Citizen Science AI
Artificial intelligence for Citizen Science is an emerging field to apply machine learning (ML) to various data-driven operations in public engagement apps. With AI machine learning, data and media uploads like photos can be quickly analyzed, and an output can be presented directly in a Citizen Science App. A common use-case for AI in Citizen Science is the use of image recognition to e.g., identify species in a biodiversity program and output a list of probabilities of which species has been photographed by a participant in the mobile app.
AI in general is an umbrella term, which can mean a wide range of applications of artificial intelligence. While many AI applications like generative AI and large language models (LLM) have a multitude of ethical and sustainability issues, tailored machine learning or deep learning can be a valuable asset in science and data analysis.
On SPOTTERON, we have developed a neural network framework that forms a flexible basis for project-specific AI machine learning integrations. With the SPOTTERON AI system, Citizen Science projects from all kinds of fields can apply image recognition in their interactive mobile apps.
From User Upload to Identification
The SPOTTERON AI seamlessly integrates into the Citizen Science App. When uploading a new observation, participants can send the provided photograph to the AI with just a click/touch of a button. The AI system processes the uploaded image and outputs the results. The interface of the AI tool presents the user with a list that shows the classification and also their probability. The user can now choose the best fit, and the selection is taken over automatically as data for the new submission.
To support the user's decision, not only are the probabilities displayed, but also every outcome is linked to an "Info-Panel". On the right side, a thumbnail of, e.g. the AI-identified species or object is linked to a glossary page with further information. By that, a user can simply take a closer look at the result before submitting to improve data quality and even learn more about the species or object with just a touch.
Further options in the SPOTTERON AI interface provide a manual selection to cover an outcome if the results do not match the photo and a correction is necessary. The SPOTTERON AI toolkit can be integrated into every app, running on the platform. The development of the framework has already been completed, which saves costs and makes the Citizen Science AI toolkit a cost-efficient and best-value option for Citizen Science, Community Science, and Public Engagement projects.
From defining the goals to a live AI
Due to the flexibility in the design of the SPOTTERON AI system, it is possible to train and apply AI models for nearly every purpose. From classic species identification to object recognition, even extended possibilities like occurrence counting, written text transcription, or other forms of data output can be effectively integrated. Our expert team can support your project from the start with planning the AI's scope and goals to AI training and model deployment. Furthermore, we can help with the training data and provide reports in the training process about which objects of species require more data for training.
Ethical AI made in Europe
When working with AI, it is essential to consider and apply strong ethical principles to mitigate negative impacts. On SPOTTERON, digital authority and safety without abusing the participant's privacy are always a core goal of all our work. In terms of AI, we do not build on disruptive third-party large language model services but train and run our custom citizen science AI integrations in the European IT landscape. We have developed the SPOTTERON AI base modular framework in the course of our participation in the Horizon 2020-funded "IPM Popillia" app development, and are extending the technology further as a partner in other Horizon Europe funded projects.
