Speaker
Description
The use of chatbots in higher education has gained significant popularity, and research shows that the number of students consulting chatbots is increasing rapidly. It is well known that chatbots can be a valuable learning aids, however, they also rise ethical concerns. Therefore, it is very important is to understand how accurate these chatbots are [1].
The aim of the current study was to investigate how well one of the most popular chatbots, ChatGPT, responds to practical physics questions and solves different physics problems in different languages.
We compiled about 500 multiple-choice questions and physics problems in four languages: Latvian, English, Kazakh and Russian in order to analyse the correctness of answers. The results showed relatively good performance in English but weaker performance in Kazakh and Latvian. Since ChatGPT is trained mainly on English language data, it is less proficient in other languages. Thus, there is still a language barrier and some important information may “get lost in translation”
