I know there is plenty of skepticism among posters here with respect to language testing. I also think it is difficult to put a single number on something so multivariate.
About a year ago, when I wanted to start actually speaking Spanish and start lessons, curiosity got the best of me and I bought a language assessment on italki. The language test they use is called “Emmersion”, purportedly adaptive AI based and spits out scores and relates them to various other language assessment scales, including CEFR. It is apparently used by some universities and companies for their pre-employment/admission language assessments.
The majority of the test is listening and repeating audio sentences of increasing length (Elicited Imitation). Then there are very few traditional language test questions (fill in the blank, multiple choice, and answering questions about passages you read), and finally long form spoken answers to open questions (talk about a time when…). Since it is adaptive (i.e. questions vary based on how you do), I’m not sure how much this varies.
That was it. It’s a really short test, and the results were instant. It gave me a score of 5.7 for speaking which they say correlates to B1 on the CEFR, and 561 for grammar (B2 CEFR) for an overall B2, which correlated with my self assessment and those of my italki teachers. I remember feeling like quite the dummy because it was such a challenge to repeat the sentences as they got longer. I wasn’t sure I could even repeat English sentences that long. I felt even more skeptical it could assess anything with so little information.
It turned out the package I bought came with 4 tests. I’m pretty confident in my Japanese, so I decided to test the test and took one in Japanese. 9.4, C1-C2 (it is speaking only). Not a bad result, but I still felt like a dummy after because it was still quite challenging to remember and repeat those longer sentences.
Well, I’m about a year on from that point and I still had the two tests left as of this morning, but they are going to expire soon. I decided to take the Spanish one again today. It intrigued me, because while I have been very consistent with reading and listening, I’ve not been consistent with italki lessons, and I don’t really “feel” like I’ve improved much. (Intermediate plateau? ). The result was 6.6 / B1-B2 in speaking and a grammar of 911/C1.
I still felt every bit the dummy after, maybe more that last time. I’m surprised the result was better based on how I felt. The grammar score seems like it’s over rating me, but it is interesting to see that it assessed an improvement in both, and that it recorded more improvement in the area I spent more time on (written/grammar over speaking production).
Have any of you taken these assessments? What was your experience? Will AI provide us with more capability in assessing language ability in the future?
Anyway, I’m certainly not taking these results too seriously just yet. I am curious about the topics of language assessment and AI. I am tempted to use the last one to try English (my native language) and see how it goes. I fear I’ll get a dose of humble pie in that one, too.