Ukrainian migrant asking ChatGPT to explain different types of rental contracts in Austria (still from Yudytska & Androutsopoulos 2025)

For nearly four years now, I’ve been heavily involved in online (Telegram-based) communities for Ukrainian forced migrants in Austria. Such communities sprang up all across Europe after Russia’s full-scale invasion of Ukraine in early 2022, serving as grassroots digital info-points. In those early, chaotic days, the groups focused on a Ukrainian’s first steps upon getting off the train in their new country. They have since expanded to cover everything from where to buy buckwheat (a very pressing question!) to the fine details of the local education system, health insurance, job market, etc.

The communities are run for Ukrainian refugees, by Ukrainian- and Russian-speakers, that is, overwhelmingly by migrants – forced, labour, student; first-gen, second-gen; etc. It truly is mutual aid: those who migrated earlier answer questions on integration, those who left recently give tips on how best to reach relatives in occupied territory. My family left in 1999, when I was five; I co-admin one group for Ukrainians in my home state of Upper Austria (ca. 3,700 members) and one group focused on how to change from the Temporary Protection status to a longer-term work/residence permit (ca. 8,300 members), plus I occasionally help out in other groups.

In 2025, our communities gained one more, less welcome, “member”: ChatGPT.

Migrants’ search for information

Understanding the role of ChatGPT in these communities requires understanding the complex informational space newly arrived migrants are in. There is an avalanche of info to take in. Some involves the everyday differences: when your child has a fever, at what temperature can you call the ambulance without getting in trouble for wasting their time? Other aspects are specific to the (forced) migrant status, and would stump a local too. Language barriers make accessing information difficult. Even with machine translation, it’s hard to know what term to google or how to interpret a bureaucrat’s answer.

This is where the Telegram communities come in, providing an easy place to ask for help and discuss problems.

An NGO’s German-Ukrainian sign about food and clothing distribution; the understandable but incorrect Ukrainian suggests machine translation (photo taken by author)

Still, they’re not a panacea. Us volunteers learned only on the fly the ins-and-outs of the Austrian asylum system in all its kafkaesque glory: we sometimes misremember, misunderstand, and mistranslate. Grasping a question can take real detective work, especially when there’s a dozen ways to translate a single German term. Official Austrian sources give a lot of information orally; this leads to bewildering discussions where one person informs the chat their social worker said X, another says it was Y, and the third Z. Maybe it’s miscommunication, maybe the official sources have no clue either.

Above everything loom the rumours, best illustrated by the (admittedly pretty sexist) Russian acronym, ОБС – одна бабка сказала, lit. “some old lady said”. It’s used in the sense of, “Uh-huh, riiight, you got that info from your cousin’s friend’s mom’s neighbour”. In 2027 all Ukrainians in the EU will be put on the next train home, even if the war is still ongoing. Truth or ОБС? Great (unanswerable) question!

Enter ChatGPT

Into all this chaos slams the iron certainty of ChatGPT. If it worked as advertised, it would be an invaluable resource for migrants. You ask a question, it searches for and provides a summary of official information. No linguistic barriers, no mistranslations, no complex legalese; no petty online fights to wade through. The appeal is clear.

The only tiny problem is that it doesn’t work as advertised.

The biggest issue is in how ChatGPT functions: it doesn’t search for and copy-paste information from a database, but rather generates a statistically likely sequence of tokens (words) based on the data it was trained on. This is why it can “hallucinate”, that is, make up answers which are linguistically coherent but factually untrue. It has also obviously been fed more data (law databases, official websites, forum discussions, etc.) from Germany than Austria. For example, it has confidently explained to a Ukrainian that she’s legally obligated to have health insurance – true in Germany, not so in Austria.

Linz Castle on the Danube, Upper Austria, lit up in the colours of the Ukrainian flag (photo taken by author)

The AI technique ‘Retrieval Augmented Generation’ (RAG) is meant to resolve this: ChatGPT first searches for and pulls relevant information from a website, then incorporates it into the answer. (ChatGPT uses RAG sometimes, but not constantly. It costs more energy, thus more money.) But the answer is still generated, so hallucination is still possible. This leads to ChatGPT claiming, for example, “According to the City of Vienna website, Ukrainians have to hand in their old refugee ID card. [Link to website]” Except its generated summary missed a negation: the website explicitly states Ukrainians do not have to hand in their old refugee ID card. RAG can thus lead to even greater misinformation, as it implies a direct source.

The other huge problem is the limited information actually available on migrant issues; even if hallucination was somehow solved, the adage of ‘garbage in, garbage out’ for machine learning remains. For example, to switch to a longer-term residence permit requires the migrant to have an ortsübliche Unterkunft (“housing according to local standards”). ChatGPT physically cannot answer what these local standards are. I know this because, according to my old-fashioned research techniques of searching law databases, skimming court cases, and asking lawyer friends, neither can anybody else! Unfortunately, it is too often the case that there is too little concrete info for ChatGPT to be trained on.

In short, using ChatGPT for information is a gamble. Sometimes it works great, sometimes it generates nonsense. To know which is which, the migrant is left with the same problem of surmounting the language barrier they started with.

Mutual aid with and against machines

Rows of donated women’s shoes at SUNUA, a grassroots organisation supporting Ukrainians in Upper Austria, 3 weeks post-invasion (photo taken by author)

I don’t consider myself a techno-pessimist – actually, I rely on language technologies heavily in my online volunteer work. My phone’s autocorrect is a life-saver. While I can read and write in Russian, I left Ukraine before starting formal education, and find it slow and frustrating to spell without autocorrect. Similarly, machine translation is a great help. I occasionally need it to double-check my understanding of more complex Ukrainian-language questions; I also machine translate German-language official updates for the sake of speed, then post-edit the Russian text to correct mistakes and stilted phrasing before posting.

That is, due to my background as a child migrant from a primarily Russian-speaking area of Ukraine, I have varying levels of competencies in the three languages I need: Russian, Ukrainian, German. For all their faults, language technologies are an invaluable resource for stuffing the gaps so I can help people successfully. And efficiently, as this has always been 100% unpaid labour in my free time, next to my full-time uni work.

But all this is why I find the intrusion of ChatGPT into our spaces so infuriating on a personal level. In the midst of a horrible situation, between fear, grieving, trauma, burnout, we’re all trying to use the linguistic and technological resources available to us to help each other. I accept arguing against a community member’s cousin’s friend’s mom’s neighbour’s experience as part of that – that connection is also a resource, and it’s human to trust an acquaintance over a fuzzily written law in a language you don’t speak yet. I’m willing to spend my free time picking apart where the confusion lies.

The author’s post in a Telegram group explaining that AI as a source must be clearly stated, with over 60 users leaving reaction emoji of agreement (screenshot taken by author)

I don’t accept arguing with ChatGPT screenshots.

ChatGPT adds beautiful formatting, with eye-catching emoji as bulletpoints. ChatGPT switches between Cyrillic and Latin easily, writing out German acronyms and translating them to Russian in brackets: ÖGK (Österreichische Gesundheitskasse – Австрийская касса медицинского страхования), wow. ChatGPT cites laws using a fancy § paragraph sign I have to copy-paste each time.

Sure, the actual information may or may not be correct, but that’s less important than the style, which so neatly mimics that of official sources. It simply looks trustworthy – the complete opposite of my own messages, written one-handed on a moving bus, with at least one butchered case suffix apiece. It’s unsurprising that people cling to ChatGPT’s information more stubbornly than to the usual ОБС.

Arguing against it is thus extra tedious and, frustratingly, requires me to do additional work. “Wait, no, where did you get that information from, is that an official source?” “Uh, nope, that website ChatGPT cited says the opposite.” “Dude, did you actually read the law ChatGPT ‘references’ – it’s about industrial chemicals, not document translation.” Most unsatisfactorily for all involved, having to prove a negative: “I’m sorry, but there’s no information on that. I don’t know why ChatGPT says there is, maybe it exists for Germany, but not for Austria.”

As a migrant who’s also stared at bureaucratic German in confusion and anxious despair, I don’t blame people for turning to AI. As a volunteer, it’s genuinely made me want to quit: in anger, in exhaustion, with a childishly vindictive, “Well, if they prefer machine over human, so long and thanks for all the fish.”

Where to next?

In principle, the issues explored here are no different from those we’re facing in other areas: education, academia, news, etc. Misinformation is rife everywhere; so is a lack of digital literacy on what current AI can and can’t do. For me, the crucial point is how vulnerable forced migrants are. Misuse of ChatGPT can lead not to a failed homework assignment but to problems on an existential level: with the legal status, with housing, with having enough money for food. Similarly, to put it bluntly, I’m paid to deal with students’ AI use; in volunteer work, it’s just one more weight tipping the scales in favour of finally quitting. It’s also important to add that not all my fellow admins share my worries. Some eagerly embrace ChatGPT answers themselves, which of course save time and energy for volunteers who have little of either to spare.

Our current ‘solution’ is that people must state openly that the information they’re posting is AI-generated. Then other members can decide themselves to what extent they trust it. In this, I would say, we’re ahead of the curve compared to many organisations, and for now, this will have to be enough.

Reference

Yudytska, J. & Androutsopoulos, J. (2025). The use of language technologies in forced migration: An explorative study of Ukrainian women in Austria. In M. Mendes de Oliveira & L. Conti (eds.), Explorations in Digital Interculturality: Language, Culture, and Postdigital Practices (pp. 135-166). Transcript. https://doi.org/10.1515/9783839476291-007

Some references made in this episode include:

• Language and the Internet (2001) by David Crystal
• The recontextualization of Multicultural London English: Stylising the ‘roadman’ (2023) by Christian Ilbury
• The Dolly Parton meme
• What’s the difference between AAVE and Gen-Z slang?
• Shetland Islands accent

If you enjoy the show, support us by subscribing to the Language on the Move Podcast on your podcast app of choice, leaving a 5-star review, and recommending the Language on the Move Podcast and our partner the New Books Network to your students, colleagues, and friends.

Transcript (coming soon)

***

Number of languages present in the ELP repository (March – June 2020) against the background of the world’s ca. 6,000 languages

The outbreak of the COVID-19 pandemic has stimulated an unprecedented amount of translation work in both major and minority languages of the world. The online repository of the Endangered Languages Project (ELP)—which houses links to most of the materials about COVID-19 that are available online in “Indigenous, Endangered, and Under-Resourced Languages”—currently contains around 1,500 resources in close to 600 languages. About 520 of those languages can be considered under-resourced or otherwise marginalized.

No one would question the fact that this is good news: if we are to counter a pandemic, then health-related information has to reach every single person on the planet, and it looks like we are going in that direction. Right? Well, at a closer look, the picture is not as crystal clear as it would appear on the surface, as I will try to outline in this post.

Understanding a message means accepting it? A key but forgotten issue

Let’s start by saying that, from a quantitative point of view, there still remains much to do: if we consider that there are more than 6,000 languages currently spoken in the world, resources are still lacking for more than 90% of them.

But, one could argue: does this matter? Reaching every person on earth is not contingent upon having resources in all the world’s languages because most people on earth are multilingual, right? And here comes the key point: Can we consider translation work as a purely mechanistic endeavor so that, by making a verbatim translation of a resource from language A to language B, we will be assured that those who understand language B will respond to the message as expected? If people understand a message, will that automatically mean that they accept it?

Learning from the past

Previous experience during the various Ebola outbreaks (2014-2019) in West Africa and the Democratic Republic of the Congo (DRC) has shown that without a clear understanding of the specific cultural situation where communication about risk takes place, translation efforts may be in vain, if not counterproductive. In reporting research done by Translators Without Borders in DRC, Kemp (2020) informs us that “study participants voiced frustration with information like ‘You have to go early to the Ebola treatment centre to be cured.’ [study participants] want details on complex issues to inform their decisions, and they want them presented in what they referred to as ‘community language’—meaning in a language and style they understand, using words and concepts they are familiar with.”

The “do the 5″ poster in Torwali, a minority language spoken in the Khyber-Pakhtunkhwa province of Pakistan

Some further suggestions are provided by Bastide (2018), who worked in West Africa: “one cannot understand the [Ebola] outbreak in West Africa without clarifying its relation to colonial and postcolonial medical practices across the region; in the same spirit, it is difficult to understand the crisis without looking at prevailing social imaginaries [i.e. culture-specific ways of interpreting the world] to understand vernacular rationalizations of the event, involving culturally and socially formed expectations; and it is difficult as well to make sense of the situation without looking at how the event reshuffles social arrangements … as when Ebola intersected with political elections, or with social relations by restricting body contacts or disturbing funeral rituals.”

Put differently, having a message that complies with the grammar and the lexicon of a given language is just the start of a translation work. The primary goal of risk communication is that people change their behaviors based on the new information they receive. So, the point is not that the message merely be understood, but that it be understood and accepted by as many people as possible. For this to happen, one must keep many aspects in sight in addition to the narrow linguistic ones: the words and metaphors that are chosen (as stressed by Kemp), the meaning that a message will evoke by the simple fact of being delivered by certain institutions or individuals, the use of certain channels and styles (as alluded by Bastide), to name a few, are all factors that affect how people will respond to a message. How many specialists conceive of the possibility that much of the actual impact that the translations listed in the ELP repository will have on the ground depends on these “emotional” aspects, rather than on the mere “well-formedness” of the message? It is difficult to tell. It remains a fact that little attention overall is being given to this key problem in the public discourse with regard to COVID-19 translations.

From a message-centered to a community-centered perspective

Conventional translation work pivots on the message itself: this message-centered approach is based on a high degree of control of the source message and of the type of translation—i.e. verbatim translation—aimed to obtain an output that can be created easily and diffused quickly. A typical example is the “do the 5″ poster model that Translation Commons has promoted. More than 10% of the languages in the ELP repository have a “do the 5 poster” as their only COVID-19 health information resource.

This is in stark contrast with what Kemp and Bastide (amongst others) are advocating, which is an approach that puts the particular community and its socio-cultural context, rather than the message itself, at the center of the translation efforts. This community-centered approach entails empowering the community itself as the producer of messages, and this is what a number of communities around the world have done for COVID-19 (e.g. listen to this audio message in Triqui, Mexico, or the video messages in Nahuatl described here). However, most communities speaking minority or endangered languages lack the necessary means, knowledge, and tools to do the same.

A “Town crier” in Bafut, North-West Region of Cameroon (ca. 1960, but the practice is still very much alive) (Image credit: Ritzenthaler, R. & Ritzenthaler, P. 1962, p. 101)

Out of the many commendable projects that have been launched during the COVID-19 pandemic, there is one that I have the honor to coordinate which is based on this community-centered approach to translation and aims to empower communities so that they can arm themselves with accurate information and create their own content. This project is called virALLanguages and is a volunteer-run initiative organized by the KPAAM-CAM project, the SOAS World Languages Institute, the Community for Global Health Equity of the University at Buffalo, and the Department of Linguistics of the University of Hawai’i at Mānoa.

virALLanguages in a nutshell

The founding principle is simple: only community members will know how to deliver a message that has the best chance of being accepted by their community. What virALLanguages does, then, is (i) transfer key information about the pandemic to motivated community members, along with some additional background knowledge, (ii) foster the creation of language teams made of at least two people (i.e. a speaker and a “judge” or reviewer, who has the key role of guaranteeing that the final message conveys accurate information), (iii) stress the importance of translating “authority” locally, which means having speakers that are widely known and highly esteemed in the community, and (iv) provide technical and logistical assistance for recordings to be created and diffused according to the various teams’ requests (e.g. videos on social media, audio messages broadcast in local radio stations, etc.).

To date, virALLanguages has produced more than 60 resources in over 30 languages of Cameroon, Pakistan, and Ghana: there are “big” languages like Pashto and Akan, but most are minority or endangered languages in which no other resources about COVID-19 prevention measures exist. A drop in the ocean of the world’s languages, one might argue. One feature that makes these numbers more significant, though, is that most languages are concentrated in specific areas due to the work of local collaborators: northern Pakistan (12 languages), northwest Cameroon (11 languages), and northern Cameroon (6 languages) are three areas where virALLanguages has concentrated its efforts, and we hope to be soon able to add Indonesia to this list thanks to the collaboration with the Department of Linguistics of the University of Hawai’i at Mānoa.

What is different in virALLanguages resources?

Along with the pandemic of COVID-19, the world is experiencing an infodemic of misinformation about its causes and treatments. In such a context, one of the key problems that virALLanguages aims to tackle is the fact that the credibility of a message must be evaluated from a local perspective. Just to take one example, most people on Earth don’t know what WHO (World Health Organization) is and, among those who do, there are many who have mixed feelings about it or only have a vague sense of what it does. This means that a message branded by WHO is not likely to have a great chance of being taken seriously in many communities around the world. This is the reason why many of the virALLanguages videos feature traditional leaders or other personalities that are meaningful to the local audience, like PhD students who are studying abroad, researchers, or singers.

Open embedded content from YouTube

This video is in Mendankwe, a language spoken by about 30,000 people in the area of the city of Bamenda, in the North West Region of Cameroon. It features Mr Herick Abongwa Forsuh, who comes from the family of the Mendankwe traditional rulers (Fon), as is visible from his necklace. The video hit 500 views in less than 2 days.

Over the course of the project, we learned that authority and credibility can be locally constructed in many other ways as well. For example, in some societies of the Cameroonian Grassfields there is a traditional figure that is expected to disseminate information about risks that threaten the society as a whole: this person could be described as the “Town Crier”. The Town Crier is normally associated with specific objects (a drum, a double-gong) and ways of speaking.

You can hear below how this traditional resource was leveraged by the Babanki language team in their recording for virALLanguages (Babanki is spoken in the Cameroonian Grassfields along with Bafut and about 70 other languages).

Open embedded content from SoundCloud

The audio message can be downloaded here.

Credibility can also be constructed through the use of rich language and the appropriate ways of involving the audience, as Dr Margaret Neh Chenemo (sociolinguist and speaker of Bafut) tells us in this video.

Open embedded content from YouTube

Conclusion

In a talk I recently gave with Mandana Seyfeddinipur (co-director of virALLAnguages along with Jeff Good), we provocatively raised the question “Can linguists save lives”? Our claim was that, in order to save lives, linguists must be clear on the fact that risk communication is in and of itself an issue of what Dell Hymes called “communicative competence”, rather than linguistic competence alone. As I tried to argue in this post, message-centered, verbatim translations are better than nothing, sure, but they leave too many gaps and potential sources of misunderstandings to be considered sufficient if the goal is to convince people to change, albeit temporarily, their behaviors. What we actually need is community-centered, credible messages produced by empowered community members and delivered by locally accepted authority figures and through locally meaningful ways of speaking. VirALLanguages is just the first initiative to take up this challenge and illustrate not only that it can be done, but also how it can be done effectively. As lead coordinator of virALLanguages, my greatest hope is that we will soon be able to measure its impact so as to test, rather than just argue, that this is a promising way forward for impactful risk communication—even beyond COVID-19—in today’s connected and threatened world.

Acknowledgments

virALLanguages would have never seen the light without the determination and intelligence of four Master’s students in Language Documentation at SOAS – University of London (in alphabetical order): Leonore Lukschy, Yẹwá Ògúnṣẹ̀yẹ, Sydney Rey, and Vasiliki Vita. This post is based on the talk I and Mandana Seyfeddinipur gave on May 28 2020 at the SOAS Alumni “Continuing the Conversation” event series. I thank Sydney Rey and Jeff Good who have commented on a previous version of this post. Thanks go also to Anna Belew, who provided me with updated figures about the ELP repository.

References

Bastide, L. 2018. “Crisis Communication During the Ebola Outbreak in West Africa:​ The Paradoxes of Decontextualized​ Contextualizatio​n”. In M. Bourrier and C. Bieder (eds.) Risk Communication for the Future. Towards Smart Risk Governance and Safety Management, 90-104. Springer. (This is a free ebook)
Kemp, E. 2020. “Replacing the language of fear: language and communication in DRC’s latest Ebola response”, March 2020, odihpn.org
Ritzenthaler, R. & Ritzenthaler, P. 1962. Cameroons village: an ethnography of the Bafut. Milwaukee: The North American Press.

Language challenges of the Covid-19 pandemic

Visit here for our full coverage of language aspects of the COVID-19 crisis.

My “audience” as I was recording the first online lecture for the new term

Yesterday, I spent six hours pre-recording a puny little lecture of 15 minutes for the postgraduate “Literacies” unit I’m teaching this term. The unit has gone fully online this year due to the Covid-19 pandemic and I have been planning for interactive delivery in a variety of formats.

One element in the overall mix is a podcast series. I’ve taught the unit a couple of times already so have the content down pat and figured all I needed to do was sit down and deliver my lecture into a microphone. It did not turn out to be a smooth experience.

The content I was covering yesterday – features of written vs spoken language – usually takes about 40 minutes of class time to deliver. That includes asking questions, taking student responses, and summarizing those responses. A standard teacher question-student response-teacher feedback cycle.

Without dialogue, the lecture shrunk to not much more than a third of the time it would normally take but producing it blew out by about nine times.

Most of this production time is a one-off, as I needed to learn how to use Adobe Audition and spent a lot of time designing an intro and an outro, and figuring out how to overlay them with a signature tune (I chose a few bars of Vivaldi’s Spring Concerto :-). Including a signature tune is a playful option that is obviously not strictly necessary but was fun to learn.

Quite a bit of time also went to editing in order to smooth out bloopers.

Hot tip: If you are unhappy with anything you’ve recorded, don’t stop recording. Instead, pause, click your tongue three times, and repeat whatever went wrong. This way you can easily identify the bits you’ll need to cut in your voice editor.

I may have smoothed out major bloopers but the final product still doesn’t please me and doesn’t meet my usual standards of work. I’m dissatisfied with recurring disfluencies, with too much detail in some parts and not enough in others, a joke that I started and then trailed off because it seemed silly delivering it to the unmoved microphone.

Seeing how much time I invested, I’m wondering where did I go wrong?

Maybe it’s not me at all but the problem is the genre of the academic lecture?

What’s wrong with lectures?

Lectures are odd creatures at the intersection of reading and writing, as a quick look at the table listing the key differences between written and spoken language will show.

The academic lecture, including in its pre-recorded version, is obviously a form of spoken language. However, most of its characteristics are typically associated not with spoken but with written language:

• The lecture is technologically mediated (recording device at my end, audio player at yours).
• Speaker and audience are distant across time and space (I recorded the lecture yesterday in my home and students will listen to it at other times and places).
• In terms of context, the lecture sits somewhere in the middle between high and low context (it’s part of a unit taught in the Applied Linguistics program at Macquarie University but it could be taught in any Applied Linguistics program in an English-medium program).
• The recording is durable and not as fleeting as the spoken word usually is.
• The lecture is not quite as scannable as a written text but you can certainly stop and rewind if there is something you didn’t understand, or jump ahead if you get bored.
• The lecture is planned and tightly structured.
• In terms of syntactic complexity and formality, I was aiming for a simple and casual style – the desired “conversational tone” of a podcast. However, on listening back, I discovered that I used a garden path sentence to exemplify one, and I also used words such as “therefore” and “thus” – clear traces of written language.
• I don’t even need to mention that the content of the lecture is relatively abstract (“Features of written language”) and that I delivered a monologue.

These mismatched criteria produce a “damned if you do, damned if you don’t” genre. For instance, I did not write up the lecture beforehand and so did not read out a script. In face-to-face teaching, I don’t need one and for pre-recordings the general advice seems to be that a script will make the lecture sound unnatural. Even so, I’m now I’m beating myself up for uneven delivery – there are a few unfinished thoughts and dysfluencies.

How did such an awkward genre become the main mode of university teaching?

Miniature drawing of a medieval lecture (Image credit: British Library)

The academic lecture has its origins in the European Middle Ages, when both literacy – the ability to read and write – and books were scarce. In a world where writing is cheap and literacy is almost universal, it is hard to imagine just how scarce they were back then. Only the most valuable information was committed to writing. Hand-written manuscripts took years to produce and books were a rare and extremely valuable commodity. Online courses, textbooks, even notebooks were still far in the future.

To teach the valuable information committed to manuscripts, early university education therefore consisted of a “lecturer” reading to an audience. A lecturer is literally a “reader”, a title still used in UK academia today for what is an Associate Professor in the Australian and US systems. The lecturer read the set text out loud, sometimes providing running commentary or explanations as they went along.

That explains why the lecture is such an odd cross-over genre between written and spoken language. It’s a written text read out loud.

What it doesn’t explain is why we are reverting to this mode of teaching as we transition from face-to-face to online teaching. The best explanation I can come up with is that technical affordances of the digital world have changed both written and spoken language in fundamental ways, and we are all still working out how to harness them best for learning.

What do you think about pre-recorded lectures? And what are your most and least favorite teaching genres? Have they changed between face-to-face and online?

As for me, I’ll try and mix genres as much as possible. Even if they make me cringe, I’ll keep podcasts in the mix for now, mainly because I want my students to experience another form of writing to learn: note taking. To commit something to memory and process it deeply, writing continues to be the medium of choice.

“To reach the mind, knowledge has to flow through the hand,” as one of my lecturers in teacher training kept insisting.

***

“The trend of the future is working from home. The big question is: Are Filipinos ready for this kind of work?” This question was asked by an employer of Filipino virtual accountants providing offshore services to clients overseas during my fieldwork in June 2018. Two years later, office workers all over the world find themselves forced to do just that—work from home—as a social distancing measure in light of the COVID-19 pandemic situation. In Australia alone, 1.6 million reported this significant change in their working conditions.

Unsurprisingly, there are many different reactions to this global shift to remote-work setup—some readily embracing it as the new normal, others taking a more critical stance. In the interim, as working from home continues to be the norm for some occupation groups, the experience of offshore accountants, who are employed to work remotely, pandemic or no pandemic, provides a picture of how this work arrangement works on a permanent basis.

A sociolinguistic analysis of the globalized accountant experience of working from home was the subject of a webinar co-organized on 3 July 2020 by the Linguistic Society of the Philippines (LSP) and the Lasallian Institute for Development and Educational Research.

In this online lecture, I explore how working remotely has shaped communication practices and ideologies in globalized accounting practice in the Philippines. Examining communication in this context is important as the demand for off-shored (including home-based) accounting services is increasing. This trend comes with the positioning of the Philippines as an emerging global provider of knowledge process outsourced services to businesses headquartered overseas. Ethnographic data collected from this work context for my PhD thesis (in progress) is analyzed using the lenses of performance and audit. Findings show that the way accountants communicate has evolved to fit the shape of virtual work environments. Digital solutions are making communication skills more salient and creating new norms and protocols of transparency that contribute to tensions between autonomy and accountability. The lessons highlighted from accountants’ experiences potentially reflect communication challenges and opportunities in other work domains especially during this period of COVID-19 pandemic, when mandatory physical distancing is redefining workplace interactions.

You can watch this virtual presentation uploaded in the LSP YouTube channel.

Language challenges of the Covid-19 pandemic

Visit here for our full coverage of language aspects of the COVID-19 crisis.