Search

Artificial intelligence (AI) is increasingly positioned as a design partner in materials optimization, enabling accelerated exploration of vast composition–processing–structure spaces under multiple, often conflicting, targets. Yet sustainability-centered materials design is not simply a larger version of multi-property optimization: it requires negotiating trade-offs across heterogeneous objective types such as performance, cost, safety, emissions, toxicity, circularity, and resource criticality, while accounting for lifecycle shifts and stakeholder-dependent priorities. Many current AI-enabled optimization workflows implicitly treat trade-offs as static Pareto-front problems with stable objective meanings and fixed feasibility boundaries. This conceptual manuscript argues that such assumptions are structurally incompatible with sustainable materials decisions, which involve trade-offs that are contextual, value-weighted, and regime-dependent. We introduce a novel theoretical framework—Trade-Off Sensitivity Theory (TOST)—which models sustainability optimization as a decision process governed by objective incompatibility geometry, lifecycle constraint migration, uncertainty-to-consequence coupling, and preference volatility. Rather than proposing algorithms or empirical evaluation, TOST provides a theoretical map linking Pareto efficiency to sustainability legitimacy through three layers: objective semantics, trade-off sensitivity, and action admissibility. The framework clarifies when AI outputs support responsible selection, when optimization is ill-posed, and how sustainable decisions can be justified under conflicting criteria.

The discovery of next-generation materials remains a slow, iterative, and resource-intensive process. Conventional approaches rely on sequential cycles of hypothesis generation, synthesis, characterization, and interpretation. Although this process has produced transformative materials, its pace is increasingly misaligned with urgent technological needs in energy, sustainability, electronics, and advanced manufacturing. Recent advances in artificial intelligence, robotics, high-throughput experimentation, and computational physics have created new opportunities to accelerate materials discovery. Self-driving laboratories and closed-loop experimentation systems can now propose experiments, execute them, learn from results, and refine subsequent decisions. These developments suggest the emergence of autonomous materials intelligence as a new paradigm for scientific discovery. However, current approaches often treat artificial intelligence, physics-based simulation, and human expertise as separate instruments rather than as mutually reinforcing partners. AI models may generate predictions without sufficient physical grounding, simulations may remain disconnected from experimental feedback, and human judgment may enter only after automated decisions have already been made. This fragmentation limits the development of truly autonomous and scientifically trustworthy materials discovery systems. This conceptual framework article develops a Human–AI–Physics framework for autonomous materials intelligence. The framework positions human expertise, AI algorithms, and physics-based models as co-equal pillars in a self-driving discovery pipeline. It explains how these pillars interact across discovery, optimization, and validation cycles. The article synthesizes 26 peer-reviewed publications published between 2017 and 2024 across autonomous experimentation, materials informatics, active learning, generative models, graph neural networks, physics-informed machine learning, and self-driving laboratories. The synthesis is not presented as a review or meta-analysis. Instead, it is used to construct a systems-oriented conceptual architecture for integrating human judgment, machine learning, and physical laws. The proposed framework defines autonomous materials intelligence as an iterative workflow in which AI proposes, physics constrains, humans guide, and experiments validate. By linking these functions into a closed-loop system, the framework offers a blueprint for discovering, optimizing, and validating next-generation materials with greater speed, interpretability, and scientific rigor.