Abstract

The rapid rise of citizen science is seen as a solution to the mismatch between data demand and supply whilst simultaneously engaging citizens with scientific topics. In the environmental domain for example, lay people form often extensive biodiversity sensor networks. However, such recording schemes require careful consideration of how to motivate, train and retain volunteers, especially where schemes concern species groups for which knowledge levels in society are low. The typical lack of resources within citizen science programmes for engaging with volunteers thus presents a substantial bottleneck.

In this context we evaluated the role of automated textual feedback aimed at improving volunteer’s species identification skills and enhancing volunteer experience and retention. The NLG component used a species identification key and data collected from within a focal citizen science program to contextualize a submitted record, explain reasons for any misidentification and highlight key features that facilitate correct identification.

Experimental trials showed increased species identification accuracy and enhanced retention of participants who were provided with such in-depth automated feedback, compared to those who only received the correct identification without explanation. This feedback component is now incorporated into BeeWatch, a citizen science program focused on bumblebee species in the UK.