Artificial Intelligence and the ESL Teacher
English language assessment tools have come a long way since the introduction of TOEFL, the first English proficiency test, in 1964. Back then, everything was done by pen and paper, without a computer in sight. But as technology has advanced, so has the way these tests are administered, designed, and used. While computer grading is now the norm for many companies, artificial intelligence doesn’t eliminate the importance of using human graders—especially English as a second language (ESL) teachers—in assessing the writing and speaking sections of an exam. There’s no doubt AI plays an important role in the future of English language assessment tests, but there are many advantages to using ESL professionals to judge the competency of test-takers.
Computers Can’t Detect Nuance
We communicate in a subjective world. The purpose of language is to pass information from one human being to another. Artificial intelligence is not yet advanced enough to answer complicated questions or place the response in context of what else is around it. An ESL teacher has been trained to think about the big picture and ask questions like “Is the idea expressed here complete or incomplete?” and “Does this response make sense?” when grading an exam.
Language is highly dependent on context and on the different denotations and connotations of words and word combinations. The important thing to understand with AI evaluation of language is that the evaluation does not happen directly. Instead, AI evaluation depends on ratings arrived at through algorithms that compare speaking and writing to set models, and statistical analysis is often employed to “predict” the likely proficiency of the test taker. AI evaluates writing and speech characteristics not the communicative quality and critical analysis of a response. Human beings are still needed for that.
ESL-trained graders can detect this critical thinking and complex sentence structures. An exam grader, whether human or computer, needs to understand the language at several different levels, including the appropriate meaning of words in a sentence and how they interact grammatically to convey meaning, as well as the situation and contexts in which the words are used. Human graders are much more adept at this than any computer, no matter how good the AI is.
Often times, tests that use AI are simply asking the test taker to repeat a short utterance or read directly from a displayed text. In some cases, test-takers are asked to transcribe a short listening passage. Again, statistical analysis is used to “predict” the likely proficiency level of the test taker. Many have raised concerns that this invites test-takers to craft their responses with the automated grading system in mind, incorporating elements in their speech or writing that will trigger the best rating by the algorithm. This raises larger concerns about the washback effect of AI graded testing, especially in terms of promoting language learning. Specifically, these concerns involve sending a misbegotten message to language learners that successful communication means checking off component features of formulaic speech or writing rather than focusing on the effectiveness with which a unified idea is communicated to another person. Shifting the focus to the effective and authentic communication of an idea between human beings has been a theme in the field of language learning and teaching for many decades now.
Certified Teachers Make a Difference
The typical iTEP grader is an active ESL teacher who works in a classroom in addition to grading language assessment tests for us. Our graders have years of experience working with students of different levels of English comprehension and a strong knowledge of the psychology behind language learning. In addition to their qualifications, each of our graders is trained on a grading rubric we provide and have completed an extensive iTEP Grader Certification program.
Once they begin grading exams for us, our graders are required to participate in regular norming exercises that show how their evaluations compare to both their peers and the standards we set out for them. Our graders don’t live in a vacuum—this essential and frequent training gives them the skills to recalibrate and avoid internal bias. iTEP also employs a master grader who is frequently in contact with our graders to offer feedback, answer any questions, and help eliminate human error.
An Extra Level of Care
At iTEP, we sincerely care about what we do. iTEP International was founded in 2002 by former TESOL professionals with deep roots in the international education field. The goal was to create an English proficiency test that addressed the needs of the international education community, and the members of this community are iTEP’s constant partners in the continual development and improvement of iTEP tests. The company and its offerings have their roots in the communities they serve.
We don’t have outside investors looking to make easy money through a sale or public offering. We value the importance of language learning and how transformational it can be for people to become competent in a foreign language. We hire and train educators who apply their expertise to designing exams that provide reliable results. Without reliability, an English assessment tool has little value. While AI is more reliable than humans at many tasks, evaluating the skill level of spoken and written language is not one of them, and isn’t likely to be any time soon.