AI does it again: Adaptive testing is changing the way we measure language proficiency

However, while some data is easily quantifiable – such as demographics – other types of data points – like language proficiency – are notoriously difficult to pin down. For a start, each language functions differently and needs its own form of measurement. On top of this, the understanding of ‘fluency’ differs from place to place, and from context to context. In short, accurately measuring someone’s language level is no easy task. 

So, why does this matter to businesses, and what’s it got to do with AI?  

Our ability to communicate – particularly in English – can determine more than our travel plans. With 85% of international companies using English as one of their working languages, and nearly three-quarters of employees with low English proficiency facing limited participation at work, language level can have a direct impact on career development and inclusion.   

Proving your abilities in English in this landscape is therefore vital. Without certification, English skills may remain hidden behind a lack of confidence or biased assumptions. This is where accurate, non-biased language assessments play a critical role.

Until now, measuring language proficiency has been the job of human teachers or a machine asking ‘yes/no’ questions. Spoken language, in particular, has always required a human to make a call on the proficiency level of the speaker. Naturally, this method leaves a lot to chance depending on who your teacher is.  

With AI now a part of language assessments, English speakers can test all four language skills – reading, writing, speaking, and listening – without fear of human error or bias. With a rich data set of international speakers to draw from, AI minimizes the possibility of subjective analysis. The AI simply adapts the test to accommodate the measured level of the speaker and assesses based on accuracy and comprehensibility, including in the nuanced spoken word.  

The value of this cannot be overstated. Not only does this mean that certification is in reach for anyone with an internet connection, increasing accessibility for those without a teacher, but it also delivers accuracy and consistency faster than a human could possibly achieve. With a certificate granted within minutes of completing the test, speakers can prove their competency in a fraction of the time it would usually take. 

What does this mean for larger organizations and governments? 

Language proficiency can tell you a lot about your people. By mapping demographic data to accurate AI-powered English proficiency data, leaders can gather a heightened awareness of the trends that shape their people. For example, on a global scale, our English testing has revealed that there is a growing English proficiency gap between men and women, with women experiencing a significant decline in recent years. This data is invaluable for understanding how access to education or changes in culture are impacting practical skills and opportunities. 

With AI in the mix, your business can gather such data in real time and at scale with ease. Understanding how English is impacting your teams is no small feat; it has the power to uncover the communication and inclusion needs in your organization, and, crucially, to solve them. 

Certifiable testing, though driven by AI, still requires a formalized structure to be recognized and valuable. However, in the virtual classroom (such as the EF Hyperclass), already we are seeing how AI can gather data in the informal setting of a live class and transmit that data in real time. With unbelievable insight into how much a student speaks, their vocabulary count, and overall accuracy, a version of ongoing ‘AI assessments’ is already happening in online language learning.  

Both of these testing techniques have value for the learner, teacher, and training manager in organizations. But for ultimate consistency and accuracy at scale, the AI-driven four-skill test is unbeatable.