Search

In the rapidly expanding domain of artificial intelligence applied to materials science, default assumptions embedded within machine learning pipelines—ranging from software library choices and architectural presets to data preprocessing routines and evaluation protocols—are routinely treated as neutral, inconsequential background elements that require no explicit justification. Yet these defaults operate as hidden parameters, subtly yet powerfully constraining the hypothesis space, directing optimization trajectories, and ultimately shaping the predictive behavior of models in ways that rival or even exceed the influence of explicitly tuned parameters, as theoretical analyses of deep networks have long emphasized. This paper advances the theoretical claim that default assumptions in materials AI function as implicit priors, encoding unacknowledged inductive biases that propagate through every stage of a pipeline and determine what counts as a valid or reliable prediction about material properties. Building directly on foundational examinations of inductive bias, we distinguish defaults from both explicit parameters and tunable hyperparameters, develop a taxonomy of four primary default types specific to materials informatics, and derive corollaries concerning the epistemic consequences of unexamined defaults for model comparison, reproducibility, and knowledge transfer. We further examine why such defaults persist—owing to cognitive convenience, historical path dependence, and systematic attribution errors—and clarify their subtle yet critical relation to formal Bayesian priors, while noting that understanding deep learning requires rethinking generalization when defaults remain hidden. The analysis culminates in concrete implications for practice, proposing that defaults must be elevated to first-class objects of documentation, justification, and sensitivity analysis if materials AI is to achieve genuine epistemic transparency and scientific robustness. By theorizing defaults as hidden parameters, this work identifies an overlooked dimension of model epistemology in materials science and offers a conceptual framework for making the invisible visible.

Thermal transport, specifically lattice thermal conductivity, in disordered materials such as glasses, high-entropy alloys, and polymers is critical for applications ranging from thermoelectric energy conversion and thermal barrier coatings to electronic cooling and flexible electronics. The structural disorder inherent in these systems breaks translational symmetry, generates heterogeneous local atomic environments, and produces complex, non-perturbative phonon scattering that defies conventional analytical or simulation-based prediction. Equivariant graph neural networks (GNNs) have demonstrated strong empirical promise in this domain by embedding specific inductive biases directly into their architecture. This paper delivers a purely theoretical analysis of the inductive biases that are necessary and sufficient for reliable thermal transport prediction in disordered materials using equivariant GNNs. We formally define inductive bias as any assumption that constrains the hypothesis space and thereby enables generalization from limited data. Through conceptual arguments we identify three key biases—rotational equivariance, locality (finite cutoff), and energy conservation—as fundamental, while showing how supporting biases of smoothness, permutation invariance, and size extensivity reinforce them. Rotational equivariance ensures that tensorial thermal conductivity transforms correctly under arbitrary material rotations; locality reflects the short-range nature of phonon scattering even in globally disordered systems; and energy conservation guarantees that derived forces remain physically consistent for phonon-frequency calculations. We prove conceptually that these biases collectively reduce sample complexity by shrinking the effective dimension of the hypothesis space, improve extrapolation to new disorder realizations provided local environments remain statistically similar, and enforce physical consistency without requiring explicit regularization. Disordered materials constitute a worst-case regime for machine learning because they lack the global symmetries that crystalline systems implicitly exploit; the biases therefore become not merely advantageous but essential. The analysis concludes with direct implications for GNN architecture design, establishing equivariant GNNs as the theoretically preferred framework for thermal transport in disordered materials. This work supplies the missing theoretical foundation that explains why equivariant architectures succeed where generic models fail and offers precise guidance for future model development in computational materials engineering.