spinofflive
Image: Tina Tiller
Image: Tina Tiller

InternetFebruary 3, 2022

When you can’t learn a language from your ancestors, learn from an app

Image: Tina Tiller
Image: Tina Tiller

Many people in Aotearoa long to learn the languages that their families could speak in the past. For IRL, Shanti Mathias explores the potential of online learning for those who want to reconnect. 

In March 2020, the month the pandemic started, Dani Pickering discovered something amazing: their ancestor had written journals about being forced out of the remote Hebrides Islands by British colonisers and starting a new life in 19th century New Zealand. Better yet, the journals still existed five generations later. One was in the Auckland War Memorial Museum. After a long journey discovering their ancestry, Pickering, a PhD student in Te Whanganui-ā-Tāra, was delighted to learn that they could access a piece of their past. 

There was just one problem: the journals were mostly written in Scottish Gaelic, a language that nobody in Pickering’s family had spoken over those five generations. But there was a way to learn, because the Scottish Gaelic course had just launched on Duolingo. 

“I took that as a sign from the universe to learn my ancestral language,” Pickering says. Two years later, Pickering’s phone screen is filled with colourful cartoon badges signifying language learning achievements, and they have almost finished the Scottish Gaelic course. 

US-based corporation Duolingo is by far the most well-known and popular language learning website and app; the company claims to have more than 40 million users, and was worth more than $6.5 billion dollars at its IPO last year. It is one of many sites, with names like Mango Languages, iTalki and Babbel, that promise the opportunity to learn a new language on your phone. 

Dani Pickering visits their Scottish ancestor’s grave in Auckland. (Photo: supplied)

“I feel really gutted about it,” says Gina Dao-McLay, a student based in Wellington, of their inability to speak fluent Vietnamese. While Dao-McLay was able to speak Vietnamese as a young child, later exposure to English at school meant that they lost their language. When, as an adult, Dao-McLay wanted to practise Vietnamese in preparation for a trip to see family, Duolingo seemed like a good option.

“I can’t really connect with people … without knowing the language,” Dao-McLay says. In their desire to learn, they aren’t alone: according to Duolingo’s 2021 Language Report, more than 70% of people who have started learning a language in the past two years say they are motivated by family heritage, ancestry or culture. 

“[Without language] you feel cut off from a big part of your culture,” says Alex de Vries, a masters student based in Auckland. He was delighted that Barakat, a film in Kaaps, the language from where his family lives in South Africa, was released last year, “but it was very confronting to have to read the subtitles for the whole film – like you’re a part of it, but you’re not.”

Gina Dao-McLay has made multiple attempts to learn Vietnamese on Duolingo, but has found the process frustrating. (Photo: supplied)

I know how de Vries, Dao-McLay and Pickering feel, because I, too, have longed for language. My father’s parents spoke to him only in English when they moved to New Zealand from India in the 1970s, so he never became fluent in Konkani, the language of our South Indian ancestors. Instead, I grew up learning Hindi, a language I know from classrooms and bazaars, but my halting fluency is not enough to make me comfortable speaking it with friends or at home. Missing India, at 19 I downloaded Duolingo and started tapping through the Hindi course, but filling in the blanks and parroting simple sentences couldn’t give me what I wanted: to not flounder when I try to describe a thunderstorm; to reach for words in Hindi and find them waiting; to be certain I belong. 

Being able to learn a language without talking to a real person who speaks it is a relatively new phenomenon. If Pickering’s ancestors hadn’t been driven out of their homeland; if de Vries’ family had never left South Africa; if Dao-McLay had been able to attend a school where they weren’t the only Vietnamese person; if colonisation hadn’t violently eradicated people’s abilities to speak their own languages, all over the world, things might have turned out differently for many of us. We might still be speaking te reo Māori or Vietnamese or Kaaps or Gaelic with our families, not learning from a machine. But while there’s no way to change the past, it is possible, even easy, to download an app.

Duolingo, which claims it is the most effective way to learn a language online (as do many of its competitors), uses particular strategies to get users hooked. The colourful achievement markers Pickering has gathered is one such motivator. The app also sends insistent reminders to users who aren’t opening it frequently enough, and offers various advantages (as well as the removal of ads) to those who pay.

After I re-downloaded Duolingo as I worked on this story, the passive-aggressive owl notification popped up each day, cajoling me to practise. The prod to maintain a streak successfully convinced me to open the app. But behind the bright icons was a reminder that algorithm-generated sentences about Raja going to Delhi for five months couldn’t fulfil my desire for a different kind of fluency. 

Duolingo staff celebrating after a successful IPO. Despite a billion dollar valuation, the app relies on volunteers to develop new courses. (Photo: Duolingo)

Duolingo is an example gamification, where features of a game are folded into activities that are not normally game-like. It can be motivating, but it also flattens the nuances and beauty of a language into just more content to consume. While Pickering notes that Duolingo “isn’t my daily Candy Crush,” they do feel conflicted: proud of their 700-day learning streak while also aware that the reasons they’re learning Gaelic are far deeper than the desire to maintain a position in the Diamond League

While the digital world seems limitless, those who seek to learn online are often reckoning with a lack of materials for all but the biggest languages. “There are no resources to learn Kaaps,” de Vries says, although he’s done his best to find what is available. It’s important to him to learn Kaaps, specifically, as a Black language, a dialect of Afrikaans drawing on Javanese, Khoekhoe, and Dutch – a history of empire and resilience held in vocabulary.

Another issue: the vocabulary and pronunciation taught by apps teach is decided by the creators of the course, erasing regional differences. While Dao-McLay’s family is from northern Vietnam, they found that Duolingo taught the southern dialect of the language. “With some of the words, it was like – that isn’t how my family pronounces it, that isn’t what I want to learn,” they say.

Lack of resources means that smaller languages don’t have additional features on the app, like the ability to speak into the phone to practise pronunciation. “I found Duolingo limited for speech but there were other apps in the ecology of online language learning that fill that role,” Pickering says. 

While all humans are uniquely hardwired to learn language, that doesn’t mean it’s easy. Pickering walks me through their Duolingo routine: a half hour of practice a day; adding each new vocabulary word to a set of flashcards; joining a Zoom group of other Gaelic learners; listening to a Gaelic-learning podcast on the weekends while doing the ironing. How long does this all take, I ask. They add it up on their calculator: at least four and a half hours a week. For them, Duolingo works because they treat it like a course, not a game. 

In Aotearoa, where the demand for te reo Māori teachers is at record levels, could online learning plug the gap? A muchanticipated te reo Māori course has been in the Duolingo Incubator for a number of years, although the release of the beta version has been pushed back several times. A Duolingo spokesperson told The Spinoff there is still no release date confirmed for the course. One reason for the delay has been a lack of resources, as the course was originally developed by volunteers (the company’s volunteer contributor programme has since been closed, and all contributors are now paid). 

Pickering has thought about what it means to learn a language from a company that commodifies their ancestral language by making them watch ads or pay a subscription to keep learning. “Duolingo’s interests are not our interests. It’s a corporation. This is capitalism.”

But while important, those concerns fade against the sense of connection they’ve gained from learning Scottish Gaelic. Since starting the Duolingo course, they’ve visited the grave of their ancestor – the one who wrote the journals – in Waikumete Cemetery four times. Each visit is a reminder of why they have chosen to learn the language that their family lost. “I speak to him in Gaelic … each time, I can put together more complex and articulate sentences,” Pickering says. “It’s given me a place to stand.”

Image: Tina Tiller
Image: Tina Tiller

InternetFebruary 2, 2022

What’s the privacy cost of your ride across town?

Image: Tina Tiller
Image: Tina Tiller

A new report has found that Uber gathers more user data than almost any other ride hailing app worldwide. For IRL, Dylan Reeve explores what this means, and how worried it should make us. 

Recently, VPN company Surfshark published a blog post titled ‘The most data-hungry ride-hailing and taxi apps‘, in which it looked at popular taxi-style apps worldwide, like Uber and Ola, and ranked them based on an assessment of their privacy impact. 

“Ride-hailing has become completely digital over the last decade,” the blog post reported. “And it’s not just money we pay for this convenience, but it’s also our data.”

The study ranked 30 applications from around the world according to the “data sensitivity index” the researchers created. The highest (worst) scoring was GrabTaxi, an app widely used in Singapore and Vietnam, with 114; of the services available in New Zealand, Uber scored highest with 80 points, followed by Ola with 52.8 and Didi with 38.4.

New Zealand’s other major ridesharing provider, Zoomy, was not included in the Surfshark study, but using the framework provided with the research, The Spinoff calculated a score of 31 for Zoomy’s iPhone app.

But are these numbers useful, and what are they really measuring? 

What Surfshark has done seems unique: they gathered privacy data from the Apple App Store for each of a number of rideshare apps and created a formula to measure the privacy impact for each one. The data comes from application privacy disclosures that have to be made by software developers who are making apps for Apple devices, and data points are broken into three categories: data used to track you; data linked to you and data not linked to you. 

Uber’s app privacy summary in the App Store.

Surfshark uses a pretty simplistic formula: for each of the 32 data points Apple potentially provides details about, they allocated three points for each time it showed up in the “data used to track you” category, two points for “data linked to you” and one point for “data not linked to you”. Add all those up and you get a total. Then, to account for the added privacy implications of sharing your info, they add a 20% penalty for companies that say they share your data with third-party advertisers. 

It can all seem pretty daunting and worrisome, and when you see the details laid out, it can be hard to imagine just what a glorified taxi service needs with any of it. In the old days, we called a number and a car turned up. If things went well, the sum total of our communication with the driver was just saying where we were going. So why does Uber collect so much data?

Privacy advocate and NZ Council of Civil Liberties spokesperson Thomas Beagle points out that New Zealand’s Privacy Act contains the principle that companies should only be able to collect the data that they need to use the service offered by the company. “I support the practice of collecting just enough data and no more,” he said. “If companies are failing to do so, we need to look at ways of enforcing the law.”

But who’s to say how much data a company needs in order to run a modern taxicab service? Obviously they need to know who you are. And your credit card info to pay. And I suppose they need your location in order to pick you up. And where you’re going. And maybe your phone number for future contact. And maybe they want a photo of you for your profile. And their programmers probably want info about how well the app runs on your phone.

It’s hard to say say how much data a company needs in order to run a ride-hailing service. (Photo: Nathan Stirk/Getty Images)

When the data points are all laid out it can seem excessive, but as you look at each one it is easy to understand what the legitimate use could be. So what does Surfshark’s study really tell us?

Well, not very much. For a start, it relies on Apple’s system of demanding information from developers and showing it to users, which doesn’t actually tell us a whole lot. Firstly, it is largely reliant on an honesty system of sorts, with developers ticking check boxes when they submit their software. Apple would probably catch any big omissions, but mostly they’re relying on what’s supplied. Secondly, as users, we can’t know what actually happens to any of that data once it’s collected: does it just sit in a big file somewhere, or are the software developers using it to build advanced profiles about our day-to-day lives? 

Another limitation is that Surfshark was only able to consider the Apple versions of these apps. (At this stage, Google isn’t disclosing the same type of information for Android applications, but they’ve announced plans to implement something similar in the near future.)

The Surfshark study gives you a nice overall number, but not a clear indication of what privacy you might be giving up. Is your real name more or less “private” than your email address? In the Surfshark methodology, there is no differentiation between the sensitivity of information. 

“The privacy point-scoring system is a brave attempt at creating an index, but it’s too one-dimensional,” said Beagle. “Most importantly, there’s a big difference between a company collecting the data for itself and passing it on to others, and the 20% surcharge for data sharing doesn’t reflect that.”

All of this raises another question: what can we do about any of this, anyway? You’d assume, given the source of the research, that perhaps using a VPN might protect your privacy in these cases, but nope. Surfshark’s blog post offers no clear answers because, it seems, there aren’t any.

Almost all of the data in question is connected to information you provide when signing up to the application, or is somehow derived from your phone itself. While a VPN could disguise one or two small pieces of information, it does little to preserve your privacy from the developers of applications you’ve chosen to install. 

“Most people don’t read privacy agreements, they’re not going to review app privacy settings, and even if they did, it’s not like you can negotiate for a better arrangement,” said Beagle. “If you want to use the same app your friends are using, you have to accept the deal they offer, and that’s that.”

Ultimately, we each make choices about how worried we are about our private data, and which companies we trust with it. Despite improvements from Apple, and eventually Google, we still find ourselves really having only one decision to make about protecting our privacy: use the app or don’t. Beyond that, we don’t get much say in what data our phones will give up: trusting some big company with assorted private data is the trade-off we make in exchange for carrying a super-computer in our pocket – and for getting a cheap ride across town.

Unfortunately, Surfshark’s study, while presenting a simple number to look at, doesn’t actually shed much light on the bigger picture.