It’s been a while since we did this, but after a year as … interesting … as 2016, we thought we’d have a go at coming up with some predictions for the year ahead. The rules were simple: two predictions each from Nick, Tim, Laurie and Jo. Then in 12 months, we can all look back and wonder just what we were thinking.
1. Automated content creation will be everywhere
We’re already seeing evidence that EdTech companies are beginning to use automatically-generated learning content. For example, Duolingo’s assessment product features cloze activities that are clearly generated from out-of-copyright texts that happen to include a requisite number of target items (see screenshots below).
We also recently saw WeSpeke take over content creation duties on Newsmart – a product for which we at ELTjam used to create content using human writers and editors. WeSpeke will “use its proprietary and patent-pending Natural Language Processing (NLP) technology to create English lessons using world-class content from The Wall Street Journal and other Dow Jones brands.”
In many cases, the technology is nowhere near where it needs to be in order to create the level of content quality that an ELT publisher would expect, but bear in mind that a) it’s just a matter of time until it is and b) how many students actually know or care, especially if a product is free. One way to look at content automation is on a continuum that started with the switch from royalties to fees and ends with the use of algorithms to create learning materials that publishers would have had to pay authors to produce in the past. Bad news if you’re an ELT materials writer not able to retire in the next five years.
Nick Robinson
2. AI will give 1-2-1 tutors a run for their money
It’s no longer just about adaptive algorithms that are able to make suggestions for you based on your behaviour (Facebook feeds, adaptive learning content). 2017 will see the ascendance of AI-driven products and services that provide a much deeper level of service, particularly in the area of Intelligent Tutor Systems. These are able to replicate the experience of 1-2-1 tutoring and have been shown to give real live tutors a run for their money in terms of their effectiveness. Examples of such systems include Tabtor, Carnegie Learning and Front Row.
We’ve already seen an AI entrant in English language learning in the form of ELSA, and I’m sure we’ll be seeing many more to come.
Tim Gifford
3. Machine Learning will make automated translation mainstream
A theme already picked up on in Nick and Tim’s predictions, but we’re going to be hearing a lot more about machine learning in 2017 – that is, the ability for software not only to apply complex algorithms (that’s nothing new), but to actually learn in response to incoming data and user interaction and adapt accordingly. Look at how tools like Google Translate have improved in recent years (check out its ability to translate written and spoken language in real-time), then think about the potential of things like Skype Translator. The more these tools are used, the better they get, and we’re going to see exponential improvement. With the full force of the tech ‘big 4’ (Apple, Google, Facebook and Amazon) behind this, I think AI-powered real-time translation utilising machine learning might actually start to seem credible by the end of 2017. If that happens, then the impact on ELT won’t be far behind.
Laurie Harrison
4. Mobile will continue eating the world
In this 30-minute presentation, Benedict Evans of Andreessen Horowitz brilliantly captures where we are in the evolution of mobile. As we pass 2.5 billion smartphones on Earth and move towards 5 billion, the case for going ‘mobile first’ has never been stronger. If you combine that growth in device adoption with the huge leaps forward being made in machine learning, you can start to imagine a world of almost ubiquitous device ownership, with those devices carrying out tasks for us that we can only start to imagine now. In other words, this is about far, far more than trying to shoehorn coursebooks onto phone screens. If you’re currently working for an ELT product company that doesn’t have or doesn’t plan to have a mobile first strategy in place, then you might want to find one that does.
Nick Robinson
5. An ELT publisher will either pull out of ELT or radically change what they do
Every major ELT publisher has had at least one major restructure in the last couple of years, and the changes keep on coming without any sign of anyone finally having ‘cracked it’. ELT is not a core concern for any of the larger businesses that own the major ELT publishers, so there must be a limit to how long the current instability will be allowed to continue.
Laurie Harrison
6. Learner Experience Design will become the sought after skill in education
The increased focus on experience is something that is playing out in a number of domains. An interesting trend to compare this to is the increasing attention being given to the experience of candidates applying for jobs by HR teams. A poor candidate experience can turn off a whole tranche of talent to the prospect of working with you. Richard Branson has taken this to heart and is in fact crafting a fairly lucrative revenue stream out of taking care of the candidate experience for Virgin.
The same is happening (and will continue to happen) for learner experiences. ELT content publishers and creators can no longer assume that delivering language learning products is enough. The success of a product (and therefore the ongoing stability of a company working in the learning space) will increasingly be measured in terms of learner delight and the efficacy of the experience.
Tim Gifford
7. We will make more of what we already have
As more and more platforms and LMSs are built and left to wither and die due to unrealistic expectations of what could be achieved, we will see educational organisations making more of the tools that have bedded in as technological staples; things that will be with us for the long haul. This will mean more education through Google app suite, Facebook and Whatsapp, email, sms. Nothing groundbreaking or new in using these tools, but I predict that we will shift more towards relying on existing well-used technology rather than racing to build new platforms that try to do the same things, but manage it less well.
Jo Sayers
8. Harder times for immigrants learning English
As UK government cuts come into effect, and the impact of Brexit becomes more evident, times will be even harder for poorer people here in the UK, especially immigrants. People will feel more comfortable telling immigrants that if they are in ‘our country’ they need to speak ‘our language’ and this will be directly at odds with the fact that funding to support this language development is being cut, as are other benefits and services, making it even harder for immigrants to find the time and resources needed to learn English. This will make social integration even more challenging and further divide an already divided population. I predict that, on the whole, edtech advances will continue to be aimed at wealthy folks who can afford it rather than on effective solutions for supporting EAL needs in UK schools, or functional language and integration programmes for immigrants arriving in English-speaking countries.
Jo Sayers
Featured image from https://www.flickr.com/photos/manoftaste-de/
Happy New Year ELTjam!
Thank you for a very insightful blog as we look ahead to 2017. Technology will continue to make inroads in the are of ELT but it is important to recognize that there will remain a role for human intervention and interaction in the process. We appreciate the shout out to WeSpeke and our work with natural language processing and machine learning. As you said, this work is in its early stages; we’re exciting to see how it develops. WeSpeke’s mission is to create new opportunities for people across the globe to learn English (and other languages) in order to help them accomplish their personal and professional goals. We think the NLP work is a piece of the solution. Keep up the great work!
Hi Laurie
Re Point/Prediction No. 5, which ELT publisher is this likely to be in your opinion?
My money is on Macmillan as ELT has always been peripheral for them anyway but I would like to know who you think will withdraw from ELT first. Cengage is another possibility for the same reason but at least they appaear to be growing.
Thanks and HNY!
Hi Alasdair – you won’t draw me to speculate on that, I’m afraid! Everyone’s facing the same challenges, and from very similar positions.
Just a note: ELT has been peripheral to Macmillan in the UK, but not in places with a still high ELT demand as Latin America and Spain. In fact, it’s the opposite.
Thanks – the trends are totally on point. The ‘cool’ automated experience is going to be key, along with the expectation ‘it’s online so should be free’. However, I agree with Cathy about the human element. Fundamentally, people learn languages to connect and be sociable. At Learningonline.xyz, we are excited to be playing with the newer content platforms such as Articulate Rise, H5P and AI/Bot capabilities. Yet still, our customers want to connect with a ‘human’ as part of it, to make new friends and reach out on a human level. Communicate!
Hi ELTjam team,
I’ve responded to Nick’s comments about ‘automated content creation’ in language assessment in a blog post here: https://pedagogablog.wordpress.com/2017/01/04/playing-with-fire-when-you-dont-even-know-its-hot-eltjam-duolingo-and-irresponsible-practices-in-language-assessment/
In a nutshell, if the Duolingo English Test example you’ve provided above – the cloze based on a passage from James Joyce’s ‘Ulysses’ (!!) – and my own experience of Duolingo’s Practice Test are anything to go by, it will be well beyond 2017 before ‘automated content creation’ technology gets *anywhere near* where it needs to be to create decent quality test items. That’s assuming that Duolingo really cares all that much about ‘validity’ and ‘reliability’ beyond paying lip service to such concepts. As it is though, Duolingo’s automated test item creation betrays a disgracefully cavalier attitude to test design and is reason to be pessimistic about any expanded use of such technology in 2017.
My hope for 2017 and beyond is that test designers such as those at Duolingo develop enough respect for the complexity and seriousness of assessment to fully explore the extensive literature and reflect deeply on the validity of their assessment beliefs and practices. They could start with Denny Borsboom, Glenn Fulcher, Mantz Yorke and Peter Knight, whose work would give them pause for thought. It might cause them, for example, to question whether claims that their test is “scientifically designed to provide a precise and accurate assessment of real world language ability” and “significantly correlated with the TOEFL iBT (a standardized English test)” (https://testcenter.zendesk.com/hc/en-us/articles/207815213-How-predictive-are-Duolingo-s-test-scores-) have any substance at all.
Cheers,
Kyle.
Hi Kyle – agree with everything you say about the Duolingo tests. I literally laughed out loud the first time I saw them – it’s almost unbelievable that something like this would be put forward as a serious product by a well-known brand. Our prediction is simply that we will be seeing a lot more of this kind of thing in the near future – most of it will be bad, but the quality will improve over time, perhaps quite quickly.
I also agree with you that, in the case of Duolingo specifically, it’s irresponsible to make the claims they do and charge for their tests, presumably to people who believe their ‘science stuff’ and therefore assume there is some genuine assessment expertise behind the product. It feels like a Beta product, and should be pitched as such – but I imagine they’re desperately looking for sources of revenue, and hit upon assessment as a potential goldmine if they could differentiate themselves from ETS, Cambridge etc on cost and convenience rather than quality.
The automatically generated content in their main learning app is also famously dodgy, of course, but at least that’s free…
Yes, Duolingo are still trying to figure out how to make money. They’re valued at about half a billion dollars (double the valuation of Rosetta Stone), but they don’t come close to breaking even. They’ve clearly given up on their attempt to make dollars out of crowd-sourced translation databases. They may see their test as a potential revenue-earner: they’re currently recruiting a ‘North America Partnership Manager’ to ‘drive adoption of the Duolingo English Test in the admissions process of higher education institutions’. But they’re going to struggle to compete with Pearson and others there. Development of reliable tests is extremely expensive, and that’s not what they’re putting their money.
At the moment, they’re making some money from advertising. With however many million daily users, they can sell quite a lot of advertising (and they could sell more) and they must be collecting a fair amount of data which can be sold, too. My bet is that they have come to the conclusion that it’s the sheer number of users that is their greatest asset and, therefore, the priority is to maintain and develop that user base. In that context, a cheap test gets them into the news, attracts more punters and raises the brand profile. Similarly, their recent foray into bots and their even more recent launch of Duolingo Clubs could be seen in the same light.
Sure, Duolingo want to improve the quality of their products but it doesn’t seem to be their priority (except for UX). If it were, their ‘Jobs Vacant’ board would look rather different. Language learning seems to be more of a smokescreen than anything else. With a $45 million investment from Google Capital in 2015, I don’t think we’re likely to see Duolingo getting any more ‘educational’, even though the edtech clichés will continue to roll out. Google’s #1 mantra is ‘Focus on the user and all else will follow’. With Duolingo, that ‘all else’ is likely to be advertising revenue.
But even if advertising revenue fails to generate enough … if you continue to grow your user base, the value of the company will grow, meaning that Louis Van Ahn is sitting on a huge goldmine. So, perhaps, his business strategy is simply an exit strategy?
Re prediction 1 – for 2017 at least, the state of the art of NLP will be more appropriate to power authoring productivity tools rather than spit out content directly. But moot is the point that learners may not opt for the quality, curation or scaffolding that old-school publishers hold as gospel. They won’t care whether content is NLP-powered or lovingly hand-crafted. They probably will care about how engaging it is and how productive it appears to be. Apps will be built, business models will be tested, learners will vote with their credit cards, and publishers will look at their bottom lines ever more carefully…
I agree, Paul – do most Duolingo users care much about (or even notice) the quality of the content, when the app’s fun to use and provides a clear and motivating sense of progress? I doubt it. It’s the overall experience that matters. Unlike entertainment, ELT isn’t a content industry, so is a prime target for auto-generated content. You may be right about the timeframe in which this happens, though – let’s see!
The problem, Laurie, is that no one has ever paid for either Duolingo content or the Duolingo experience, so we have no evidence to hypothesise meaningfully on how learners (as Paul says) will vote with their credit cards. And, we also have to acknowledge that the scenario discussed describes one small part of ELT – Adult private and Adult self-study: the vast majority of learners are in state institutions where consumer choice is limited and where task design (more difficult to automate well) is an integral part of the ´content´.