Opportunities and Risks for Citizen Science in the Age of Artificial Intelligence

Abstract

Introduction Technologies that allow computers and machines to perform tasks normally requiring human intelligence are often referred to as artificial intelligence (AI). These technologies allow machines to complete tasks with traits or capabilities ordinarily associated with human cognition, such as reasoning, problem solving, common-sense knowledge management, planning, learning, translation, perception, vision, speech recognition, and social intelligence (Kaplan and Haenlein 2019). Research in AI is rapidly increasing, as indicated when c omparing the annual publishing rate of papers focused on AI between 1996 and 2017 against the publishing rates of papers focused on any topic or against the publishing rates of papers in the field of computer science (see the growth of annually published papers by topic in Shoham et al. [2018; p. 9]). This growth in AI publications has prompted researchers to critically explore the potential promises and risks of AI (Scherer 2016; Webb 2019; Yudkowsky 2008) as well as ethics and responsibilities (Miller 2019; Cowls and Floridi 2018; Scherer 2016; Dawson et al. 2019). AI has been used in citizen science projects for about 20 years. It was first used in this context in 2000, in collaborative AI databases such as the Generic Artificial Consciousness (GAC)/Mindpixel Digital Mind Modeling Project (McKinstry 2009) and the Open Mind Common Sense project (Singh et al. 2002). In these models, usersubmitted propositions were meant to create a database of common-sense knowledge that could function as a kind of digital brain. This relationship between collective knowledge and algorithmic processing evolved in many directions and, in 2019, is predominantly represented by machine learning, especially applied to computer vision, which includes diverse methods of automatically identifying objects from digital photographs. For example, the iNaturalist platform, a citizen science project and online social network, is designed to enable citizen scientists and ecologists alike to upload observations from the natural world, such as images of animals and plants (Van Horn et al. 2018). The platform is one among many (Wäldchen et al. 2018) that include an automated

Community science participants gain environmental awareness and contribute high quality data but improvements are needed: insights from Bumble Bee Watch

“…The Evolution MegaLab project gave immediate automated feedback to users upon record submission, including text and figures comparing their submission to historic records in the same area, which also helped to motivate their participants (Worthington et al, 2012). Several community science programs, such as iNaturalist, Pl@ntNet, and Merlin Bird ID, use an artificial intelligence and deep learning computer programs to suggest identifications based on comparison to other photos in their database and nearby records , which provides instant feedback to users (Wäldchen & Mäder, 2018;Ceccaroni et al, 2019). The use of new technologies, including new software programs, will help to remedy some currently existing challenges in both Bumble Bee Watch and community science programs in general to improve data collection and the user experience going forward (Newman et al, 2012;Wäldchen & Mäder, 2018;Ceccaroni et al, 2019).…”

Machine Learning in Citizen Science: Promises and Implications

“…Deep learning can help solve certain types of difficult computer problems, most notably in computer vision/computer hearing and natural language processing (NLP). Computer vision or hearing defines a subset of AI which automatically extracts information from image, video, and audio data using algorithms (see Ceccaroni et al 2019). The 'deep' in deep learning refers to the many layers that are built into a model, which are typically neural networks.…”

“…Most projects in our list involve supervised learning by using image recognition software in the realm of computer vision. Computer vision is used on citizen science data and camera trap data to assist or replace citizen scientists in fine-grain image classification for taxon/species detection and identification (plant or animal) (Ceccaroni et al 2019). A good example is the project Wildlife Insights (Ahumada et al 2020) covering images from camera trap databases.…”

“…The use of ML continues to grow, but scholarly reflection and discussion on the role of ML in citizen science are still in their infancy. In other words, a solid research overview on this topic is complicated by the fact that not only is citizen science research not settled science, as Ceccaroni et al (2019) argue in their review essay, but also by the fact that 'AI is not settled science either; it inherently belongs to the frontier, not to the textbook' (p. 8).…”

“…In scientific research, they find many applications in different fields such as biology, astronomy, and social sciences, just to mention a few (Jordan and Mitchell 2015). Although AI is not new to citizen science (Ceccaroni et al 2019), the convergence of advanced computing, availability of data, and learning algorithms can introduce something dramatically new in this area. The opportunities are many, and in some cases not yet foreseen, but so are the challenges, including the need to advance the explainability, accountability, and fairness of algorithms from the perspective of ML research and from that of citizen scientists using the applications.…”