Den artificiella intelligensens begynnelse i skolväsendet – ChatGPT och validitetsproblemet med summativ skrivbedömning i gymnasieskolan

Abstract

The present essay examines the apparent problem in the validity of writing assessment in the Swedish upper secondary school. The problem lies in the introduction of ChatGPT, which the students can use to write abstract texts of high quality. The main analyzing tool to examine this problem is Bachman’s (2005) assessment use argument (AUA). The empirical material comes from seven interviews of secondary school teachers in Swedish. The result is comprised by the answers from the interviews applied to the assessment use argument. The results suggests that there are two main ways of solving the validity problems in the assessment argument. The first way is that teachers can control the assessment situations. This is done by having the students write in a locked writing program, and only having them write in the classroom. This ensures that the teachers know that ChatGPT was not used to write the texts. The second way is that teachers can change the writing construct that they wish to grade. The teachers mean that in the future this is a likely consequence of the introduction of ChatGPT. The assessment would then shift focus from the product, the text, to the students writing process. These results are then discussed, and two conclusions are formed to help develop a solution to the validity problem in writing assessment presented by ChatGPT. These conclusions are as follows:

- The teachers can use a secure writing program to solve the validity problem. - The teachers can also change the writing construct and the target domain to solve the validity problem, however that would require more information on how to do that, as well as changes in the policy document of the subject of Swedish.