The influence of AI-generated speech characteristics on knowledge acquisition
Workgroup | Realistic Depictions |
Duration | 10/2020 – 09/2023 |
Funding | Sondertatbestand Data Science |
Project description
Artificial intelligence-based applications, especially deep learning techniques, allow the manipulation of visual and auditory information. In the context of knowledge acquisition, these technological developments offer the potential to personalize learning materials and optimize their utility. Aim of the project is to investigate the extent to which aligning the tutor with the learner influences the learning process.
Early studies on model learning examined which characteristics of a model influence the learning process. It was shown that a similarity between the model and the learner can promote knowledge acquisition. This result can be explained by the assumption that learners find it easier to identify with a similar model, which in turn facilitates the imitation of the observed behavior. Pedagogical agents used in interactive learning environments have not yet been able to show clear positive effects on knowledge acquisition when similar to the learner. However, only global aspects, such as age and gender, were manipulated. Research, on the other hand, that looks at the imitation of the counterpart in social interactions has shown that verbal imitation can influence the actions, thoughts, and feelings of the imitated. This verbal imitation includes the use of similar words and sentence structures, as well as the assimilation of speech melody, speaking rate, and pitch. This project investigates to what extent pedagogical agents that use such verbal mimicry have a positive effect on knowledge acquisition. For this purpose, state-of-the-art Deep Learning methods are used to analyze speech features of learners and transfer them to virtual tutors.