May 27, 2016

Comments about FIT’s position statement on crowdsourcing (guest post)

A response by Attila Piróth to a recent statement published on the web site of the Fédération Internationale des Traducteurs [1].
Crowdsourcing is certainly a very effective term; calling some of the practices it enables “digitally distributed sweatshop labor” – for this seems like a much better description of what’s happening on crowdsource-for-money platforms like Amazon’s Mechanical Turk – wouldn’t accomplish half as much.
(Evgeny Morozov)[2]
Digitally connected mobs will perform more and more services in a collective volunteer basis, from medicine to solving crimes, until all jobs are done that way.
(Jaron Lanier)[3]

In the past few years, crowdsourced translation and machine translation have received a great deal of attention. Both are frequently called “disruptive technologies”, and are claimed to drive growth for businesses. Professional translators are often advised to get used to the idea that machine translation and crowdsourcing are “here to stay” and to adapt “to the changing landscape of the profession”. Machine translation post-editing is frequently cited as a new “niche” for translators.

The topic choice for the two FIT position statements thus reflects important and interesting realities. However, in its stated role as the “voice of translators worldwide”, FIT should not shy away from discussing some crucial issues that go beyond the simple technicalities presented in the paper. And if FIT is to reasonably call its paper a statement of position, it should dare to state one.

Finding a consensus on the more contradictory aspects will not be easy within FIT. The socio-economic issues that lie at the heart of the heated debates around crowdsourcing and machine translation boil down to the conflict between value creation by independent professionals and value extraction by those who own certain technologies (e.g., MT), linguistic resources (e.g., TMs) or platforms. Once again, we are faced with the labor versus capital debate – which is perhaps one reason why corporate interests like to use the term translation industry. Effectively, crowdsourcing and machine translation aim to ensure the necessary ingredients for the industrialization of an intellectual activity, and (by redefining expectations) to propose alternatives for the scarcity of the required competences. This is precisely why both trends have attracted major capital investments.
Example: Duolingo is a language-learning website that received 15 million dollars of capital funding at an early stage of its development. The core idea as represented to students was to teach languages through translation exercises. The more advanced the learner, the more difficult the sentences to translate. Peer-to-peer voting provides feedback on the participants’ performance. Courses are free, because the core idea as represented to financial backers is that the company generates its income by selling the translations produced by the crowd. The patchwork translations thus provided were meant to be sold to major content creation hubs – gawker, huffpost, etc. This “disruptive” model would thus enable the translation of a huge amount of text (for which “there would have been no traditional budget”). If one consults individual professionals such as language teachers and journalists, they will also add that this platform creates competition not just for translators but for them too – thereby disrupting several professions at once.
This model gives a clear translation-related example to the main thesis of Douglas Rushkoff’s new book, Throwing rocks at the Google bus: how growth became the enemy of prosperity.[4] Crowdsourcing does not enable a sustainable professional career for those who perform it: crowdsourcing is fundamentally a winner-takes-all scheme, in which the only real winner possible is the entity that owns or controls the platform. As the casino business knows, the house always wins.[5]

In the introductory quote, Evgeny Morozov calls crowdsourcing “digitally distributed sweatshop labor”. Given that recent reforms to the French labor law have lead to massive protests, this is also an opportune moment to assess the sort of legislative treatment this digitally distributed sweatshop labor receives.

The short answer is: it is entirely overlooked. Crowdsourcing’s diffusely distributed nature – it is literally everywhere and nowhere – seems to cast an impenetrable veil obscuring links to any physical jurisdiction.

Consider a brick-and-mortar bookstore, which, to increase its profit, invites volunteers to unload the delivery trucks, fill the shelves, clean the floor, etc. The volunteers bear their own costs and have no protection with regard to health, safety, work hours and insurance; they contribute because they identify in some way with the company and its products, and may hope to be offered some kind of paid work eventually.[6] In most countries, that has long been against the law: the company should hire the workforce, pay them at least the minimum wage, pay the various contributions/taxes after the employees, etc. When a company makes a profit, workers are paid, and the state also gets a share in the form of taxes and other contributions.

Over the past several years, many brick-and-mortar bookstores have been driven out of business by a virtual bookstore that has developed one of the most sophisticated platforms in the world: Amazon. As explained in Wikinomics by D. Tapscott and A.D. Williams,[7] hundreds of thousands of volunteer programmers participated in the “collaborative effort” to build the Amazon platform – which debuted as a bookstore, then added consumer electronics (bankrupting Circuit City and Best Buy), and only continues to grow and diversify.

Since the boom of the digital knowledge economy, numerous volunteer “community” projects have been launched under the banner of “harnessing the unused intellectual capacity of the community (the cognitive surplus[8]) for the benefit of all”. But who will extract that “cognitive surplus”? Will the resource extraction models developed in the 20th century for oil, gas, minerals etc. be followed – with notional “competitors” forming close alliances behind the scenes to control ownership of the resources? Cognitive surplus may be even more attractive to mine than physical resources because there is no sovereign owner and there are no cross-border issues requiring negotiations, contracts, royalties or trade agreements. But are nations really OK with having their workers deliver free, untaxable labor to, among others, private foreign interests?[9]

A typical example is when major IT companies can slash customer support costs because an enthusiastic user community is at their disposal to provide peer-to-peer help for free. IT giants like Google, Microsoft, Amazon, Symantec, etc. all benefit from such volunteer help. For these companies, the potential to use unpaid labor in handsomely paid (or even publicly subsidized) projects is not some kind of unexpected but fortuitous glitch: it is a system feature by design.

A perfect example along these lines is the ACCEPT project, in which crowdsourcing meets machine translation. Through this project, the EU generously offered a million-euro check to US digital media companies Symantec and Acrolinx and French translation company Lexcelera to cover some of their machine translation R&D costs. One of the promises these companies made was to scale up the volunteer operations of Translators without Borders (TwB), a nonprofit organization that they control,[10]  and whose actual work is completed by unpaid contributors sourced from all over the world. Thus, although the charitable efforts of the volunteers constitute the most publicly visible aspect of this apparatus, certain companies represented at the top of the hierarchy also benefit much less visibly by deriving privatized profit from free socialized labor.

In a remarkable article published over five years ago, the Northern California Translators Association (NCTA) unveiled the real character of crowdsourcing. That analysis – and hopefully the present one, too – shows that the translation profession is not isolated: it is as strongly affected by social (media) trends as any other profession where telework has become the norm. Legislation lags seriously behind technology, and representative bodies of freelancers must act to close that gap.

A “position statement” by an international federation of professional associations can be a good step in that direction – but as noted at the outset, such a paper will accomplish little if it fails to take a clear position.

Professional associations whose member base is comprised solely of individual professionals are in a much clearer situation than those associations in the FIT family that also admit corporate members. The former should accordingly step forward and raise the issues that are omitted from the FIT paper and negatively affect their members. Raising these critical questions may ultimately mean that no FIT-wide consensus can be reached about crowdsourcing (or machine translation). But that is a much healthier outcome than remaining a silent signatory to the current position statement – and hence tacitly agreeing that there is nothing to see here and we should all move along.

Attila Piróth

Acknowledgement: Some ideas presented above have emerged or crystallized in conversations with colleagues, in particular with Vivian J. Stevenson, who also read the manuscript.

[2] Evgeny Morozov, To save everything, click here. Penguin, 2013. ISBN: 978-0241957707.
[3] Jaron Lanier, You are not a gadget. Vintage, 2011. ISBN: 978-0307389978.
[4] Douglas Rushkoff, Throwing rocks at the Google bus: how growth became the enemy of prosperity. Portfolio, 2016. ISBN: 978-1617230172.
[5] “The bigger, centralized solutions offered by corporations with traditional, extractive, and monopolistic strategies are more attractive to investors, who are themselves betting on winner-takes-all outcomes.” D. Rushkoff, ibid.
[6] Interestingly, this kind of effort looks similar to sweat equity. According to Investopedia, “Sweat equity is contribution to a project or enterprise in the form of effort and toil. Sweat equity is the ownership interest, or increase in value, that is created as a direct result of hard work by the owner(s)…” The difference is that with unpaid crowdsourcing, the owners get the equity increase while the crowd contributes the sweat for free with no guaranteed return. Appearing on Stephen Colbert’s talk show in March 2014, Jaron Lanier gave a brief overview of his book, Who owns the future (Simon & Schuster, 2013, ISBN: 978-1451654967), and noted that “…we talked ourselves into this weird double economy, where if it’s about stuff, we believe in markets, if it’s about information, then we think it should be shared, it should be open...”. He also outlined a possibility of how those who contribute to the improvement of Google Translate could be rewarded through a micropayment system that logs the reuse of individual contributions.
[7] Don Tapscott, Anthony D. Williams, Wikinomics: How Mass Collaboration Changes Everything. Portfolio, 2006. ISBN: 978-1591841380.
[8] See for example Shirky, Clay, Cognitive Surplus. Penguin, 2010. ISBN: 978-1594202537.
[9] This is especially interesting in view of the various tax minimization strategies that have also proliferated with globalism. Many of the same corporations that stand to benefit from a given nation’s cognitive surplus can sell back into the same population while enjoying minimal exposure to the domestic tax system. While all this is legal, it nonetheless poses a clear potential strain on any national economy.
[10] For a detailed criticism of the ACCEPT project and the conflict of interest in Translators without Borders’ board, see this post.