human computer collaborative learning in citizen science project full details
Timeline:01 Nov 2019 - 31 Oct 2022
Human-Computer Collaborative Learning in Citizen Science
This project explores the potential for collaborative learning between humans and machines within the framework of environmental citizen science.
This project explores the potential for collaborative learning between humans and machines within the framework of environmental citizen science. The term `citizen science' encompasses public participation in science and scientific communication to the public. Although not new, citizen science has gained renewed attention because of the opportunities arising from citizens' access to digital technologies in terms of data collection and annotation. While the vast majority of citizen science projects are aimed at data gathering, we instead propose a transformational shift to a new citizen science in which the public and technology are regarded not just as sensors or data recorders, but as a collective and empowered human--artificial intelligence that can help each other in science learning.
We will focus on the task of species identification from images. Citizen science projects such as iSpot invite the public to submit photos of wildlife. These are identified to species level and verified before being contributed to science. We will explore artificial intelligence as a means to automatically identify species in images. While this can save human effort, we are concerned about impact this might have on nature lovers. The introduction of technology is often associated with concerns of de-skilling. For naturalists, the honing of species identification skills is a key motivator of the recording activity. Hence, designing technology that provides opportunities for learning for both citizens and machines is essential, as is co-creating the technology to ensure that it is not only user friendly but responds to their motivations. Our approach will involve citizens collaborating with AI to arrive at
a species identification. AI will narrow down the choices and inform the citizen about how to distinguish the options. The citizen in turn will through providing an identification help the machine in its learning. We will study this learning interplay with respect to collaborative species identification, but will also explore technologies that foster wider science learning, environmental consciousness and data literacy through better communication of complex citizen science data. For this we will develop technology for Natural Language Generation that can communicate complex data through language.
Our proposed work programme seeks to bring about quantifiable benefits to (a) science, e.g., through the production of new knowledge and through monitoring key scientific processes at challenging temporal-spatial scales; (b) diverse stakeholders including the citizens themselves, e.g., through meaningful science learning for sustainability in formal and informal education contexts; and (c) wider society, e.g., through better societal understanding of current sustainability issues, leading to individual and societal action in support of the environment.
- Imperial College London
- University of Aberdeen
- Learning through Landscapes
- St Alban's CoE Primary School