Loading
Online English language testing calls for validity and reliability - two factors large language models cannot provide. Here, we look at the reasons to opt for bespoke programs built with tested and informed data for accurate results.
Educators looking for a precise and fair way to assess English language students are likely to be sizing up AI and how it’s utilised in English language testing. While large language models (LLMs) such as Chat GPT can fulfil some basic tasks, they don’t have the customised solutions to accurately measure and assess a student’s level of proficiency.
Envoy utilises a bespoke AI solution that evaluates students’ language skills with validity and reliability. It does this using real-life simulations such as open-ended questions and responses as used in everyday discussions, classroom question and answer sessions, and presentations.
While it can be tempting to use readily accessible tools like Chat GPT, these are simply not built to test a student’s English language proficiency with the same rigour as Envoy.
Envoy, by IDP Education, features the latest AI solutions combined with the expertise of linguists to provide a personalised, relevant and secure online English language testing platform. Built on what is referred to as a ‘golden set of data’, Envoy is informed by the Common European Framework of Reference for Languages (CEFR) levels and various L1 backgrounds including different English variants and accents. Language experts have rated each piece of data, which have also been tagged by six CEFR-trained raters.
Any inconsistencies in the tagging, are moderated or discarded from the ‘golden set of data’. The remaining data is used to train Envoy’s algorithms.
Powered by this high level of specification and with ongoing data updates training the AI, Envoy is an advanced English language proficiency test.
“Envoy’s test design and quality come first; we utilise customised technology and a golden data set to ensure quality and accuracy of every assessment,” said Reza Tasviri, IDP’s Head of Assessment.
Envoy uses Chat GPT in a very limited way for item generation and it is carefully monitored with human oversight. It does not use Chat GPT for rating or marking because open AI and LLMs have low reliability, high variability and lack the constructs mapping required for accurate assessment.
On the other hand, Envoy employs AI technology with detailed criteria and benchmarking to provide valid and transparent testing scores.
Envoy is simple and easy for both educators and students to use with questions based on real-time responses and a dashboard that allows tests to be set up in minutes.
Envoy is designed to support students at every stage of their English language journey. It can be used for diagnostic purposes and as a placement test for a language learning program. It can also used by teachers to check the progress of students and assess their overall proficiency at the completion of their learning.
No software downloads are required, so students and teachers can use the online tests at any time, anywhere.
The four skills tests - listening, reading, speaking and writing - can be completed in as little as 90 minutes and a detailed and accessible score report is produced within just two hours of test completion.
Envoy also integrates comprehensive anti-cheating mechanisms including copy/paste prevention, multiple voices and noise detection and face and eye-movement tracking. Any tests flagged for malpractice are reviewed by a human rater.
To learn more about Envoy or trial our test please contact us today.