Towards an ethical framework for evaluating paid translation crowdsourcing and its consequences

ABSTRACT

The emergence of new technological tools for translating online has potential wide-reaching impacts for translators and the industry. However, very little research has so far systematically analysed the ethical dimension of such practices. To this end, this paper proposes an ethical framework to evaluate the consequences of translation platforms that use paid crowdsourcing on translators and clients. Based on the values that inform a good life, the paper proposes the use of multiple ethical principles: beneficence and non-maleficence, autonomy, and justice, to consider the different outcomes of the use of crowdsourcing platforms. Applying this framework to paid crowdsourcing platforms, the findings suggest that the translators’ well-being and respect for their autonomy are undermined by platforms in several ways, while clients stand to gain from their use.

KEYWORDS:

• Crowdsourcing
• ethics
• consequentialism
• translation platforms
• online collaborative translation

1. Introduction

Translation platforms have existed since around 2008 (Garcia 2015). With the integration of artificial intelligence (AI) into project and data management workflows, they are one of the most prevalent examples of ‘disruptive’ translation technologies (Sakamoto 2018). Ethical challenges faced by the industry at large are heightened through platforms using these digital technologies. Many studies have already shown low job satisfaction (see, e.g. Moorkens 2020), industry sources also report grievances concerning work–life balance and unstable workflows, with 40% of freelancers reportedly not earning enough from translation/interpreting and needing an additional source of income (ELIA 2020). This situation is unlikely to improve with a global competition on prices and platforms that establish themselves as mediators. Fırat (2021) finds that translators are exposed to similar highly alarming issues on translation platforms, such as price dumping, loss of translator autonomy and bargaining power, poor work–life balance and questionable use of their data. Already, several of the language service providers (LSPs) on Nimdzi’s (n.d.) list of the largest 100 LSPs operate translation platforms: Lionbridge (nr. 6) bought Gengo, the Acolad group (nr. 10) acquired Textmaster, CSOFT (nr. 57) owns Stepes, and OneHourTranslation rebranded as Blend (nr. 82). So, it can be assumed that platforms are here to stay and will only become more prominent.

Even though scholarly interest in paid crowdsourcing platforms and the ensuing ethical issues is increasing, no ethical framework has been proposed so far to systematically analyse their possible consequences. They have thus far been explored in terms of their impact on project management (Sakamoto 2018), on translation workflows (Jiménez-Crespo 2021), on translators’ status (Jiménez-Crespo in this volume) and on translators’ working conditions (Fırat 2021). While none of these studies focussed explicitly on their ethical impact, other online collaborative translation practices have been studied in terms of ethics, such as unpaid crowdsourcing on Facebook (Zwischenberger 2022), community translation as a form of ‘playbour’ (Rogl 2016), and Wikipedia translation regarding the notion of linguistic justice (McDonough Dolmaya 2017). This paper aims to fill this lacuna by employing principles of consequentialist ethics to formulate a framework (for future research). It will be applied to paid translation crowdsourcing platforms as the prototypical form of ‘disruptive’ technologies in the translation industry to explore some of the consequences that their business practices have on translators and clients.

2. Paid crowdsourcing on translation platforms

Translation platforms have recently garnered interest from Translation Studies scholars (Fırat 2021; Garcia 2015, 2017; Jiménez-Crespo 2021; Sakamoto 2018). However, there is neither a consensus on what to call these platforms, although several qualifiers such as ‘paid crowdsourcing platforms’ (Jiménez-Crespo 2021), ‘digital labour platforms’ (Fırat 2021) and ‘translator platforms’ (Heinisch and Iacono 2019) are used, nor how to differentiate between different types and what even constitutes a platform. Fırat’s (2021) study encompasses bidding marketplaces such as Fiverr and ProZ as well as platforms such as Stepes that act as mediators by setting the rates and distributing tasks on a first-come-first-served basis. Gough et al. (2023) also include translation environment tools that allow translators to work concurrently, such as SDL Trados Groupshare and Wordbee. Jiménez-Crespo (2021) differentiates between platforms that use paid crowdsourcing and those that use professional paid crowdsourcing, i.e. they offer (near-)market rates to their translators. Garcia (2015, 27), on the other hand, places platforms and conventional language service providers on a continuum rather than in clearly delineated categories. This fragmentation points to the fluidity that translation platforms exhibit, both in the types of translators they employ, i.e. certified/experienced translators vs. novice/bilingual translators, as well as in the services and prices clients can choose from.

The shared characteristic of paid crowdsourcing platforms is the outsourcing of translation jobs to a large pool of novice or ‘hobby’ translators and/or translators with experience/a university degree. Microtask crowdsourcing involves the segmentation of a text into ‘strings’ which are then distributed among multiple translators. This type of translation crowdsourcing on platforms has been researched by, e.g. Jiménez-Crespo (2021 and in this volume). This paper refers to them as ‘on-demand platforms’ as they are characterised by their particularly fast turnaround times. These include platforms such as Gengo, Blend, MarsTranslation, ProTranslate, Stepes, and Motaword. Macrotasks are considered more complex tasks that are not necessarily split into smaller parts:

Macrotask crowdsourcing refers to crowdsourcing that is designed to handle complex work of different degrees of structure and decomposability, assumes varying levels of (expert) knowledge over one or more domains, requires a range of 21st century skills, benefits from worker communication, collaboration, and training, and incorporates flexible work management processes that potentially involve the workers. (Lykourentzou et al. 2019, 10)

The transition to bigger crowdsourcing tasks – even whole texts – as the next step of translation crowdsourcing on platforms allows them to keep experienced translators engaged and expand their services to meet high-risk translation demands. Viewing macrotasks as part of the crowdsourcing spectrum highlights the expertise that is required in professional translation crowdsourcing. These ‘agency-like platforms’ follow the traditional TEP-process (translate-edit-proofread) and employ technological tools to streamline the ordering process as well as project management to nurture a large pool of translators with more experience and/or a university degree as their curated crowd. Examples of this type are Translated, Tomedes, Tolingo and Textmaster. Collaboration on these types of translation crowdsourcing initiatives takes place, for example, through the sharing and co-construction of resources, such as translation memories and glossaries. The community may also be fostered by additionally providing forums or chat functions. Contrary to conventional LSPs that outsource translation jobs, crowdsourcing platforms automate a considerable part of their project management, oftentimes with the help of AI. Tasks can be allocated either on a first-come-first-served-basis or through an automated assignment where an algorithm takes into account aspects such as language pair, previous translation performance, availability and rates. It should be emphasised, however, that on-demand and agency-like platforms represent overlapping rather than clearly delineated categories.

The ever-evolving nature of these internet companies necessitates a flexible and adaptable ethical framework to evaluate possible consequences for translators and other industry stakeholders.

3. The need for a consequentialist ethics

As Zwischenberger (2016) has argued, ethics in Translation Studies is commonly discussed from a deontological point of view, i.e. in terms of ethical codes of practice that deem certain actions as inherently morally correct and others as morally wrong. Consequentialism, on the other hand, evaluates actions based on their outcomes. To an extent, the deontological view is spread by professional associations that strive for greater recognition of translation as a profession, but it is also embraced by academia in hopes of finding some guidelines for their students. An example of this is Chesterman’s (2001) suggestion of a Hieronymic Oath, based on the idea of the Hippocratic Oath, using translator virtues from professional codes of ethics to formulate a deontic oath for the translation profession. The need to move away from such deontological codes has already been demanded by e.g. Inghilleri (2012) and Lambert (2023). Disillusioned by the frequently used deontic dichotomies of ‘foreignising’ vs. ‘domesticating’ translation, Pym (2012) suggests a consequentialist framework based on cooperation as the ultimate goal. His ethics is inspired by economic theory and therefore employs variables such as risk and transaction cost that need to be calculated in order to determine if the translation effort is worthwhile.

Pym’s (2012) theory opens pathways for new ethical reasoning. Moving away from deontological ethics and towards a more consequentialist ethics enables a crucial shift in perspective: it allows us to go beyond the micro-ethical decisions of individual translators as outlined in codes of practice and view the bigger, societal picture to determine the impact of certain practices and decisions. By its nature, consequentialism is also more adaptable to changing societal settings (Birnbacher 2013). The research area of online collaborative translation would especially profit from a more consequentialist view; not only is its impact on translators, other industry stakeholders and the translation process as a whole rather underexplored, it is also a very dynamic field due to innovations introduced by the internet and masses of volunteer translators. The fact that the line between novice, unpaid collaborative practices and the paid industry is blurred, e.g. since crowdsourcing has entered the paid industry, leads to a myriad of consequences to take into account.

To judge certain online collaborative practices and their consequences, the concept of exploitation has recently been applied. Rogl (2016) discusses the concept of ‘playbour’: a combination of labour and play, in an online translation community and relates it to possible exploitative practices. Zwischenberger (2022) similarly argues that the impact of unpaid crowdsourcing for profit-oriented companies constitutes a form of exploitation, not merely because of the lack of remuneration, but also because of the consequences it has for third parties and the translation profession as a whole. This will be discussed further in relation to the principle of justice.

Yet, where deontological ethics offer a clear set of rules to follow, consequentialist ethics might appear like a jungle to practitioners due to the multitude of consequences that need to be considered. Such a framework, especially a pluralistic one as suggested here, is therefore not very suitable to guide day-to-day translator decisions, but it can be applied to evaluate current acts based on their ongoing consequences to suggest possible ways to move forward. A consequentialist approach to translation ethics takes into account a broader societal view of ethics. The following section will sketch some of the underlying values of being morally good that form the basis of the framework.

4. Principles of a good life

A theory of ethics aims to answer meta-philosophical questions about morals. Thus, the object of an ethical investigation are the standards, rules and norms that make up morals. An ethical framework can therefore give us the tools to judge decisions that have been made in the past and to guide future moral choices. However, in order to evaluate possible actions and choose the best alternative based on their consequences, we need to look at which consequences are desirable in the first place (Birnbacher 2013).

Consequentialism oftentimes propagates a maximising approach to decide which action to take. This means that ‘[a]n action ought to be performed (is right) iff¹ its outcome is intrinsically better than (at least as good as) that of every alternative, and an action is wrong iff it is not right’ (Carlson 1995, 13). Accordingly, a person is obligated to choose the action with the best possible consequences. Another option is a satisficing approach that postulates that an action is right if the outcome is deemed as ‘good enough’, e.g. it crosses a certain threshold or is better than a certain percentage of its best alternative outcomes (Hurka 1990). Actions that go beyond are then considered supererogatory, i.e. exceeding the expectation. Evaluating outcomes in a way that they can also guide future actions or bring about change hence requires a dialogue among industry stakeholders to establish a common baseline for obligatory and supererogatory actions. There are, however, some values that are considered desirable for a good life in general.

Depending on the interpretation of consequentialism, there are different values associated with a good life, and thus, different actions may be regarded as morally right or obligatory. Utilitarianism, for example, presumes that the action that produces the highest amount of pleasure and the least amount of pain is the right action to perform (e.g. Mill 1864/2014). Assuming that pleasure is not the only value people strive for, it is necessary to consider what a good life entails. Frankena (1994, 107) lists the following values, among others: happiness and pleasures, truth and knowledge, beauty/beatitude, love and cooperation, freedom, peace and security, and a fair distribution of these values. These will form the axiology for the framework proposed in this paper and thus inform the desirable outcomes.

In its pursuit of the single value ‘pleasure’, Utilitarianism needs to formulate only one principle: the morally right action is the one that brings the greatest happiness to the greatest amount of people (Mill 1864/2014). However, this postulation already results in two moral demands: (1) the consequences of an action should be a surplus of desirable outcomes vis-à-vis undesirable ones, and (2) this surplus should be equally distributed among as many people as possible (Frankena 1994, 60). Hence, Frankena suggests splitting the greatest-happiness-principle into two and using the principle of justice alongside the principle of beneficence to account for the double moral obligations. Ranking different principles then also allows the resulting norms to be diversified. If the principle of beneficence takes precedence over the principle of justice, increasing the amount of happiness would be the priority, followed by efforts to spread the happiness to the highest amount of people.

Still, two principles cannot account for the array of possible desirable outcomes, as outlined above. This paper therefore draws on Beauchamp and Childress (2012) who, in formulating their ethics of biomedical sciences, propose four principles: the principles of beneficence, non-maleficence, autonomy and justice. The first two follow similar assumptions as the ‘greatest-happiness-principle’. The goal is to avoid causing harm and increase well-being. Since these two principles relate to each other, the proposed framework will consider them as one principle. Table 1 below gives an overview of how the values taken from Frankena (1994) relate to each of the principles.

Table 1. Ethical principles and their corresponding values.

Principle of beneficence and non-maleficence	Principle of autonomy	Principle of justice
Happiness and pleasures	Truth and knowledge	Cooperation
Beauty/beatitude	Freedom	Peace and security
Love		Fair distribution

The principle of autonomy concerns the right to self-determination and freedom. Autonomy also means that decisions that are made this way are based on sufficient understanding of the subject (Beauchamp and Childress 2012). As McDonough Dolmaya (2011, 102) postulates, translators participating in the Facebook or Twitter crowdsourcing initiatives engage in the crowdsourcing effort under a false altruistic motive as they are not aware of the huge profits those companies make. The same applies to translation platforms. Translators might not be aware how their contributions, for example their data and the shared translation memories, are used, and how this may result in additional profits for the companies who sell said data or use it to train machine translation engines. In those instances, the respect for autonomy is violated.

The third principle is the principle of justice, which determines how the desirable outcomes of an action should be distributed. Frankena (1994) calls for a fair distribution of the good, which creates the need to characterise ‘fairness’. Pym (2021, 10) attempts to address this in his principle of cooperation: ‘all parties act in their own interests but do so in a way that they all acquire more value than what they started with’. So, as long as each party gains something out of the transaction, regardless of whether those gains are equal, it can be considered a morally right action. However, applying Wertheimer’s (1996) concept of exploitation, Zwischenberger (2022) argues that an exchange can be deemed unfair or exploitative if, for example, the profits are distributed in such a way that one side makes much more profit than the other. This profit does not necessarily need to be monetary, for example, as I will discuss later, translators on platforms generate a huge amount of data that can be used by these platforms for their purposes. In light of this, Pym’s (2021) approach seems shortsighted as it could serve as an excuse for unreasonable profits made off the back of translators. Rawls’ (1971, 60) second principle of justice thus becomes more useful. He outlines some circumstances under which an unequal distribution may be tolerated: ‘social and economic inequalities are to be arranged so that they are both (a) reasonably expected to be to everyone’s advantage, and (b) attached to positions and offices open to all’. Rawls (1971) allows for some inequalities in the distribution of wealth and status because he assumes that using a higher reward as an incentive for certain groups can lead to larger societal advancements that are beneficial for everyone. However – and this is the main difference to Pym – the benefits for society must be especially in favour of the least advantaged people. In the context of translation platforms, this means that any technological innovations and generated surplus value need to lead to reasonable benefits for all users, especially translators and clients.

5. A consequentialist framework to study paid translation crowdsourcing

The framework I am proposing here is similar to the ethical matrix brought forward by Mepham (1996) to evaluate food biotechnologies, based on the work by Beauchamp and Childress (2012). While the academic disciplines are different, there is a shared context: new technological innovations are deemed as potentially disruptive and ethically ambiguous while the tools to adequately evaluate their possible consequence on the affected parties are missing. Because of this, the framework also aims to be adaptable to different settings in Translation Studies, especially within online collaborative translation where the introduction of new technologies enables new forms of translation that could potentially have wide-ranging impacts on translators, clients, users and the status of the profession. Consequently, the principles need to be interpreted in terms that are relevant to the affected parties before they can be applied to translation platforms. This is shown in Table 2.

Table 2. Consequentialist framework showing the interpretation of respect for the three principles considering the interests of translators and clients (for more detail see sections 5.1 and 5.2.).

Affected parties	Beneficence and non-maleficence	Autonomy	Justice
Translators	Adequate reward & good working conditions	freedom to make translation-related decisions	Equitable and fair treatment
Clients	Satisfaction of translation demand & adequate translation quality	Respect for choice (transparency)	Access to translations & universal affordability

For the purpose of this article, the column for affected parties only includes translators and clients, the two most important sides on a translation platform. Platform owners would be a very salient group to look at as well, but that goes beyond the scope of this article. The column for affected parties could also be further expanded to include other industry stakeholders, such as traditional translation service providers, users of the translation, or technology providers. Moreover, as Mepham (1996) has observed for his matrix, the categories could also be subdivided. In my framework, for example, translators could be differentiated into those who work or have worked for the platforms, and those who do not participate in crowdsourcing but whose livelihood may be affected as a result of the new technologies introduced by platforms. Alternatively, novice translators and expert translators could form different sub-categories. However, the differentiation between a novice and an expert translator on a platform cannot be clearly drawn at this stage. For the same reason, the term ‘adequate reward’ instead of ‘adequate payment’ was chosen to also account for non-monetary forms of reward. For instance, translators might feel the recognition or experience they receive is an additional reward for their work.

It should be noted that the categories of the framework might also result in contradictory moral obligations that cannot be fulfilled at the same time, especially concerning the two opposing sides of translators and clients, which constitute differently affected parties with different needs. The weighing and ranking of the results is therefore a necessary step if the consequences are meant to be evaluated, rather than simply described. Weighing and ranking alternative courses of action is an integral part of consequentialist ethics and involves deciding which principle takes priority over the obligation that results from a different principle (Frankena 1994). The difficulty in deciding which consequences outweigh other consequences is one of the critiques of consequentialist ethics. This paper can only provide tentative directions for this undertaking. The main aim is to outline possible ethical consequences that may serve translation associations, educational facilities and other industry stakeholders as a basis for a discussion.

In the following sections, I will apply the framework to paid translation crowdsourcing on translation platforms, thereby testing its usefulness. As this paper is not a longitudinal study, its focus will be on current outcomes resulting from the use of translation platforms. Possible wide-reaching long-term consequences can only be addressed tentatively.

5.1. Translators

5.1.1. Beneficence and non-maleficence

This part will mainly address possible harmful consequences for translators working on platforms, since the key objective of the principle of beneficence and non-maleficence is to prevent harm.

This group consists of translators that only work part-time as translators and only use translation to supplement their income, as well as those who make their living from translation. Furthermore, it can be assumed that a cline of novice to experienced translators are present on translation platforms (Jiménez-Crespo 2021). Because I am analysing the paid form of translation crowdsourcing, i.e. all translations are remunerated by the platforms, it can be concluded that income is directly related to this group’s livelihoods, which is contrary to other forms of online collaborative translation, such as fan translations. Hence, low income is likely to have a potential negative impact on translators’ well-being and may lead to longer working hours, which would also negatively affect their work–life balance.

Due to the global nature of the translation profession and ensuing different costs of living, it is difficult to determine which translation rate is adequate. Garcia (2015) compared rates on translation platforms with the average rate that translators set themselves on the marketplace ProZ. It was found that while some translation platforms seem to be charging near market level rates, others fall well below that. Gengo’s (n.d.-d) standard rate for translation services is $0.08/word for an English to German translation, with translators presumably getting only a fraction of that, while ProZ’s (n.d.) poll among their freelancers reports average translation rates from $0.08/word to $0.11/word. Gengo’s (n.d.-d) advanced services come closer with $0.14/word charged to clients. A poll of the Austrian professional translators association Universitas revealed that on average, members charge from €1.90 upwards per standard line (Universitas 2022). This roughly comes to $0.23/word, which is a significant difference to Gengo (n.d.-d). Since there is no way to compare and determine fair prices because translators work from all over the world, translation platforms setting a fixed rate does not seem ethically justified. To ensure that translators receive adequate remuneration for their corresponding cost of living would mean letting them choose their own rate. The platform Translated, for example, gives translators this possibility. However, the translation rate is also one of the factors in choosing which translator will ultimately receive a job. This competition might also incentivise price dumping beyond translation platforms, as translators may have to offer lower rates in order to compete with the prices on platforms.

In terms of working conditions, platforms typically emphasise the benefits, such as flexible working hours and working from anywhere in the world, which could have a positive impact on translators’ well-being. The platform Translated uses a tool for translators to track availability. On Gengo, translators can login and search the job board. On Stepes, translators can even work from their phone. However, this does not guarantee that any translation jobs will be available when the translator has time. On several platforms, translators will be notified via email if a job corresponding to their skills is available. Jobs are oftentimes allocated on a first-come-first-served basis, meaning that translators have a limited amount of time to decide if they want to take them or not. Gengo even recommends the use of an RSS feed reader so translators can be notified as soon as possible if something in their language pair becomes available (Gengo n.d.-c). Not only does this create a sense of competition among the translator community but it is also a potential source of stress. Concurrent translation, such as on Motaword, also leads to increased pressure through competition (Gough et al. 2023, 19–20). In Gengo’s general public forum, 5 out of the 42 user posts in 2022 and January 2023 were related to complaints about low workload (Gengo n.d.-a, n.d.-b). While this is not an exhaustive quantitative analysis and cannot be generalised by any means, it does point to a recurring concern voiced by translators on Gengo as a result of the race to accept first-come-first-served jobs.

If platformisation continues to rise in the translation industry, ‘conventional’ LSPs might be pressured to implement similar practices, which could lead to an emphasis on speed and availability instead of experience and quality in the industry. Even on agency-like platforms, translators risk being devalued in the crowd. Consequently, if investing in a formal education may seem less beneficial for translators, there will be ramifications on funding at universities and subsequently on Translation Studies as a discipline.

5.1.2. Respect for autonomy

Translators’ autonomy is interpreted as the freedom to make translation-related decisions. Translation platforms should act in such a way as to infringe as little as possible upon this freedom. The decisions refer to the skillset that is expected of translators as experts and are in general considered to go beyond mere translation, such as technological skills, e.g. evaluation and choice of tools, as well as management skills, e.g. time management, setting of rates (Krajcso 2018). Thus, translation platforms that use AI to calculate and set the delivery time as well as the rates and demand translators to adopt their proprietary CAT-tool, infringe upon the translators’ autonomy. As Olohan (2017, 277) has pointed out, such cloud-based translation technologies ‘produce a misleading impression of autonomy by “allowing” translators the “freedom” to complete their work anytime, anywhere, while their lived experience may be that of a translator on call’. The feeling of loss of autonomy is aptly summarised by one of the translators in Moorkens’ (2020, 23) study on job satisfaction: ‘if working for large agencies, we are now just a tiny cog in a large machine’. A survey among translators who work for translation platforms also reported a ‘lack of control over the workflow and on the final quality’ (Gough et al. 2023, 19).

Platform owners gain certain benefits from employing a proprietary CAT-tool as it allows them to retain control over the generated data, such as glossary terms and translation memories, but also user-specific data, such as translation speed. A platform can gather huge amounts of data on their users which they might use, for example, to improve their machine translation or other parts of their translation service. Translators oftentimes have no say on how their data is collected and used, which results in a loss of autonomy. Additionally, translators on platforms have no way to reuse their contributions on projects outside of said platform. This data collection also potentially undermines the principle of justice, as will be outlined in the next part.

Another potential issue for the translators’ autonomy are the ratings on translation platforms, as ratings that are received on one platform cannot be transferred to another. So, if translators have been performing well and would like to switch, they have to start from the beginning. The consequence is that translators’ freedom to choose where they want to work while keeping their performance score is reduced. The main purpose of the reputation seems to be to signal trustworthiness of the platform to potential clients, not improve the trustworthiness of specific translators. On top of that, another common concern in the general forum on Gengo, with 8 out of 42 posts, relates to unclear evaluations in reviews and re-reviews (Gengo n.d.-a, n.d.-b) as the language specialist who did the first review also performs the re-review (Gengo n.d.-e). Furthermore, some platforms, such as Blend and Stepes, will let customers rate the translators, which makes it an even less adequate evaluation tool. Thus, translators stand to gain very little from ratings but risk to lose a great deal if they become a source of stress.

The status of the whole profession might be negatively impacted if platforms disregard translators’ special knowledge, e.g. if translators are seen as replaceable crowd workers and are employed merely as a resource to generate translation data (e.g. for machine translation). This casts a view of translation as a simple mechanical task with experienced translators acting only as revisers and quality controllers. On top of that, the loss of autonomy as outlined in this section risks leading to a complete loss of their influence on the development of translation technology and translators in general losing their bargaining power.

5.1.3. Justice

In principle, the translators, the platforms and the clients are in a mutually beneficial relationship: In this interaction, the platforms gain money from the client, the clients receive the translated text, and translators are paid by the platform. This does not mean, however, that the profits are distributed fairly and that no exploitation takes place. Rather, this principle can be examined under two lenses: (1) exploitative behaviour as defined by Wertheimer (1996) whereby an unequal distribution of profits is viewed as morally wrong and (2) Rawls’ (1971) principle of justice that allows for an unequal distribution under certain circumstances. Looking closer at the issue of generating and utilising data, the fine line between these two points can be elucidated. With platforms gaining valuable data from their translators as outlined above on top of monetary payments from their clients, this exchange can be viewed as highly unequal and thereby exploitative according to Wertheimer (1996). Explicit consent to such transactions does not redeem their exploitative context (Wertheimer 1996). As Zwischenberger (2022) has argued, individual participation might negatively affect third parties such as the translation community as a whole.

Applying Rawls’ (1971) principle of justice to the interaction raises the question of whether unequal distribution is justified by benefits that are mostly for the under-privileged and least-advantaged. To give an example, in the terms and conditions, the platform Translated informs their users that the data generated by the usage of their free online CAT-tool MateCat may be processed internally to improve their services even if the privacy setting is chosen (MyMemory n.d.). Under the principle of justice according to Rawls (1971), the use of data generated for free but given back to the community (under the public setting), is ethically permissible. However, the sole internal use of the data would undermine this principle. The adherence to Rawls’ principle of justice needs to be evaluated on a case-by-case basis. Otherwise, it risks devolving into an arbitrary interpretation of fairness. Nonetheless, control of huge amounts of data in the hands of a few companies could result in monopolisation, as pointed out by Fırat (2021), thus further skewing the power dynamics in the translation industry.

5.2. Clients

The clients’ perspective has so far seen very little consideration in online collaborative translation research, especially in the context of platforms.

5.2.1. Beneficence and non-maleficence

The principle of beneficence and non-maleficence regarding clients mainly concerns whether their translation needs are adequately met. The dynamic or fit-for-purpose quality that is prevalent on translation platforms (Jiménez-Crespo 2017) means that, in theory, clients receive their desired quality, be it a social media post or a contract, and pay depending on the difficulty of the source text. Translators with a matching expertise are then allocated to the respective translation jobs. From the clients’ perspective, translation platforms also fulfil the requirements of our fast-lived world by offering speedy delivery times. Thus, the respect for their well-being is met if platforms ensure that sufficiently qualified translators work on the translation jobs. This can be done through extensively screening the translators or building some other form of trusting relationship with them.

However, some on-demand platforms such as Gengo, Blend and MarsTranslation offer proofreading only as an optional service. This has potential consequences on the translation quality clients receive. Not only do low-quality translations potentially lead to reputational and legal ramifications for the clients, published mistranslations could also harm the users. Societal risks might follow, such as a lack of translator accountability and trust in translators, if jobs are done by a faceless crowd rather than a qualified translator.

5.2.2. Respect for autonomy

Respect for choice, as the second principle, depends on sufficient available information. Platforms aim to establish themselves as trustworthy mediators thereby potentially taking away from clients’ autonomy. Some translation platforms, such as Gengo, Translated and Textmaster, let clients choose their required translation quality and charge them accordingly. There is no need to connect with a project manager and wait for a quote. The easy ordering process can be viewed as a benefit for clients, but it might in fact end up leaving them overwhelmed by the number of choices. If clients do not know their specific requirements and are not properly advised, they cannot reasonably be expected to select the required level of quality. Similarly, the ratings of translators are not clarified on the websites of platforms and are thus not transparent for clients. On the other hand, letting clients browse translators’ public profiles and ratings, as on Stepes, Gengo, and Mars Translation for example, increases autonomy.

If ‘conventional’ signals of trustworthiness, such as translator certificates (Pym, Orrego-Carmona, and Torres-Simon 2016), are absent, clients do not place their trust in translators. Rather, they trust in the technological infrastructure that is the platform. This potentially takes away power from the translators and clients and shifts it to the platforms that set the quality standards by controlling translator ratings.

5.2.3. Justice

With respect to the principle of justice, which is interpreted here as universal affordability and access to translations, it cannot be said that the platforms have a considerable impact as of yet. Translation platforms, however, might have the means to offer and deliver more languages than local translation agencies. The European Commission (2012, 36) noted this as a potential benefit of unpaid crowdsourcing. Nevertheless, the platforms’ decision to do so will depend on considerations regarding profitability. Prices then might still be out of the range of clients that rely on these translations. One could also argue that the responsibility of subsidising translations into minority languages ultimately lies with the governments.

5.3. Weighing the consequences

The last step of a consequentialist framework is to weigh the consequences, by, for example, ranking the principles, taking the context into account, finding the solution with the least amount of harm or converting everything into monetary terms and choosing the most profitable option. This is usually carried out before the action is taken, but since this evaluation is looking at unfolding consequences, this paper is dealing with current and past actions. On the one side, the respect for translators’ well-being, especially regarding good working conditions, is subverted by first-come-first-served task allocation, non-transparent evaluations and ratings, and possible low rates vis-à-vis the costs of living. On top of that, translators’ autonomy, in particular with respect to the choice of tool and rate, and the control over their data, is threatened. On the other side, the clients’ well-being, interpreted in the framework as meeting their translation demand and providing adequate translation quality, is potentially improved by platforms offering an easy ordering process and fast translations.

An overarching goal to guide the ranking might be a trustworthy industry that cooperates (Pym 2012, 2021), a technologically versed translation industry, universally affordable translations or a sustainable translation industry (a demand that has been made before (see, e.g. Moorkens 2020)). Depending on which of these perspectives is taken, different compromises will be made. Targeting a trustworthy industry, one might promote actions with the aim of good translation quality. Targeting a technologically versed industry, one might prefer actions that further technical developments to speed up translation workflows. These two might then condone practices such as rigorously testing and evaluating translators and strict monitoring of translators, which might diminish translators’ well-being. Similarly, universally affordable translations might negatively influence translators’ income.

In the end, however, one might argue that a sustainable industry is the one that is in the best interest of all stakeholders. While the translation industry is flourishing, with a steady growth in revenue (DePalma and Lommel 2022), the same cannot be said for translators. If good translators left the industry over poor working conditions, the industry could be faced with a downward spiral of decreasing quality and rates. As the risk of receiving an unsatisfactory translation increases, the willingness of clients to pay adequate rates decreases, which further deteriorates the translators’ working conditions, prompting more translators who know their value to leave the industry (see Pym, Orrego-Carmona, and Torres-Simon 2016 for a more detailed description of this type of market disorder). To prevent this talent crunch, ensuring good working conditions for translators, by following the principles outlined in this paper, should take precedence over universal affordability and convenience for the client.

6. Conclusion

The aim of this paper was to present an ethical framework for the study of ‘disruptive’ online collaborative translation phenomena, particularly paid translation crowdsourcing platforms. I have used the principles of beneficence, non-maleficence, autonomy and justice to take into account a range of possible consequences and their moral implications. The ethical analysis carried out based on the consequentialist framework revealed that, regarding translators, the principles are undermined in several ways by the use of translation platforms while clients stand to benefit. In the end, the consequences that were presented, such as price dumping, stress on translators due to constant ratings and competition over first-come-first-served jobs, and loss of control over their generated data, need to be weighed against consequences for other industry stakeholders. As a possible solution, it was argued that striving for a sustainable translation industry necessitates addressing translators’ needs first. Translation platforms are significant players in the industry and therefore bear a certain responsibility in shaping a sustainable future. On top of that, considering the principles of justice by Rawls (1971), it might be argued that platforms are obligated to contribute their surplus value to the common good, by, for instance, sharing their translation memory data or developing new technologies that benefit translators and clients. Another issue that should be taken into account which, however, this paper was not able to address, is the environmental impact, namely that developing and sustaining the technologies that run platforms consumes a tremendous amount of energy.

While this framework has certainly shown promises in the study of translation platforms, it is important to address its limitations: its usefulness depends on the available data to determine the impacts shown in Table 2. This paper has presented a theoretical framework that can guide studies that wish to analyse possible consequences. Interviews and questionnaires with translators that have worked on platforms should be conducted to obtain information about their experiences and opinions. To further expand on this research, I also recommend longitudinal studies to gain a deeper understanding of the resulting consequences. Lastly, actually making a change in the industry, based on this consequentialist framework, will very much depend on the effort of scholars, translator trainers and professional associations to spread awareness about the possible consequences of the use of translation platforms.

Disclosure statement

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

Additional information

Notes on contributors

Leandra Cukur

Leandra Cukur holds a BA in Transcultural Communication and an MA in Translation from the University of Vienna and is currently working as a research assistant as part of the research group Transcult.com at the Centre for Translation Studies in Vienna. Her research interests include the impact of online collaborative translation on the translation profession, especially the use of translation crowdsourcing in professional settings and its ethical implications. In her PhD thesis, she aims to shed light on the design and role of translation platforms and to explore their usefulness and drawbacks for translators.

Notes

1. if and only if.