Search

In goal-directed materials optimization powered by artificial intelligence, researchers routinely employ teleological language such as “target properties,” “design objectives,” and “optimal structures,” implicitly assuming that materials evolve toward purposes or that optimized outcomes represent intended final causes. Scientific teleology, defined here as the explanatory practice of invoking goals, purposes, or final causes as causal factors within material systems that lack inherent intentionality, constitutes a distinct conceptual failure mode in artificial-intelligence-driven materials science. This failure arises through three primary mechanisms—reification of goals, retrospective teleology, and purpose projection—that systematically distort the epistemic relationship between human-specified objectives and the contingent structure–property relationships uncovered by optimization algorithms. The present analysis articulates a typology of four specific teleological failure modes: teleological overclaim, design-versus-discovery conflation, objective naturalization, and teleological explanation. Detection principles based on language audits, objective genealogy, counterfactual testing, and agency attribution enable researchers to identify these assumptions before they propagate, while five mitigation principles—explicit objective contextualization, literal-versus-metaphorical clarity, multiple-objective transparency, avoidance of agency language, and consistent design-versus-discovery distinction—provide practical safeguards. By treating scientific teleology as an identifiable failure mode rather than an innocuous heuristic, the materials artificial-intelligence community can preserve the epistemic integrity of discovery processes and prevent the misinterpretation of optimized materials as possessing purposes they do not inherently possess.

The progressive integration of artificial intelligence into materials discovery has introduced systems capable of generating hypotheses autonomously. Yet, the problem of scientific autonomy remains largely unexamined as a distinct failure mode within the field. Scientific autonomy is defined here as the degree to which an AI system independently performs hypothesis generation, experimental design, or result interpretation without meaningful human oversight or intervention. This concept must be rigorously distinguished from mere automation, which can still preserve human decision rights. This autonomy introduces multiple mechanisms of failure—including opacity of internal reasoning processes, speed mismatches between AI generation rates and human cognitive capacities, goal misalignments between optimization objectives and epistemic goals, and authority erosion wherein human scientists increasingly defer to machine outputs—each of which undermines the foundational norms of scientific inquiry in materials science. The analysis further articulates a typology of four specific autonomy failure modes—hypothesis proliferation, pathological focus, unaccountable hypotheses, and epistemic lock-in—that manifest uniquely in materials AI contexts such as self-driving laboratories and closed-loop Bayesian optimizers. Detection principles are proposed to identify when autonomy becomes problematic, while mitigation principles emphasize deliberate design strategies to restore appropriate human control. By framing scientific autonomy as a core failure mode rather than an inevitable byproduct of progress, this paper argues for a recalibration of current practices in automated materials hypothesis generation, ensuring that technological advancement does not come at the expense of human epistemic authority or scientific understanding. Ultimately, the work calls for explicit attention to autonomy levels in the design and deployment of materials AI systems to safeguard the integrity of discovery processes.

The scaffolding problem represents a critical yet underrecognized failure mode in artificial intelligence for materials science, wherein seemingly autonomous AI systems depend on vast, unstated layers of scientific infrastructure that remain invisible until they fail. At this point, the AI models themselves cease to function reliably or reproducibly. This failure mode arises because materials AI pipelines are not standalone artifacts but are instead supported by extensive scaffolding—databases, software ecosystems, computational resources, measurement standards, tacit community knowledge, and institutional frameworks—that enable data ingestion, model training, and inference but are rarely documented or maintained as core components of the research. The scaffolding problem is formally defined as the failure to recognize, document, and sustain these invisible support structures, leading to unrecognized vulnerabilities that undermine the reliability of data-driven discoveries in materials design. Materials AI depends on six distinct types of scientific infrastructure, ranging from data repositories such as the Materials Project and AFLOW to software libraries like pymatgen and institutional funding mechanisms, each of which carries hidden assumptions about stability and accessibility. Scaffolding failures occur through mechanisms including infrastructure decay, dependency drift, access loss, and knowledge erosion, producing a typology of four specific failure modes: silent dependency failure, reproducibility collapse, infrastructure lock-in, and knowledge gap failure. Detection relies on systematic dependency mapping, version pinning, access monitoring, reproduction testing, and knowledge auditing, while mitigation demands explicit documentation, containerization, data archiving, dependency minimization, infrastructure independence, and knowledge capture. By articulating the scaffolding problem as a distinct failure mode, this analysis reveals how unexamined infrastructure dependencies threaten the long-term viability of materials AI and calls for a fundamental shift toward treating scaffolding as an explicit, first-class concern in research practice.

Competing scientific ontologies represent a pervasive yet under-analyzed failure mode in artificial intelligence applications for materials science. Different classification systems for the same materials, structures, properties, and processes create fundamental incompatibilities that cause AI models to fail in ways that are difficult to diagnose through conventional performance metrics. This paper defines the ontology problem as the inherent challenge of representing material knowledge when multiple, partially incompatible ontologies coexist within the domain, each encoding distinct conceptual boundaries and relational assumptions. It articulates four primary types of ontological competition—category boundary differences, naming conflicts, relationship differences, and granularity differences—that arise repeatedly in materials informatics. These competitions trigger specific failure modes, including transfer failures, evaluation incompatibilities, data integration failures, and communication breakdowns between research communities. Detection relies on explicit ontology audits and cross-ontology testing, while mitigation centers on mapping strategies, ontology-agnostic representations, and community harmonization efforts. By framing ontology competition as a distinct failure mode, the analysis draws on existing literature to propose an ontology-aware framework that strengthens semantic interoperability and model robustness in materials AI. Ultimately, acknowledging and managing competing ontologies is essential for translating data-driven discoveries into reliable, reproducible knowledge.

Rotational invariance, a core design principle in graph neural networks (GNNs) for materials property prediction, has enabled remarkable accuracy for scalar quantities such as formation energy and band gap yet becomes a fundamental liability when the target property is a tensor of rank two or higher. Anisotropic elastic constants, which describe the directional stiffness of crystals and enter engineering design as a 6×6 Voigt matrix with up to 21 independent components in triclinic systems, must transform covariantly under rotation; any architecture that collapses all orientational information into a single invariant scalar cannot, in principle, recover the required tensorial structure. This failure-mode analysis articulates four distinct mechanisms by which invariance induces systematic error—diagonal indistinguishability, off-diagonal blindness, shear underspecification, and symmetry-induced degeneracy—and demonstrates how each arises directly from the interplay of global pooling, distance-only edge features, and symmetry averaging that characterize rotationally invariant models such as CGCNN, MEGNet, and SchNet. Detection principles based on equality tests, anisotropy-ratio checks, and differential error audits provide practitioners with practical diagnostics, while mitigation principles centered on equivariant architectures, directional encoding, and stress-tensor multi-task learning chart a clear path toward reliable tensorial prediction. The implications extend beyond elasticity to any directional property in low-symmetry crystals and call for a principled redesign of materials GNNs that treats invariance as a deliberate choice rather than a default.

Deep graph neural networks (GNNs) have become a cornerstone of computational materials engineering because they can learn representations directly from crystal graphs. Yet a fundamental failure mode remains hidden when these models are applied to thermal transport: over-smoothing. After only a modest number of message-passing layers, node features converge to a single global vector, erasing the atom-resolved local environments that govern phonon scattering. The problem is dramatically amplified in large unit cells—structures containing 50 or more atoms, common in skutterudites, clathrates, and complex perovskites—because the graph diameter forces the model designer to choose between insufficient propagation depth (missing long-range interactions) and excessive depth (catastrophic smoothing). Lattice thermal conductivity κ emerges from local force constants, anharmonicity, and defect-sensitive scattering; over-smoothing replaces these variations with a featureless average, producing systematically erroneous predictions. This failure-mode analysis demonstrates that the central conflict is structural: thermal transport demands precisely the short- to medium-range distinctions that deep GNNs destroy. Four distinct failure modes are identified and mechanistically linked to the over-smoothing process. Detection principles based on feature similarity, depth sensitivity, and defect response are proposed. Mitigation strategies grounded in architectural modifications—residual connections, normalization, multi-scale pathways, and attention—are shown to preserve local information without sacrificing global coherence. The analysis reveals why standard crystal GNN architectures succeed for formation energies or band gaps yet fail systematically for thermal transport in realistic materials. By reframing over-smoothing as the dominant limitation rather than a minor training artifact, this work provides a conceptual roadmap for next-generation GNNs tailored to phonon physics.

Active learning is widely used in materials discovery to reduce the number of expensive density functional theory calculations required for surrogate-model training. Its dominant acquisition rule, uncertainty sampling, assumes that the most uncertain configuration will also yield the greatest information gain. Although this logic performs well for many bulk properties, it breaks down for defect formation energies. In this setting, uncertainty sampling repeatedly selects uninformative structures, misallocates computational budget, and fails to reach the defective configurations that govern vacancy, interstitial, and substitutional behavior in technologically important materials. This failure reflects a structural mismatch between the acquisition rule and defect physics. Defect configurations are rare in configuration space, energetics are highly localized, and DFT labels often contain substantial aleatoric noise from finite-size effects, charge corrections, and supercell artifacts. These conditions weaken the link between predictive uncertainty and useful learning signal, while representation bias in models trained mainly on perfect crystals further erodes selectivity. This study develops a failure-mode analysis of uncertainty sampling for defect formation energies, identifying recurrent breakdowns in sampling, representation, calibration, and budget use. It also outlines practical detection principles and defect-aware mitigation strategies, including initialization with defective structures, hybrid acquisition functions, epistemic-only scoring, and budget partitioning. The central implication is that active learning for defect discovery cannot rely on generic uncertainty-based querying. Effective acceleration in this domain requires acquisition strategies designed around the rarity, locality, and noise structure of defect energetics.

Machine learning interatomic potentials achieve high accuracy at molecular-dynamics scales but rely on local descriptors that truncate long-range interactions. In charge-transfer-sensitive materials, where electrostatics decay as 1/r, this assumption fails. The resulting mismatch produces systematic errors, including non-convergent energies, force discontinuities, unstable charge distributions, oxidation-state ambiguity, suppressed ionic transport, and collapsed dielectric response. These effects are not incidental but arise directly from the absence of global electrostatic coupling. This work presents a failure-mode analysis that links these breakdowns to their physical origin and introduces practical diagnostics based on standard simulation outputs. It further outlines mitigation strategies that restore long-range physics through electrostatic solvers, charge equilibration, multipole representations, and hybrid architectures. By establishing when and why locality fails, the study defines the minimal requirements for charge-transfer-aware machine learning potentials and provides a pathway toward reliable simulation of ionic and polar materials.

Graph neural networks (GNNs) routinely achieve low errors in formation energy prediction, yet they systematically fail for metastable phases lying above the convex hull. These kinetically accessible materials, critical for batteries, catalysis, and thin-film applications, are severely underrepresented in training data dominated by stable ground states. This failure-mode analysis identifies four interlocking issues: (1) convex hull bias that pulls predictions downward toward stable energies, (2) energy range compression that collapses the predicted dynamic range, (3) local environment extrapolation failures caused by strained or unusual atomic geometries unseen in relaxed training data, and (4) error asymmetry producing consistent underprediction for metastable structures. We describe the origins, signatures, and discovery consequences of each mode, propose simple diagnostic tests for practitioners, and outline targeted mitigation strategies including metastable data augmentation, hull-aware regularization, and physics-informed extrapolation. Current GNN performance on convex-hull benchmarks does not guarantee reliability for metastable discovery. Addressing these failure modes is essential for trustworthy machine-learning-assisted design of functional materials that exist because of their position above the convex hull.