Abstract:
The present study empirically investigates the impact of ChatGPT’s voice mode regarding pronunciation fluency, and its effectiveness when it comes to practicing English oral skills, within the specific sociolinguistic and academic context of Algerian higher education. This research employs a quasi-experimental design with pre-test and post-test control group methodology. A number of 26 student participants was selected for this study, which was divided into two groups of 13 students in the experimental procedure: a control group that received no AI-mediated oral practice or intervention, and an experimental group who completed four sessions of ChatGPT voice interaction over a month. Both groups were assessed at pre-test and post-test using a five criterion pronunciation and fluency rubric (encompassing: speech rate, pause frequency, pause length, repairs and overall fluency) which was applied to the evaluation of their audio-recordings and oral performances during the two tests. The pre-test and post-test involved two identical speech tasks: paragraph reading task and a spontaneous image description task. The scores were measured using the Speech Rate Meter application to ensure the objectivity and precision of the measurements. The data were analyzed using SPSS Statistics software, with gain scores outcome measurements, and using an independent samples t-tests as an inferential procedure. The experimental group recorded a positive gain of +0.08 points compared to the control group which declined by -0.85 points, demonstrating a statistically significant difference (t=2.19 , p=0.038). The results indicate statistically significant improvement in pronunciation fluency
