Elgazette Logo newtrans  The magazine for English language teaching and English medium education

Iatfel: 2018: Rise of the machines?

clouds 1845517WEB

As new players enter the market, technology is becoming a vital part of language testing – but the human element is still vital, Dr Michael Milanovic tells the El Gazette

For a while, ELT testing seemed to be increasingly dominated by two or three mega exam providers. But we are suddenly seeing new players, like LanguageCert, entering the market. Language school chains are also marketing exams to government and Silicon Valley is promising cheap online alternatives. How do you see this playing out?

The idea of global dominance in language testing is perhaps a bit dramatic. While international exam providers play a key role, it’s always been the case that far more learners take locally produced tests than international ones.

Clearly, for some purposes, such as study or migration, there is a restricted range of tests. The increasing interest in language testing, though, is being driven by the importance people around the world place on language knowledge in general and English in particular.

As you suggest, there are numerous new entrants in the language assessment market, one of these being PeopleCert, the group behind LanguageCert, which has now turned its attention to language provision in English and a range of other languages.

Testing is like any other industry – traditional market leaders are always likely to be challenged by new entrants. I see this as a healthy situation which increases choice and drives up quality.

Technology facilitates the emergence of new entrants. It provides a great way of spreading the word about new products as well as supporting their delivery and processing.

But technology alone doesn’t deliver a full understanding of customer needs. Being able to deliver what people actually need is essential to success.
I think that the number of organisations interested in developing language assessment is likely to grow. Those that focus on quality products and quality service are most likely to do well.

Where do you see the marriage of technology and language testing going next? Will we see the day when computers not only mark the exams but set them?

Digital technology has been used in rating language tests for many years, although mainly with regard to reading and listening skills, where machine marking is ideal for items like multiple choice questions. More recent advances in computational linguistics and machine learning are making this type of technology in scoring the performance of productive skills a reality.

Making it work on a large-scale remains challenging, however.

I think it’s also fair to say that many educationalists view ‘black box’ marking engines of speaking and writing with suspicion. They often feel that the features being evaluated are either inadequately described by the provider or seem divorced from the features of real-world communication.

Technology already impacts significantly on all aspects of large-scale language testing, for example to support strong customer service, to store/retrieve/combine test material stored in item banks, and to ensure speedy and accurate marking.

Does that mean that computers ‘set’ the exams?

They are clearly an important part of the process. Computers might help select suitable authentic materials, generate options for multiple choice questions or even help to make computer-based speaking tests more interactive.

But for the moment, using computer technology alone to write other than the most simple and formulaic test items is a long way off, as anyone has tried to write test items will appreciate. There’s a useful analogy here with technology and teaching. The most effective use of digital has been to support more traditional teaching rather than replace it.

Some people are saying we’ve seen peak English and the move is towards multilingualism. How far do you think that is true?

I have long said that a knowledge of English has become necessary but not sufficient. Learning English and other languages are not mutually exclusive, and I think interest in both English and multilingualism will continue to grow, along with demand for high-quality tests.

There are believed to be 1.5 billion English speakers worldwide.

However, we shouldn’t forget that there are, for example, 400 million Spanish speakers and 21 nation states where Spanish is an official language.

Learners may want and need English and English tests, but the same is true for other languages.

For example, LanguageCert so far offers tests in Spanish and Turkish as well as English.

The EU’s Barcelona objective, which explicitly promoted the learning of ‘mother tongue plus two’ foreign languages, is a good example of that principle of multilingualism in action.

What do you think is the biggest challenge facing language testing and what is the biggest opportunity?

Firstly, there’s a need to promote assessment literacy. A need to give educational professionals as well as parents a better understanding of how testing works.

They can then operate as informed consumers when it comes to what features to look for. Secondly, maintaining test security and integrity is an increasing challenge and technology is increasingly used to try to cheat.

On the other hand, it can also be used to deter, prevent and detect malpractice, for example by using CCTV surveillance of test-takers or running software to detect improbable patterns of test statistics.

Going forward, I think it’ll only be the better-resourced organisations that are likely to be in a position to take the steps necessary to protect and maintain public confidence in their tests. Technology also remains the biggest opportunity for language testing. As ever, its use needs to be managed with great care.

Dr Milanovic is chair of the Advisory Council at Peoplecert, one of a number of roles he has taken on since stepping down as chief executive of Cambridge English Language Assessment. He is also a visting professor at the University of Bedfordshire.