Defining Quality Requirements for a Trustworthy AI Wildflower Monitoring Platform

Quality goes far beyond model accuracy: The paper demonstrates that a production-ready AI system requires consideration of many factors beyond machine learning performance metrics. For wildflower monitoring, this means evaluating data quality, software reliability, and system robustness—not just how accurately the model identifies flowers.

ISO standards can be adapted for AI systems: The research applies traditional software quality frameworks (ISO25000) to AI systems, showing that established quality standards can be structured to address AI-specific concerns. This provides a practical pathway for organizations to define and measure trustworthiness in AI applications.

Quality requirements must address the full lifecycle: The paper examines three realistic scenarios—using, extending, and improving the wildflower platform—revealing that quality needs change as AI systems evolve. This suggests that trustworthy AI requires ongoing attention to quality, not just at deployment.

A structured approach helps define trustworthiness: By using a quality model as a "structured dictionary," practitioners can systematically define requirements across data, model, and software components. This moves beyond vague notions of "trustworthy AI" to concrete, measurable quality criteria.

Implementation guidance remains incomplete: The authors note that future work is needed to develop specific metrics, tools, and best practices. This highlights that while frameworks for trustworthy AI are emerging, practical guidance for engineers to actually implement these standards is still developing.