New Zealand Statistical Association 2024 Conference

Anna Fergusson

University of Auckland | Waipapa Taumata Rau

Lost (and found) in translation: Examining the diversity and impact of languages selected on student responses to a statistical investigation of automated language translation

This is joint work with Lars Thomsen, Anne Patel

The data used for teaching statistics is often far removed from students’ lives, limiting the ability for students to make personal connections in their responses to learning tasks. As part of a broader initiative to re-design a large introductory statistics course at the University of Auckland, new educational technologies were created to provide greater support for personalised learning. These "data landscapes" utilise APIs to access large data sets and online databases, and require students to make selections to generate the data used in assignment tasks. In one such task, nearly 2000 students were given headlines from a prominent newspaper and asked to select two non-English languages. The data landscape translated the headlines from English into the two selected languages and then back to English, with a similarity score provided for each "round trip translation". Students then used statistical methods to evaluate and compare the use of Google Translate for the two selected languages, before writing a conclusion and reflection on their findings. Analysis of the student responses to the assignment task found that a diverse range of languages were selected, leading to diversity in learning experiences and student responses. However, the impact of using data landscapes was minimal in terms of marking and scores allocated, suggesting that it is feasible to provide personalisation within large scale assessment contexts. The study also suggests that data landscape tasks could enrich the learning experience by providing an opportunity for students to make connections to the data.