AI Experiment

The AI Explainability Audit service addresses critical needs in agriculture by making AI more transparent and trustworthy. Farmers and agricultural organisations often hesitate to fully embrace AI due to its "black box" nature, lacking understanding of how AI tools reach their recommendations. This service bridges this gap by providing independent assessments of AI explainability, focusing on the clarity and usability of AI explanations for non-technical users. This empowers stakeholders to make informed decisions, ensure ethical and sustainable practices, and meet regulatory requirements, ultimately fostering confidence in AI-driven agriculture. Before & after the service: Before Farmers and agricultural organisations may be hesitant to adopt AI due to a lack of understanding and trust in "black box" systems, potentially leading to suboptimal decisions and concerns about ethical implications. After Stakeholders gain clear, understandable explanations from their AI tools, enabling informed decision-making, ensuring alignment with ethical and sustainable practices, and facilitating compliance with emerging regulations, fostering trust and confidence in AI-driven agriculture.

Logistics The assessment can be a one-time engagement or repeated periodically (e.g., annually, after model updates) to ensure an ongoing check for transparency and compliance. Delivery Period The service can be performed throughout the year. However, the timing of the assessment might be strategically aligned with the AI provider's product development cycle or marketing campaigns. Duration The service duration depends on the complexity of the AI system and the scope of the evaluation. A typical assessment might take between 2 and 4 weeks. Location The service can be performed remotely or on-site, depending on preference. On-site visits may be needed for observing AI systems or meetings. No strict location requirement exists, but travel costs may apply. Customer Requirements The customer must provide access to the following: (1) Documentation describing the AI system's functionality, architecture, and the types of explanations it generates; (2) Examples of AI explanations for various scenarios; (3) Information about the target users of the AI system (e.g., farmers, agricultural consultants); and (4 - Optionally) Access to a test environment or demo of the AI system. Deliverables The customer will receive a comprehensive audit report, including: An assessment of the quality, accuracy, and consistency of the AI's explanations. An evaluation of the usability and understandability of the explanations for the target audience. Identification of any gaps, biases, or potential ethical concerns related to the AI's explanations. Recommendations for improving the explainability of the AI system.

The service is customisable to meet the customer’s needs. It offers options such as tailoring the explainability assessment to specific agricultural domains (e.g., precision irrigation, livestock management) or AI application types, focusing on particular explanation types (e.g., feature importance, rule-based), adjusting the scope of the usability testing with target users, and customising the format and level of detail in the final audit report.