In this series of posts, I’ve tried to outline some of the ways that I think we could improve the provision of public sector digital services in the languages of Welsh and English in Wales. I’m hesitant to go any further than I have done in these posts and aspire to write guidance for a few reasons.
The first reason is that good guidance, like good communities or good gardens, looks easy but takes a lot of ongoing effort to establish and prune and maintain. Good guidance needs the input of a multiplicity of views and backgrounds and experiences and it needs testing, iterating and evolving over time. (I sometimes think people forget – or don’t realise – how much hard work from so many people has gone into creating, developing and sustaining the Government Service Standard, for example.)
The second reason is simpler. There may be someone with the full breadth of skills in translation and service design and who works with services on a daily basis but it’s not me, and I’m a long way from being the best person to do this. I wrote these posts because I’ve been hoping and waiting for someone else to do it for so long that I’ve reluctantly decided that it is better to write a post than curse the empty and irrelevant google search result page. I’m sure that there’s a lot that I’ve got wrong, and a lot that I’ve missed, and I would be delighted to come back and delete these posts in future because they’ve been replaced by better work.
The final reason is more complex. The problem with writing guidance on multilingual or bilingual service design is that a lot of it becomes indistinguishable from guidance on service design:
- using machine generated outputs without human intervention and checks is a bad idea, whether you’re talking about language options or algorithms;
- a service is for life and not just for Christmas – creating a product or service without the ongoing funding to continue to improve and develop it is also a bad idea: fund teams not projects, don’t translate something once and then leave it there for years;
- if the (multidisciplinary) team is the unit of delivery, then work needs to happen within that team and not in silos, whether that’s content design in two languages or development and interaction design;
- metrics that are only quantitative and not also qualitative are misleading to the point of being harmful, whether you’re talking about services in different languages or about the social media awareness campaign I once sat through a report for where the reach of the campaign was reported as being in excess of the total population in the country covered by the service in question;
- if you test services with users to identify which parts cause them problems so that you can improve them, then you should iteratively test the digital service and the telephone service as you should also iteratively test the Welsh language service and the English language service;
- accessibility is easier if you include it as a requirement from day one rather than trying to retro-fit it, so is working in the open, so is language.
I could go on. The only unique aspect is that language is not separate from service design: a service does not exist hovering above language, simply waiting to be cloned into multiple languages; it exists uniquely and differently in each language.
When I was managing the creation of web content in the languages of Welsh and English, people often complained to me that Welsh was a wordy, verbose language. Why did they say this? They said it because they were typing Welsh language text over the top of English language text onto an existing web page and finding that it didn’t fit in the space that the English language text had used. This often annoyed them because it caused them problems, particularly in overcrowded navigational/IA areas.
I don’t know about the wider truth, but the truth about the content that I had translated was that neither Welsh nor English content was more verbose, but that any content that was translated was usually 10-15% longer than the original. If it started in Welsh and was translated into English, the translated text was about 10-15% longer; and if it started in English and was translated into Welsh, exactly the same happened. I assumed that this was because you tend to be extra descriptive in translation to make sure that you’re accurate. I think this point is also true for a lot of other scenarios: if you begin with one language, and use that as the prism through which you see your service in all other languages, then you – and your service – will end up with some inaccurate assumptions.
I started this series by saying that language is more a part of our world than we sometimes realise. It seems particularly appropriate that the first known example of written Welsh appears not in isolation but, a little time before the first known example of written English, next to another language. This series of posts stops here, but I hope that we can continue to talk a little more about language and service design.