A primer on artificial intelligence in plant digital phenomics: embarking on the data to insights journey. (February 2023)

• Record Type: Journal Article
• Authors: Harfouche, Antoine L.; Nakhle, Farid; Harfouche, Antoine H.; Sardella, Orlando G.; Dart, Eli; Jacobson, Daniel
• Is Part Of: Trends in plant science. Volume 28:Number 2(2023)
• Page Start: 154
• Page End: 184
• Publication Date: 2023-02
• Subjects: AI system architecture -- black box models -- data analytics -- digital phenomics -- explainable artificial intelligence -- interpretable by design models; Botany -- Periodicals; Botanique -- Périodiques; Botany; Periodicals; 580.5
• DOI: 10.1016/j.tplants.2022.08.021
• Languages: English
• ISSNs: 1360-1385

Abstract

Abstract : Artificial intelligence (AI) has emerged as a fundamental component of global agricultural research that is poised to impact on many aspects of plant science. In digital phenomics, AI is capable of learning intricate structure and patterns in large datasets. We provide a perspective and primer on AI applications to phenome research. We propose a novel human-centric explainable AI (X-AI) system architecture consisting of data architecture, technology infrastructure, and AI architecture design. We clarify the difference between post hoc models and 'interpretable by design' models. We include guidance for effectively using an interpretable by design model in phenomic analysis. We also provide directions to sources of tools and resources for making data analytics increasingly accessible. This primer is accompanied by an interactive online tutorial. Highlights: Artificial intelligence (AI) presents unprecedented opportunities for computational data analytics in plant digital phenomics and will have an extraordinary impact on plant science in the coming decades. In this review we describe a human-centric explainable AI (X-AI) system architecture that is based on advancing novel and creative linkages between data science, plant science, and information systems that are oriented to multiple outcomes. We elucidate the mechanisms of X-AI by illustrating the difference between inherent explainability and post hoc explainability, and dispel misunderstandings that can dilute the importance of this crucial but understudied topic. We create an interactive online tutorial to train, for the first time, an interpretable by design model that classifies cassava diseases and explains its own reasoning for each prediction.