Many students are concerned about the potential large-scale change of PTE questions while few others unfortunately faced unexplained sudden decline in their scores. Like mysteries, these are the questions which nobody can give a definitely answer. However, the mystery unwittingly points toward the obvious feature of PTE – Machine Learning, a form of Artificial Intelligence (AI).
What is Machine Learning? In the simplest form, it is when computers apply statistical learning technique to automatically identify patterns in data. These techniques then can be used to make highly accurate predictions.
Now, you probably wonder why is Machine Learning related to PTE? Well, unlike other English tests, PTE is a computer based English test and its definitive feature is the Automated Scoring System. This scoring system is very consistent, very fast (excepted for those occasional hiccups when your score isn’t available for weeks) and arguably superior than conventional methods due to its complex algorithms and most importantly, its all-powerful Machine Learning behind the scene that drives the entire process. In other words, by taking Machine Learning into consideration we could rationalize our findings and debunk the mystery.
OK, so what does Machine Learning have to do with PTE questions?
PTE score guide uses Item Response Theory (IRT), in layman’s term – for a question that is difficult then it should carry higher score weight. What this means is that item difficulties are in place for each PTE question and item difficulty level is assigned dynamically by the system via Machine Learning. This explains in some cases when some test takers think they screwed up their exam because it’s very challenging yet their score report showed improvement, and vice versa.
What it means to the overall experience for test takers?
It can be beneficial, the reason is very straightforward. Every new question introduced in PTE exam is a question that starts with minimal amount of statistical data. Pearson constantly collects new data to train its AI to improve its scoring system. Its ability at assessing responses for each new question improves as more and more data are being collected for the said question. In short, the bigger sample there is, the more accurate computer is at scoring. But, at the same time, releasing new questions pose risks because it can jeopardize the consistency of the scoring system due to low amount of statistical data.
A logical approach would be releasing new questions slowly to reduce the “damage”. By adopting this approach, it implies that those questions you have seen around during your exam will stay there for a long while (minimum 1 year), good news for most test takers!
To sum up, the release of new questions is most likely in small increments as it takes time to get enough data for Machine Learning to be effective. If your score suffered a sudden plunge in any particular section, maybe it’s because the Machine Learning is still “learning”. And for those who are undecided on switching to PTE because they are worried about the question pool or format change? Relax, it’s NEVER BEEN BETTER TIME to make that switch.
To obtain more information about PTE exam preparation, please check out PTE Study Package.