Developing and Integrating Trust Modeling into Multi-Objective Reinforcement Learning for Intelligent Agricultural Management

TL;DR

The paper addresses the challenge of deploying AI-guided fertilization in agriculture while aligning with farmers’ real-world practices by embedding a trust model into reinforcement learning. It frames the decision problem as a POMDP, solves it with a recurrent DQN, and extends to Multi-Objective Reinforcement Learning (MORL) to balance agronomic performance, environmental impact, and farmer trust, with trust quantified via Mayer’s three-dimensional framework: $Trust\_Score = Ability \times Benevolence \times Integrity$. A farmer survey informs the trust model, which is integrated into the RL training loop to produce trust-aware fertilizer recommendations; experiments use Gym-DSSAT simulations under multiple scenarios and climate variability to assess trade-offs and robustness. The results show that trust-aware policies can achieve similar or better net income with substantially higher trust and reduced environmental impact, though climate extremes reveal limitations in trust adaptability, motivating future expansion of survey data and dynamic trust recalibration for resilience. Overall, the work advances trust-aware AI adoption in precision agriculture and provides a blueprint for human-centered RL in other domains where stakeholder trust is critical.

Abstract

Precision agriculture, enhanced by artificial intelligence (AI), offers promising tools such as remote sensing, intelligent irrigation, fertilization management, and crop simulation to improve agricultural efficiency and sustainability. Reinforcement learning (RL), in particular, has outperformed traditional methods in optimizing yields and resource management. However, widespread AI adoption is limited by gaps between algorithmic recommendations and farmers' practical experience, local knowledge, and traditional practices. To address this, our study emphasizes Human-AI Interaction (HAII), focusing on transparency, usability, and trust in RL-based farm management. We employ a well-established trust framework - comprising ability, benevolence, and integrity - to develop a novel mathematical model quantifying farmers' confidence in AI-based fertilization strategies. Surveys conducted with farmers for this research reveal critical misalignments, which are integrated into our trust model and incorporated into a multi-objective RL framework. Unlike prior methods, our approach embeds trust directly into policy optimization, ensuring AI recommendations are technically robust, economically feasible, context-aware, and socially acceptable. By aligning technical performance with human-centered trust, this research supports broader AI adoption in agriculture.

Figures (5)

• Figure 1: The interaction between an RL agent and the agricultural environment.
• Figure 2: Average monthly temperature and total precipitation during the corn growth period in Ames, Iowa, in 1999.
• Figure 3: Five fertilization recommendations generated under different policy scenarios.
• Figure 4: Pareto fronts under scenarios of of temperature increase. The circled point represents the selected policy.
• Figure 5: Pareto fronts under scenarios of precipitation reduction. The circled point represents the selected policy.