It’s all about Trust.
There’s an enthralling analogy between aircraft operation and language services. To conclude both have adequate technology but still need human involvement. Like an aircraft cannot be completely flown without a human pilot; similarly it would be unwise to depend on entirely automated technology. Hence, I conclude a complete shift to machine translation will not happen in the language industry.
Ever wonder why the doors to the pilot cockpit on commercial airline flights are left open during passenger boarding? It appears that out that the passengers have a charming and very human need to peer into the cockpit and see the pilots sitting there. The airlines know this, so they leave the doors open for the passengers during boarding so we can all scratch this persistent little itch of need and reassurance.
It’s also why the pilots talk to us over the intercom during the flight.
So why all this soft-touch hand-holding of passengers? The pilots are actually flying the plane for one reason only: Trust. The plane can fly perfectly well without them and in fact mostly does. If you think translators feel threatened by the encroaching wave of next-generation machine translation software, consider the case of poor commercial airline pilots.
Computer automation today – the current state of technology – can fly commercial aircraft more safely, reliably and flawlessly; respond more nimbly and quickly to changes, and navigate seamlessly through more potential horrific disasters better than any human pilot can.
That’s a potential financial blockbuster for an eternally struggling airline industry. So why hasn’t that happened yet?
The passengers. They would stay away in droves.
Even a hopelessly romantic technophile like me would think twice about flying across the country on an aircraft flown entirely by computer. Although the airlines would sell a LOT of alcohol on those flights.
Cost of Failure
The calculus is different for a passenger because the cost of failure dramatically outweighs every other consideration. The global civil aviation infrastructure has been built up over the last 50 years to maximize safety. It’s this investment in a massive safety culture that allowed commercial air travel to become a commodity. In practical terms, modern civil aviation beat the cost of failure right out of the equation.
Airlines still struggle to differentiate on quality, but they operate in a hyper-competitive, price-driven, razor-thin-profit commodity market because commercial air travel has been engineered to be safer than any other activity in human life.
Good for us passengers. Very bad for the airlines’ bottom line.
Risk and Trust in the Language Services Market
So now we come to the language services market, where the cost of failure is very much alive. In this context, translation customers large and small find themselves faced with the following scenario:
- They have a real (often vital) need for language services
- They can’t judge or assess product quality themselves
- They have a desire (like we all do) not to look foolish or embarrass their company
- They would like to keep their jobs
- Their business must succeed in a highly competitive environment
There are risks everywhere in that description – which is ultimately about human emotions in response to doubt, uncertainty, and vulnerability.
What’s most interesting about these requirements is that they mostly lie outside the clients’ own direct control. So their objective in seeking out language services suppliers is to mitigate all these risks. That means that their decision-making process extends to areas far beyond the scope of the physical translation alone.
The persistent laser focus on “quality” and post-editing of machine translation to the exclusion of what language services clients actually need, want and buy every day undermines the financial viability of that approach.
So in addition to translation “quality,” clients value a whole host of soft-touch capabilities whose purpose is to engender trust and confidence and protect them from disaster, just like the (live human expert) airline pilots do.
Here are some examples of how companies that purchase translation services seek to mitigate those risks:
- They dilute their risk by selecting multiple suppliers
- They choose suppliers with diverse linguistic and technical capabilities
- They utilize client review teams
- They request translation certification
- They require guarantees, warranties or the right to require revisions
In addition to alleviating obvious risks, language service providers also add value in many other important and essential ways because we see things our clients are unlikely to see. That’s because our clients typically are not language people, do not inhabit the target culture, are not experts in foreign character set encoding, do not give much thought to different time and date formats around the world, etc.
What if Machine Translation becomes perfect?
A reasonable objection to this whole argument is that improving machine translation to the point where it’s indistinguishable from human translation services does in fact address the trust issue and resolves it.
In my view that objection fails because the uncertainty never goes away. Machine translation engines deal in the currency of disambiguation and semantic mapping but never eliminate ambiguity or doubt or even claim to. So error and doubt are actually assumed to perpetually exist in the solution.
Even if this weren’t true, translation users for decades have sought – and largely succeeded – in offloading every possible conceivable risk and remedy onto the translation service providers themselves. Our customers are airline passengers and they expect us – the pilots – to make sure they get home safely. If we use automation in our work, it better not harm them or their business.
a. Reduced expectations and market realities
This stand-off of sorts has led to many unusually tense – and sometimes baffling – discussions between translation users who are increasingly demanding speed, savings and scalability via machine translation and hybrid MT/TM technology, while language service companies – fully aware of the downside risks that will come back to haunt them – are urging these same clients to abandon client review, adopt “reasonable” (dramatically reduced) expectations, and introduce end-user usability testing that measures outcomes rather than translation accuracy.
b. The fruits of long endeavors
These attempts to find common ground will continue in the near-term, but the two sides would be far better served if they focused on the herd of elephants milling around the room with “risk,” “trust,” and “liability” signs hanging around their necks.
In the coming decades stand-alone machine translation and other high-speed language automation technologies are going to branch off and serve different markets than will human translation, with its focus on absolute precision, insight, creativity and impact. This is especially true of the boutique end of the market where even today there are not enough skilled human translators to meet demand.
We are already seeing new emerging markets where translation needs are beginning to explode, exact precision is less urgent and speed is paramount – online help desk forums, chat rooms, and quick feedback surveys are some examples, with the Twitter universe standing next in line.
And of course translators – the ultimate technology early adapters – will continue to be among the most ardent users of appropriate language technology in their day-to-day work lives.
Happily this means that the global language services market will diversify and expand in all dimensions to accommodate these new realities. It’s helpful if we all recognize that technology rarely destroys; it more commonly amplifies, or explodes in a galaxy of new choices. Machine translation itself is just a new river in a vast ecosystem.