a ‘new form of information capitalism [that]aims to predict and modify human behaviour as a means to produce revenue and market control’. This is what is happening in the worldwide translation market which is controlled by large companies who take advantage of translators and ordinary people to acquire their data (cf. Carreira, 2023). Zuboff (2019, pp. 136–152) describes the Big Data game as a ‘dispossession cycle’ made up of the four entrepreneurial moves incursion, habituation, adaptation and redirection, a sequence in which professionals and the general public get caught up as hapless data providers. The term incursion illustrates the unsolicited digital invasion of our professional and private lives, mostly by high-tech companies, via phones, emails, social media, etc. The notion of habituation signifies that the incursions steadily creep into the ordinary, they become normalised, simply put, we get used to subliminal digital assaults. Adaptation indicates that high-tech web-based companies keep cultivating new methods and digital technologies in order to meet social and legal requirements. By giving the impression to abide by national and international rules, laws and regulations that would respect people’s private and professional lives, such companies provide a veneer of legitimacy to otherwise ethically questionable practices. Redirection, finally, leads straight into the generation of data revenue, where all our data is processed as ‘behavioral surplus’ (ibid., p. 150), it is gathered and evaluated as raw material for present and future digital interactions. The accumulation of our online-generated behavioural surplus by and large functions as a steady source of lucrative income for big business.

The data generated in any machine translation transaction is monetised. It is worth putting the economic significance of ‘dispossessive translation data accumulation’ in historical perspective with reference to some successive Google blogs. In 2012, an estimated 2.4 billion people used Google Translate, amounting to the equivalent of one million translated books per day (Googleblog.com, 2012). In 2016, 10 years after the launch of Google Translate, more than 6 billion people were using the tool, amounting to the translation of more than 100 billion words a day (Tourovsky, 2016). And by 2021, more than one billion people had installed the Google Translate application (Pitman, 2021). By this time, Google was able to translate content into 109 languages, with technical and translational support by Google Teams and its Translate Community. The latter plays an important role in improving translation quality. In 2016, for instance, 3.5 million people made 90 million – unpaid! – contributions via Google’s Translate Community, exemplifying a well-oiled and highly profitable enterprise venture. Let’s get back to the theme of speed. In 2021, the scientific publisher Springer Nature (2021, n. p.) made the translation tool DeepL AI available for its authors, since this allows for ‘a free, easy and efficient way to have their book translated’ and it ‘gives them the flexibility to write their manuscripts in the language they prefer and significantly expand the audience for their work’. The publisher’s announcement to ensure faster translations whilst simultaneously reducing translation costs provoked a reaction from the European Council of Literary Translators’ Associations, describing Springer’s demonstration of blind faith in artificial intelligence as blatantly unprofessional (Oury, 2021). Springer Nature, however, argues that their dedication to remain competitive, innovative and at the edge of science will generate a greater number of publications for authors who otherwise would barely ever see their work in print (Anderson, 2021). The main argument for using AI-based machine translation, to be sure, lies in saving cost and time (do Carmo, 2020), yet the dominant progressivist ideology, which propagates fast-paced wealth accumulation whilst simultaneously advocating individual freedom and autonomy, leads translators and the public alike to believe – consciously or subconsciously – in some mythical idea of socioeconomic liberation through techno-scientific progress (Onfray, 2019, pp. 12–13, 49).

4. The efficiency of postediting translation

Speed and efficiency are central prerequisites for most work practices in the language and translation industries. Speedy machine translation thus may be seen to constitute one essential requirement for the digitally-versed postmodern translator. Under the régime of technological – some call it immaterial (Fuchs, 2008, p. 208) – labour in the translation industry, efficient postediting, i.e., the human revision of machine-translated texts, is hailed in some quarters as an attractive work proposition. Notwithstanding the key role of speed in postediting, this new type of translation activity primarily satisfies the demand for industrial efficiency. A cursory critical analysis of postediting translation workflows under the régime of digital neo-Taylorism highlights the pathways into which professional translators are becoming habituated. The websites of educational or non-profit organisations frequently emphasise that posteditors are not only expected to cover large amounts of words – or shall we rather say data? – within a small window of time, but also to work as precisely and, above all, efficiently as possible. One of the largest translation schools in France, the Ecole Supérieure de Traduction et Relations Internationales, features a description of postediting translation on its website which tacitly presupposes this activity as equalling to speed and productivity. Phrases like ‘l’activité de post-édition a le vent en poupe’ (engl. ‘postediting activities are booming’; ESTRI, 2023) assume a certain trendiness for postediting translation, peddling it as up and coming, as forward-moving and thus bringing some welcome fresh air to the world of translation. Mobility and flexibility are key, suggesting a surrender to the Taylorian dictates of time, cost and efficiency, nowadays however with a digital neo-Taylorian bent. Introducing students to modern translation practices by means of such playful yet subtly misleading ways appears debatable, given that posteditors are tacitly expected to meet such requirements. Even more worrisome, though, is the unspoken acceptance of cut-throat competition among colleagues, which spurs students onto become the fastest and most stress resistant translation apprentices (Voß et al., 2013, p. 80). In the globalised language industry, competition has been pushed to the limit, which is reflected both in the professional world and in translator education. NMT, combined with postediting translation, is on the verge of entirely changing our understanding of the translation profession. The French Société Française des Traducteurs boasts on its website that it has contributed to championing the ethical case for posteditors. Their imposition of the ISO 18587 standard, they claim, represents real progress for the working conditions of postediting professionals. The standard makes a distinction between the by now two well-known types of postediting, light and full postediting. Whereas light postediting ‘lightly’ corrects machine translation errors, making the text readable whilst still not being entirely correct, full postediting achieves a thorough revision of any machine translation output, making the text highly readable. Full postediting translation is of a comparable quality to human translation. From a critical point of view, however, we should ask whether the imposition of the ISO 18587 standard really does fight the corner for translators as well as the ailing translation profession. On the one hand, the creativity of the translator’s craft and graft is recognised by full postediting, and translators may also ask for a higher fee. On the other hand, light postediting demotes professional translators ‘to the status of fixers of seemingly unintelligent errors’ (O’Brien, 2012, pp. 108–109, based on Krings). It is moreover unsurprising that translators generally tend to perceive postediting as ‘boring and demeaning’, not least because it eliminates a big chunk of creative and thus rewarding labour from their skill set (Moorkens & O’Brien, 2017, p. 110). Such dissatisfaction also relates to the argument that our prototypical understanding of translating – i.e., in the sense of moving from a text’s beginning to its middle and end – only rarely applies to postediting (O’Brien, 2012, p. 114), given that human posteditors primarily are required to edit scattered pieces of text without context (Moorkens, 2020, p. 12).

Leaving aside the dreariness of postediting translation for human translators, it is impossible to dispense with machine translation and postediting in a globalised and digitalised world where the volume of content to be translated is steadily on the rise. Professionals, in any case, subconsciously strive for efficiency, since inefficient task completion is frequently felt to be a source of frustration (Stilijanow & Bock, 2013, p. 157). Significantly, however, subconscious mental processing is tightly interwoven with higher-level ideological cognitions that keep us in line with dominant value orientations. Free-market dogmas such as Time is Money, Competition is Necessary, Consumption is Good for You, and so on, are providing an ideological smokescreen that maybe finds its most prominent literary expression in the works of authors such as Douglas Adams or George Orwell. And why not comparing the situation of translators to Orwell’s (1949, p. 140) notion of doublethink, which the author himself circumscribes as ‘the power of holding two contradictory beliefs in one’s mind simultaneously, and accepting both of them’? The Siamese twins of neural machine translation and postediting undoubtedly lead to more efficient translating, but they also alienate translators – in the true Marxist sense of the word – from the fruits of their labour. In an ideal world, such forgotten fruits should materialise in the form of a sense of achievement that arises out of a creative process of meaning-making (cf. Marais, 2019, p. 22). In this line of Orwellian reasoning, translators (have to) tolerate alienation as long as they can afford to put enough bread on the breakfast table. Two centuries before the digital revolution, Marx himself did not exclusively associate technology with capitalist profit accumulation, since he saw an enormous liberating potential in it. Wendling, for instance, observes that for Marx ‘machinery is really already the emblematic means of production of a liberated society, used to produce material wealth and decrease the time spent in alienating labor’ (2009, p. 175, emphasis original).

In our postmodern, or indeed posthuman times, technology defines and governs our lives, our values, habits and behaviours to an unprecedented extent. And if translators, just like anyone else, reject the ideology of technological progressivism, they exclude themselves from the tables of freelance or wage-earning employment. About three quarters of a century ago, Jacques Ellul (1954/1967, p. 21), in his attempt to lay bare the sociocognitive ‘religious’ subtext of modern technology, claimed that ‘[t]he technical phenomenon is the main preoccupation of our time; in every field men seek to find the most efficient method’. Some years earlier, in the 1940s, and with a slightly different interpretative brush, Orwell had taken note of a new form of domination (Besses, 1984, p. 48), while nowadays an entirely new mode of disciplinary control comes to fruition in the form of techno-managerial strategies put in place by Big Tech companies (Zuboff, 2019). Popular sayings such as You need to move with the Times are emblematic of the tacit agreement to go with the flow of new technology, to uncritically embrace and use it, in all spheres of life. It is an understandable and largely existential necessity for people to conform, so who can really be blamed that ‘when a large number of individuals, slaves to the system, end up being penetrated by this discourse [of techno-progressivism], they form a terrorising majority’ (Brune, 2003, p. 103; ‘Lorsqu’un grand nombre d’individus, esclaves du système, finissent par être pénétrés de ce discours, ils forment une majorité terrorisante’). The consensual normalisation of all things technological is partly generated by fear, given that most people, especially professionals, loathe being perceived as abnormal or irresponsible citizens. It pays dividends to internalise the ideological edifice of technological utopianism rather than to question its inconsistencies, which is why scholars are able to fruitfully relate Taylor’s ideals for scientific workplace management to Foucault’s notion of disciplinary power (Hoffman, 2014). In fact, the Foucauldian premise of the workplace as a network of productive yet subjugated bodies can easily be related to the language and translation industries, whose workers and labourers are hustling along in docile serenity under the surveillance of their tech- and data-owning overlords. Which closes the loop again and leads us straight back to George Orwell since, at least from our point of view, the freedoms, liberties and riches continuously promised by the champions of techno-capitalist industrialisation remain beyond reach for the great majority of people.

5. The quantity of crowdsourcing translation

Walter Benjamin’s eponymous essay ‘The Work of Art in the Age of Mechanical Reproduction’ (1935/1969) could be rephrased for the translation profession into ‘The Work of Translation in the Age of Digital Automation’. New collaborative technologies and logistical projects such as crowdsourcing are on the verge of transforming translation work into collective mass (re)production. The harnessing of collective intelligence, of ‘the wisdom of crowds’ (Surowiecki, 2005), the wisdom of the many rather than the few, is moving centre-stage in the age of digital reproduction and automation. Where once existed an army of passive online-consumers, like in the early stages of public television, the advent of platform capitalism has given rise to new forms of audio- and televisual interaction where people are now capable of simultaneously producing and consuming – hence prosuming – their very own (digital) products and consumer goods (Ritzer & Jurgenson, 2010, p. 14). The prosumer-phenomenon sparked the phenomenon of crowdsourcing and in recent years has gained significant traction with online translation activities. The advantages of crowdsourcing are manifold: speed through spontaneous and autonomous cooperation (cf. Section 2), efficiency through flexible and round-the-clock services (cf. Section 3), and quantity through dozens and more people working on any assignment of any size. Jiménez-Crespo (2017, pp. 17–23) differentiates between two types of translation crowdsourcing: (1) crowdsourcing translation: tasks organised by an institution, (2) online collaborative translation: tasks organised by interested parties such as volunteer groups. In crowdsourcing translation, the content to be translated tends to be broken down into very small units, often just chunks of sentences that are translated by many translators. In online collaborative translation, content is translated by a potentially large cohort of people with a shared and vested interest, such as cultural or political activists, volunteers or fans. When viewed in connection with relations of hierarchy and decision-making, crowdsourcing translation may be understood as a ‘top-down effort’, since it bears less democratic potential than the ‘horizontal bottom-up efforts’ (ibid., pp. 18–19) pertaining to online collaborative translation. Within and across both types of translation, however, working conditions may differ considerably, with some translators working on a voluntary basis, others as professional freelancers and again others being paid on a per-task basis.

Crowdsourcing tends to enjoy a rather positive reputation in the public imagination, not least because it transmits an idea of collaboration and sharing where people from all over the world work together towards a common goal. Websites that offer crowdsourcing activities tend to show pictures of smiling people from different continents and countries. There is, of course, ideology at work here, where an interplay of images and words combines for an elusive sense of togetherness. Frequently used concepts such as co-worker, for instance, evoke a false sense of community where people share the same values and thus happily collaborate (Persain, 2020). A concept such as employee, however, far from communicating a spirit of progress and community, would sound rather old-fashioned and strict. It would, above all, evoke in the reader a vertical chain of hierarchy which co-worker rigidly strives to negate. The denial of hierarchical relationships in commercial organisations underpins those forms of managerial discourse that support the neoliberal and neo-Taylorist organisation of work (ibid., p. 8). Such ‘managerial newspeak’ (cf. Vandevelde-Rougale, 2017), through the use of accompanying ideological markers like partnership or cooperation not only habituates workers, and in turn budding translators, into the ideology of corporatism, but also sustains an illusion of professional autonomy, of being seated at the proverbial professional steering wheel.

It remains a truism that when a certain set of discourses, arguments and concepts is applied instead of others, there is an interest in concealing real and thus existing underlying power relations (Brune, 2003, p. 93). In his famous work Lingua Tertii Imperii, Victor Klemperer (1947/2020) demonstrated how the German language was hijacked by leading Nazi ideologues such as Alfred Rosenberg and Joseph Goebbels, both in order to habituate the Germans into a barbaric political regime and a universe of discourse that prepares them for war and suffering. By analogy, managerial newspeak borrows a similar mode of operation in that it rallies people behind its neoliberal and profit-oriented agenda. Whilst crowdsourcing translation may be an adequate solution for some humanitarian projects, crowdsourcing’s vertical and thus ‘top-down’ decision-making approach rather engenders precariousness and alienation (Ettlinger, 2019). This new form of deterritorialised labour makes it more than difficult – indeed precarious – for translators to negotiate prices for commissions, especially when they are employed by large translation platforms (Fırat, 2021, p. 62). Most translators nowadays work as freelancers and are facing global competition in the digitalised global translation industry. They operate within a commercial climate that constantly rejuvenates itself by marketing campaigns generated and maintained through the ideological discourse of neoliberalism. The non-existence of physical boundaries in this fragmented and simultaneously fluid virtual world (cf. Bauman, 2000) exacerbates a deep-seated ethical dilemma. The workforce in highly competitive markets, such as in the language and translation industries, remains under pressure to accept challenging and partly unethical working conditions just in order to remain in employment. In Bauman’s (2000) sense of a ‘liquid modernity’, precisely because there are no more physical borders in online digital labour, crowdsourcing translation represents a prime example of a deterritorialised and thus ‘liquid’ beehive-like activity. This deterritorialisation of professional activity promotes the reduction of labour cost and, like in a magnifying glass, encourages tense worldwide competition for the best-renumerated translation commissions. There also exists a political necessity – volonté politique (cf. Bourdieu, 1991) – for market actors, including political authorities, to strive for flexibility. Workers believe to be gaining from flexible working hours, but they are in fact losing out. While at least large LSPs benefit from the development of technology (NMT, translation platforms, etc.) and neoliberal globalisation, translators have little room for manoeuvre and are thus caught between technology and economic imperatives. So, is the crowd really ‘wise’ enough or, rather, do individuals follow their crowds too uncritically?

Crowdsourcing has been a contentious activity for quite some years now (cf. Klaus, 2014, p. 14). Already in 2009 there was a petition against the use of crowdsourcing translation on ProZ.com (2009) entitled ‘Professional translators against crowdsourcing and other unethical business practices’ (more recently, see Flanagan, 2016). The evolving professionalisation of crowdsourcing translation also implies a hidden, though never explicitly framed, performative agenda which orients its workforce towards specific ways of one-dimensional thought (Marcuse, 1964). This kind of deterritorialised labour serves as a mechanism and catalyst of habituation. Whatever the times or entrepreneurial agendas, ideological slogans and buzzwords such as freedom of choice or business stakeholder function to maintain and pacify an army of cheap labourers. The translation industry is looking for a multiplication of the workforce. The discourse (or managerial newspeak) that seeks to gain new translation workers, however, serves above all to hide the fact that digital forms of translation labour – such as translating with machine translation systems, postediting translation and crowdsourcing translation – generate a steady decline in wages (Sadin, 2018, p. 145). Bolstered through oft-repeated concepts such as productivity, efficiency, speed or progress, this discourse functions, following Althusser (1984), like a quasi-religious ‘interpellation’ of translators in the 2020s. And in the Foucauldian sense, this discourse may be understood as an epistemological dispositif that can be broken down into three basic components (Gordon, 1980, p. 202): if its ideological component conveys the values of our enigmatic financial overlords, its symbolic component conveys the contemporary reality of hierarchical relations of power, and its pragmatic component, perhaps most vividly, conveys the master’s power – in the truest Hegelian sense – to influence the slave’s behaviour (cf. Vandevelde-Rougale, 2017). It appears neither contentious nor plainly ridiculous to claim that the master-slave dialectics remains fresh and alive in contemporary world politics, entrepreneurship and commerce.

6. Conclusion

The development and application of AI-based technologies moves the reality of the ‘digital neo-Taylorist workplace’ centre-stage in international labour markets. In the language and translation industries, the growing predominance of AI-based machine expertise engenders a gradual devaluation of human experience and historical knowledge. It remains, however, imperative that translators, not machines, remain located at the heart of transcultural communication. Technology is synonymous with speed, efficiency and quantity, though it appears to be ‘Le mythe du progrès’ (Brune, 2003, p. 12) that is blindfolding us to imagine alternative scenarios for the workplaces of the present and future. We mentioned above that translators in the 2020s can be seen both as victims and perpetrators. Victims, as they occupy a precarious position in the world economy. And perpetrators, as due to the necessity of earning a living they reproduce the dominant ideology of technological progressivism. But hang on. Is it not a bit silly to point the finger at individuals and collectives when musing on any apparent cause for our predicament? And even if the culprit was greed or any unknown evil, is perhaps our own ignorance the biggest cause of the malaise? We will probably never know … More importantly, a global mobilisation of authorities, non-governmental institutions and other lobby groups might help preventing a quasi-universal drive towards the enforcement of neo-Taylorist principles in shopfloors around the world. The European Commission (2023) certainly is trying to improve the working conditions for digital platform workers by implementing guidelines to improve the negotiation rights of self-employed people. Whilst such initiatives remain difficult to implement on a global level, where each country has different regulations, the European Commission itself reduced its translation staff by 17% in 10 years, with 450 translators having lost their jobs between 2013 and 2023 due to automation (Sorgi & Di Sario, 2023). Promoting ethically sustainable work environments, where humans – in our case transcultural communicators – are respected for their human value, not their labour value, might belong to the terrain of utopian dreaming. It is, however, our duty as academics to keep pointing the finger at questionable labour-driven and societal developments. Félix do Carmo (2020, p. 36) says that ‘[m]odern professional lives are guided by regular and strict forms of evaluation of performance, embedded in a global pressure for increased productivity and efficiency’. In modern translation practice, every time translators are entering a source text into a neural translation machine, they are habituated into providing free data streams to companies and they are devaluing the labour of human translation. And this also applies to translators working for the European Union! Digital neo-Taylorism in translation is inextricably intertwined with the stellar ascendancy of AI technologies. The primary principle of Taylorism is to collect and classify the established knowledge of the workforce, just as the translation machine collects and indeed confiscates human expert and general knowledge. In the Taylorist office, ‘extraordinary human ingenuity has been used to eliminate the need for human ingenuity’, whilst the neo-Fordist assembly line that ‘lives in’ the ‘mean’ translation machine serves ‘to transfer knowledge, skill, and decision making from employee to employer’ (Barbara Garson in Crawford, 2009, p. 38). Margaret Thatcher’s ominous neoliberal motto TINA – There is no Alternative – bears witness to the ignorance of those who shy away from thinking outside the confines of the Taylorist box. So, once again, criminal malevolence might not be at the root of all evil in the wake of ‘the great data swindle’, it might just be our all-too-human ignorance (cf. Zuboff, 2019 and the 1980 Sex Pistols mockumentary The Great Rock ‘n’ Roll Swindle).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Stefan Baumgarten

Stefan Baumgarten is currently head of the Department of Translation Studies at the University of Graz. He is also heading the research area ‘Translation, Society and Digital Transformation’. His research centres on (critical) translation theories, the role of translation as an ideological practice and the social impact of translation technologies. He is co-editor (with J. Cornellà-Detrell) of the special journal issue ‘Translation in Times of Technocapitalism’ (Target, 2017) and of Translation and Global Spaces of Power (Multilingual Matters, 2018).

Carole Bourgadel

Carole Bourgadel works as a predoctoral research assistant at the research area ‘Translation, Society and Digital Transformation’ at the Department of Translation Studies. In her dissertation project, she investigates the application of AI technologies in translator education against the backdrop of neoliberal policies and techno-capitalist ideology. Her main research areas are the relations across translation, machine translation and artificial intelligence; translation training, historically embedded market mechanisms and monopolies; and translation services in the neoliberal market economy. She also has experience in developing translation apps.