Optimizing the frequency of question items for bird species in quiz-style online training

Abstract

Citizen science plays an important role in monitoring for biodiversity conservation. More efficient online training is needed to help citizens improve their species-identification skills, but such training remains a challenging task. We previously developed an online birdsong-identification training tool called TORI-TORE. However, its adaptive algorithm, in which question items regarding the species with lower correct answer rates for each user were presented more frequently, was not as effective as the conventional baseline training, in which the frequency of question items was uniform among species. In the present study, we introduce new algorithms, namely, the frequency-adjustment algorithm, which prioritizes species that are expected to have a large training effect, and the interactive algorithm, which allows users to adjust the frequency of question items regarding species by themselves. We then evaluate the effectiveness of the algorithms in a randomized controlled trial, based on test scores and questionnaire responses. At the same time, we compare the new algorithms with the previous adaptive algorithm. The frequency-adjustment group showed the most improvement in scores, ahead of the interactive group, the frequency-adjustment + interactive group, and the baseline group. The frequency-adjustment algorithm also improved scores to a greater extent compared with the adaptive algorithm implemented in our previous experiment in 2021. However, participants in the frequency-adjustment group had difficulty perceiving changes in their own interest in learning birdsongs, and so a future version of TORI-TORE may need to allow for the selection of not only the frequency-adjustment algorithm but also the frequency-adjustment + interactive algorithm.

Keywords

• Citizen science
• Quiz-style online training
• Birdsong identification
• Serious game
• Learning outcomes