analiza aiqbench
The paper argues that easier construction of hybrid "grey-box" models, where deep learning components sit alongside mechanistic equations, does not make their inferred mechanisms trustworthy. A good fit can hide errors because the learned component can absorb them, and many biological datasets admit multiple equally good models. Its proposed remedies, causal identifiability analysis and ensemble uncertainty mapping, are useful for ML practitioners building hybrid models in science, since they shift the goal from a single good fit to a clear account of what the data supports and which experiment would settle the question.
When Everyone Can Build a Model: Mechanistic Inference in the Age of AI
arXiv:2610.41454v1 Announce Type: new Abstract: Modern deep learning and AI-assisted code generation are lowering the technical barrier to building mathematical models of biological systems that reproduce complex observations. This lets researchers with domain expertise build semi-mechanistic, hybrid models constrained by prior knowledge and augmented with learned structure without extensive training in modelling. However, easier construction does not make interpretation easier: predictive success of such so-called grey-box models does not establish that an inferred mechanism corresponds to the one that generated the data, and the learned component can absorb errors in the mechanistic part. Many biological modelling problems are underdetermined, so several, possibly infinitely many, models may reproduce the same data equally well. This ambiguity enters at two modelling choices: which variables can affect others, and which response function describes the underlying mechanism. For the first choice, causal inference provides a formal framework stating which causal relationships and intervention effects are identifiable from the data and available knowledge. For the second, ensemble analyses quantify the uncertainty and the range of behaviours still allowed by models that fit equally well. Together, they turn a single good fit into an explicit account of what is supported, what remains open, and which experiment would decide.
Authors:Marvin van Aalst (1), Adèle H. Ribeiro (2 and 3), Anna Matuszyńska (1 and 3) ((1) Computational Life Science, Faculty of Mathematics and Natural Sciences, RWTH Aachen University, Germany, (2) Causal and Explainable Artificial Intelligence in Life Sciences, Faculty of Mathematics and Natural Sciences, RWTH Aachen University, Germany, (3) Center for Computational Life Sciences, RWTH Aachen University, Germany)
Abstract:Modern deep learning and AI-assisted code generation are lowering the technical barrier to building mathematical models of biological systems that reproduce complex observations. This lets researchers with domain expertise build semi-mechanistic, hybrid models constrained by prior knowledge and augmented with learned structure without extensive training in modelling. However, easier construction does not make interpretation easier: predictive success of such so-called grey-box models does not establish that an inferred mechanism corresponds to the one that generated the data, and the learned component can absorb errors in the mechanistic part. Many biological modelling problems are underdetermined, so several, possibly infinitely many, models may reproduce the same data equally well. This ambiguity enters at two modelling choices: which variables can affect others, and which response function describes the underlying mechanism. For the first choice, causal inference provides a formal framework stating which causal relationships and intervention effects are identifiable from the data and available knowledge. For the second, ensemble analyses quantify the uncertainty and the range of behaviours still allowed by models that fit equally well. Together, they turn a single good fit into an explicit account of what is supported, what remains open, and which experiment would decide.
Machine Learning (cs.LG)
arXiv:2610.41454 [cs.LG]
(or arXiv:2610.41454v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.41454 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 272 cuvinte
Subiecte conexe
MechReasoner: A Simulator and Benchmark for Mechanistic Reasoning in Qualitative Physics
arXiv · 29 sept. 2026
Reliable mechanistic operator recovery with biologically-informed neural networks: principles for architecture and optimisation design
arXiv · 5 oct. 2026
The Struggle Between Continuation and Refusal: A Mechanistic Analysis of the Continuation-Triggered Jailbreak in LLMs
arXiv · 7 sept. 2026
Fundamental Dynamical Units for Physics-Informed Structural Inference from Perturbation Time-Series in Networked Systems
arXiv · 14 sept. 2026