pith. sign in

arxiv: 2606.24953 · v1 · pith:Q4DVW3XInew · submitted 2026-06-23 · 💻 cs.LG · nlin.AO

How Complexity Contributes to Learning Opacity in Machine Learning

Joachim Stein , Eric Raidl This is my paper

Pith reviewed 2026-06-26 00:55 UTC · model grok-4.3

classification 💻 cs.LG nlin.AO
keywords machine learning opacityneural network trainingcomplex dynamical systemsgradient optimizationdata sensitivitylearning dynamicsirreducible opacity
0
0 comments X

The pith

Neural network learning is a complex dynamical system whose properties make the training process opaque.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that the opacity of how neural networks learn stems from their behavior as complex systems. It identifies three fundamental properties of the training process: sensitivity to initial weights, feedback loops in gradient descent, and dependence on specific training data. These create dynamical complexity that resists full explanation. Because these properties are essential to how learning works, removing them would change the nature of machine learning itself. This suggests that some forms of learning opacity cannot be eliminated without sacrificing the system's effectiveness.

Core claim

Neural network training exhibits three key complex-system properties—sensitivity to weight initialization, feedback in gradient-based optimization, and sensitivity to training data—that generate dynamical complexity and thereby produce learning opacity. These properties are intrinsic to the learning process, so efforts to reduce opacity by damping them would alter the fundamental character of machine learning. Consequently some sources of opacity are irreducible.

What carries the argument

The framing of neural network learning as a complex dynamical system driven by the three properties of initialization sensitivity, optimization feedback, and data sensitivity.

If this is right

  • The evolution of weights during training cannot be fully tracked or predicted due to sensitivity to initialization.
  • Feedback in gradient-based optimization produces unpredictable dynamical behavior over time.
  • Sensitivity to training data means small data variations create different learning trajectories.
  • Damping or removing these properties would change the fundamental nature of machine learning.
  • Some sources of learning opacity are irreducible.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same framing could apply to other iterative optimization methods beyond neural networks.
  • Practical work in ML might shift from eliminating opacity to managing its effects.
  • The properties may connect to known chaotic or sensitive behaviors studied in nonlinear dynamics.
  • Experiments could check whether stabilizing one property measurably reduces opacity without harming accuracy.

Load-bearing premise

The three properties are fundamental to the learning process such that damping or eliminating them would fundamentally alter how ML systems learn.

What would settle it

A training procedure that removes or damps one of the three properties while preserving standard learning performance and model capabilities on benchmark tasks.

read the original abstract

Machine learning (ML) algorithms are known to be opaque. We do not know the reasons for their predictions. The learning process leading to the prediction function is also opaque. We do not fully understand the time evolution of the weight values of neural nets (NN) and related dynamical phenomena. While prediction opacity is widely studied, learning opacity remains largely underexplored. This article studies learning opacity trough the lens of complex dynamical systems. We argue that NN learning is essentially a complex system and that learning opacity is due to dynamical complexity and the epistemological challenges that arise from it. We identify three key properties of training complexity -- sensitivity to weight initialization, feedback in gradient based optimization, and sensitivity to the training data -- and show how each contributes to learning opacity. As these properties are fundamental to the learning process damping or eliminating them would fundamentally alter how ML systems learn. Some sources of opacity in ML may hence be irreducible.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper claims that neural network learning is a complex dynamical system whose opacity arises from three properties—sensitivity to weight initialization, feedback in gradient-based optimization, and sensitivity to training data—and concludes that some sources of learning opacity are therefore irreducible because damping these properties would fundamentally change ML learning.

Significance. If the interpretive mapping were grounded, the work would offer a philosophical lens on why certain opacity phenomena may resist technical mitigation, potentially informing long-term research priorities in interpretability.

major comments (2)
  1. [Abstract] Abstract: the central claim that the three properties are 'fundamental to the learning process' such that 'damping or eliminating them would fundamentally alter how ML systems learn' is asserted by definition rather than derived from any model, theorem, or counter-factual analysis showing that gradient-based learning cannot be preserved while attenuating opacity.
  2. [Abstract] Abstract: no independent grounding or external benchmark is supplied for the conclusion that 'some sources of opacity in ML may hence be irreducible'; the argument reduces to labeling the listed behaviors as complex and then inferring irreducibility from that label.
minor comments (1)
  1. [Abstract] Abstract: 'trough' is a typographical error and should read 'through'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and for highlighting areas where the abstract's claims could be clarified. We respond to each major comment below, maintaining that the manuscript offers a conceptual analysis grounded in dynamical systems properties of standard neural network training rather than a formal theorem.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the three properties are 'fundamental to the learning process' such that 'damping or eliminating them would fundamentally alter how ML systems learn' is asserted by definition rather than derived from any model, theorem, or counter-factual analysis showing that gradient-based learning cannot be preserved while attenuating opacity.

    Authors: The manuscript frames the three properties as direct consequences of the standard formulation of gradient-based optimization (non-convex loss landscapes with random initialization, iterative parameter updates, and empirical risk minimization over finite data). These are not arbitrary definitions but established features of the training dynamics. While we do not supply a formal theorem or explicit counter-factual simulation, the interpretive claim follows from the observation that attenuating any of them (e.g., via deterministic initialization or non-iterative methods) would depart from current gradient-based ML practice. We will add a brief clarifying clause to the abstract to make this basis more explicit. revision: partial

  2. Referee: [Abstract] Abstract: no independent grounding or external benchmark is supplied for the conclusion that 'some sources of opacity in ML may hence be irreducible'; the argument reduces to labeling the listed behaviors as complex and then inferring irreducibility from that label.

    Authors: The grounding is supplied by the established results in complex dynamical systems theory, where sensitivity to initial conditions, feedback loops, and input sensitivity are known to produce irreducible unpredictability in system trajectories. The paper maps these properties onto neural network training and draws the logical implication for opacity; it does not treat 'complex' as a label but as a technical characterization with epistemological consequences. This is an interpretive rather than empirical conclusion, and we do not agree that it reduces to mere labeling. revision: no

Circularity Check

1 steps flagged

Irreducibility of opacity follows directly from definitional claim that three training properties are fundamental

specific steps
  1. self definitional [Abstract]
    "We identify three key properties of training complexity -- sensitivity to weight initialization, feedback in gradient based optimization, and sensitivity to the training data -- and show how each contributes to learning opacity. As these properties are fundamental to the learning process damping or eliminating them would fundamentally alter how ML systems learn. Some sources of opacity in ML may hence be irreducible."

    The text first identifies the properties and then declares them 'fundamental,' from which the irreducibility conclusion is derived by definition. The step equates 'these properties cannot be damped without altering learning' with 'opacity is irreducible' without additional grounding, making the result equivalent to the input assumption.

full rationale

The paper's central claim reduces to a single definitional move: the three listed behaviors are stipulated as fundamental to gradient-based learning, from which the conclusion that some opacity is irreducible follows by construction. No independent derivation, external benchmark, or counterfactual analysis is supplied to establish that these behaviors cannot be attenuated. This matches the self-definitional pattern exactly, with the quoted abstract text serving as the load-bearing step. No equations, self-citations, or fitted parameters are present, so the circularity is limited to this philosophical assertion rather than a technical reduction chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The claim rests on the domain assumption that neural network training qualifies as a complex dynamical system whose three listed properties are both definitional and causally responsible for opacity; no free parameters or invented entities appear.

axioms (1)
  • domain assumption Neural network training is essentially a complex dynamical system
    Invoked in the abstract as the lens for studying learning opacity and the source of the three key properties.

pith-pipeline@v0.9.1-grok · 5681 in / 1159 out tokens · 16330 ms · 2026-06-26T00:55:15.933493+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

246 extracted references · 72 canonical work pages

  1. [1]

    https://doi.org/10.1093/acprof:oso/9780199213900.001.0001

    Weisberg, Michael , isbn =. Three Kinds of Models , booktitle =. 2013 , month =. doi:10.1093/acprof:oso/9780199933662.003.0002 , url =

    work page doi:10.1093/acprof:oso/9780199933662.003.0002 2013

  2. [2]

    , title =

    Boge, Florian J. , title =. The British Journal for the Philosophy of Science , volume =. 2024 , doi =

    2024

  3. [3]

    Philosophical Studies , volume =

    Margaret Morrison , title =. Philosophical Studies , volume =. 2009 , doi =

    2009

  4. [4]

    and Fei-Fei, Li , title =

    Russakovsky, Olga and Deng, Jia and Su, Hao and Krause, Jonathan and Satheesh, Sanjeev and Ma, Sean and Huang, Zhiheng and Karpathy, Andrej and Khosla, Aditya and Bernstein, Michael and Berg, Alexander C. and Fei-Fei, Li , title =. International Journal of Computer Vision , year =

  5. [5]

    and Milan, Kieran and Quan, John and Ramalho, Tiago and Grabska-Barwinska, Agnieszka and Hassabis, Demis and Clopath, Claudia and Kumaran, Dharshan and Hadsell, Raia , year=

    James Kirkpatrick and Razvan Pascanu and Neil Rabinowitz and Joel Veness and Guillaume Desjardins and Andrei A. Rusu and Kieran Milan and John Quan and Tiago Ramalho and Agnieszka Grabska-Barwinska and Demis Hassabis and Claudia Clopath and Dharshan Kumaran and Raia Hadsell , title =. Proceedings of the National Academy of Sciences , volume =. 2017 , doi ...

    work page doi:10.1073/pnas.1611835114 2017

  6. [6]

    British Journal for the Philosophy of Science , volume=

    Explanation and Invariance in the Special Sciences , author=. British Journal for the Philosophy of Science , volume=. 2000 , publisher=

    2000

  7. [7]

    Biology & Philosophy , volume =

    Woodward, James , title =. Biology & Philosophy , volume =. 2010 , doi =

    2010

  8. [8]

    2015 , eprint=

    In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning , author=. 2015 , eprint=

    2015

  9. [9]

    Sterkenburg and Peter D

    Tom F. Sterkenburg and Peter D. Grünwald , journal =. The no-free-lunch theorems of supervised learning , urldate =

  10. [10]

    2025 , eprint=

    The Benchmarking Epistemology: Construct Validity for Evaluating Machine Learning Models , author=. 2025 , eprint=

    2025

  11. [11]

    and Dieks, Dennis , title =

    de Regt, Henk W. and Dieks, Dennis , title =. Synthese , volume =. 2005 , doi =

    2005

  12. [12]

    Philosophy of Science , author=

    Exporting Causal Knowledge in Evolutionary and Developmental Biology , volume=. Philosophy of Science , author=. 2008 , pages=. doi:10.1086/594515 , number=

    work page doi:10.1086/594515 2008

  13. [13]

    The British Journal for the Philosophy of Science , year =

    Buchholz, Oliver and Raidl, Eric , title =. The British Journal for the Philosophy of Science , year =. doi:10.1086/721797 , URL =

    work page doi:10.1086/721797

  14. [14]

    Accountability inan Algorithmic Society: Relation- ality, Responsibility, and Robustness in Machine Learning

    Bordt, Sebastian and Finck, Mich\`. Post-Hoc Explanations Fail to Achieve their Purpose in Adversarial Contexts , year =. Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency , pages =. doi:10.1145/3531146.3533153 , abstract =

    work page doi:10.1145/3531146.3533153 2022

  15. [15]

    , year =

    Mitchell, Sandra D. , year =. The landscape of integrative pluralism , volume =. THEORIA. An International Journal for Theory, History and Foundations of Science , doi =

  16. [16]

    Harman, Gilbert and Kulkarni, Sanjeev , title =

  17. [17]

    Sterkenburg , title =

    Tom F. Sterkenburg , title =. Minds and Machines , year =. doi:10.1007/s11023-024-09703-y , url =

    work page doi:10.1007/s11023-024-09703-y

  18. [18]

    Grimm, Stephen , title =. The. 2024 , edition =

    2024

  19. [19]

    Explaining Understanding: New Perspectives from Epistemology and Philosophy of Science , editor =

    Christoph Baumberger and Claus Beisbart and Georg Brun , title =. Explaining Understanding: New Perspectives from Epistemology and Philosophy of Science , editor =. 2016 , publisher =

    2016

  20. [20]

    International Conference on Learning Representations , year=

    Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability , author=. International Conference on Learning Representations , year=

  21. [21]

    , title =

    Mitchell, Sandra D. , title =. Akteure – Mechanismen – Modelle: Zur Theoriefähigkeit makro-sozialer Analysen , editor =. 2002 , publisher =

    2002

  22. [22]

    Proceedings of The 13th Asian Conference on Machine Learning , pages =

    Revisiting Weight Initialization of Deep Neural Networks , author =. Proceedings of The 13th Asian Conference on Machine Learning , pages =. 2021 , editor =

    2021

  23. [23]

    2024 , title =

    Phillip Hintikka Kieval and Oscar Westerblad , keywords =. 2024 , title =

    2024

  24. [24]

    Understanding Optimization in Deep Learning with Central Flows , url =

    Cohen, Jeremy and Damian, Alex and Talwalkar, Ameet and Kolter, Zico and Lee, Jason , booktitle =. Understanding Optimization in Deep Learning with Central Flows , url =

  25. [25]

    2025 , eprint=

    Understanding Optimization in Deep Learning with Central Flows , author=. 2025 , eprint=

    2025

  26. [26]

    2025 , eprint=

    Deep Learning is Not So Mysterious or Different , author=. 2025 , eprint=

    2025

  27. [27]

    Zhang, Chiyuan and Bengio, Samy and Hardt, Moritz and Recht, Benjamin and Vinyals, Oriol , title =. Commun. ACM , month = feb, pages =. 2021 , issue_date =. doi:10.1145/3446776 , abstract =

    work page doi:10.1145/3446776 2021

  28. [28]

    Philosophy Compass , volume =

    Grote, Thomas and Genin, Konstantin and Sullivan, Emily , title =. Philosophy Compass , volume =. doi:https://doi.org/10.1111/phc3.12974 , url =. https://compass.onlinelibrary.wiley.com/doi/pdf/10.1111/phc3.12974 , abstract =

    work page doi:10.1111/phc3.12974

  29. [29]

    Scientific Understanding: Philosophical Perspectives , pages =

    Tarja Knuuttila and Martina Merz , title =. Scientific Understanding: Philosophical Perspectives , pages =. 2009 , address =

    2009

  30. [30]

    2004 , address =

    Paul Humphreys , title =. 2004 , address =. doi:10.1093/0199266858.001.0001 , url =

    work page doi:10.1093/0199266858.001.0001 2004

  31. [31]

    Nate Rahn, Allison Qi, Avery Griffin, Jonathan Michala, Henry Sleight, and Erik Jones

    Lenhard, Johannes , title =. 2019 , month =. doi:10.1093/oso/9780190873288.001.0001 , url =

    work page doi:10.1093/oso/9780190873288.001.0001 2019

  32. [32]

    Scientific Understanding: Philosophical Perspectives , editor =

    Johannes Lenhard , title =. Scientific Understanding: Philosophical Perspectives , editor =. 2009 , pages =

    2009

  33. [33]

    Surprised by a Nanowire: Simulation, Control, and Understanding , urldate =

    Johannes Lenhard , journal =. Surprised by a Nanowire: Simulation, Control, and Understanding , urldate =

  34. [34]

    Synthese , year =

    Anouk Barberousse and Marion Vorms , title =. Synthese , year =. doi:10.1007/s11229-014-0482-6 , url =

    work page doi:10.1007/s11229-014-0482-6

  35. [35]

    SCIENCE AND COMPLEXITY , urldate =

    Warren Weaver , journal =. SCIENCE AND COMPLEXITY , urldate =

  36. [36]

    1961 , publisher=

    Cybernetics: Or, Control and Communication in the Animal and the Machine , author=. 1961 , publisher=

    1961

  37. [37]

    Zuchowski , title =

    Lena C. Zuchowski , title =. 2017 , publisher =

    2017

  38. [38]

    Alligood and Tim D

    Kathleen T. Alligood and Tim D. Sauer and James A. Yorke , title =. 1997 , isbn =

    1997

  39. [39]

    Layek , title =

    G.C. Layek , title =. 2015 , isbn =

    2015

  40. [40]

    Edward N. Lorenz. Deterministic Nonperiodic Flow. Journal of Atmospheric Sciences. 1963. doi:10.1175/1520-0469(1963)020<0130:DNF>2.0.CO;2

    work page doi:10.1175/1520-0469(1963)020 1963

  41. [41]

    Complexity , volume =

    Gell-Mann, Murray , title =. Complexity , volume =. doi:https://doi.org/10.1002/cplx.6130010105 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/cplx.6130010105 , year =

    work page doi:10.1002/cplx.6130010105

  42. [42]

    ICES Journal of Marine Science , volume =

    Volterra, Vito , title =. ICES Journal of Marine Science , volume =. 1928 , doi =

    1928

  43. [43]

    Nature , volume =

    Volterra, Vito , title =. Nature , volume =. 1926 , doi =

    1926

  44. [44]

    1925 , publisher =

    Elements of Physical Biology , author =. 1925 , publisher =

    1925

  45. [45]

    Proceedings of the Royal Society of London , volume =

    Maxwell, James Clerk , title =. Proceedings of the Royal Society of London , volume =. 1868 , publisher =

  46. [46]

    2019 , month =

    Wiener, Norbert , title =. 2019 , month =. doi:10.7551/mitpress/11810.001.0001 , url =

    work page doi:10.7551/mitpress/11810.001.0001 2019

  47. [47]

    Behavior, Purpose and Teleology , urldate =

    Arturo Rosenblueth and Norbert Wiener and Julian Bigelow , journal =. Behavior, Purpose and Teleology , urldate =

  48. [48]

    Kolmogorov , title =

    Andrey N. Kolmogorov , title =. Theoretical Computer Science , volume =. 1998 , doi =

    1998

  49. [49]

    Kolmogorov , title =

    Andrey N. Kolmogorov , title =. Problems of Information Transmission , volume =

  50. [50]

    Kolmogorov , title =

    Andrey N. Kolmogorov , title =. Sankhy

  51. [51]

    Stearns , title =

    Juris Hartmanis and Richard E. Stearns , title =. Transactions of the American Mathematical Society , volume =. 1965 , doi =

    1965

  52. [52]

    Metaphysical Emergence: Weak and Strong

    Jessica Wilson. Metaphysical Emergence: Weak and Strong. Metaphysics in Contemporary Physics. 2016. doi:10.1163/9789004310827_015

    work page doi:10.1163/9789004310827_015 2016

  53. [53]

    Metaphysics in Contemporary Physics

    Tomasz Bigaj and Christian Wüthrich. Metaphysics in Contemporary Physics. 2015. doi:10.1163/9789004310827

    work page doi:10.1163/9789004310827 2015

  54. [54]

    Nate Rahn, Allison Qi, Avery Griffin, Jonathan Michala, Henry Sleight, and Erik Jones

    Humphreys, Paul , title =. 2016 , month =. doi:10.1093/acprof:oso/9780190620325.001.0001 , url =

    work page doi:10.1093/acprof:oso/9780190620325.001.0001 2016

  55. [55]

    P. W. Anderson , title =. Science , volume =. 1972 , doi =

    1972

  56. [56]

    Simple measure for complexity , author =. Phys. Rev. E , volume =. 1999 , month =. doi:10.1103/PhysRevE.59.1459 , url =

    work page doi:10.1103/physreve.59.1459 1999

  57. [57]

    and Bar-Yam, Yaneer , title =

    Siegenfeld, Alexander F. and Bar-Yam, Yaneer , title =. Complexity , volume =. doi:https://doi.org/10.1155/2020/6105872 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1155/2020/6105872 , abstract =

    work page doi:10.1155/2020/6105872 2020

  58. [58]

    International Conference on Learning Representations , year=

    Understanding deep learning requires rethinking generalization , author=. International Conference on Learning Representations , year=

  59. [59]

    , year = 1993, month = jul, journal =

    Jeffrey L. Elman , abstract =. Learning and development in neural networks: the importance of starting small , journal =. 1993 , issn =. doi:https://doi.org/10.1016/0010-0277(93)90058-4 , url =

    work page doi:10.1016/0010-0277(93)90058-4 1993

  60. [60]

    International Journal of Computer Vision , year=

    Soviany, Petru and Ionescu, Radu Tudor and Rota, Paolo and Sebe, Nicu , title=. International Journal of Computer Vision , year=. doi:10.1007/s11263-022-01611-x , url=

    work page doi:10.1007/s11263-022-01611-x

  61. [61]

    A Survey on Curriculum Learning , year=

    Wang, Xin and Chen, Yudong and Zhu, Wenwu , journal=. A Survey on Curriculum Learning , year=

  62. [62]

    Proceedings of the 36th International Conference on Machine Learning , pages =

    On The Power of Curriculum Learning in Training Deep Networks , author =. Proceedings of the 36th International Conference on Machine Learning , pages =. 2019 , editor =

    2019

  63. [63]

    Curriculum learning

    Bengio, Yoshua and Louradour, J\'. Curriculum learning , year =. Proceedings of the 26th Annual International Conference on Machine Learning , pages =. doi:10.1145/1553374.1553380 , abstract =

    work page doi:10.1145/1553374.1553380

  64. [64]

    arXiv preprint arXiv:1412.6558 , year=

    Random walk initialization for training very deep feedforward networks , author=. arXiv preprint arXiv:1412.6558 , year=

    Pith/arXiv arXiv

  65. [65]

    The Vanishing Gradient Problem During Learning Recurrent Neural Nets and Problem Solutions , volume =

    Hochreiter, Sepp , year =. The Vanishing Gradient Problem During Learning Recurrent Neural Nets and Problem Solutions , volume =. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems , doi =

  66. [66]

    and Simard, P

    Bengio, Y. and Simard, P. and Frasconi, P. , journal=. Learning long-term dependencies with gradient descent is difficult , year=

  67. [67]

    Rumelhart, Geoffrey E

    Rumelhart, David E. and Hinton, Geoffrey E. and Williams, Ronald J. , title=. Nature , year=. doi:10.1038/323533a0 , url=

    work page doi:10.1038/323533a0

  68. [68]

    and Chuanyi Ji , journal=

    Atiya, A. and Chuanyi Ji , journal=. How initial conditions affect generalization performance in large networks , year=

  69. [69]

    Human-AI coevolution , journal =

    Dino Pedreschi and Luca Pappalardo and Emanuele Ferragina and Ricardo Baeza-Yates and Albert-László Barabási and Frank Dignum and Virginia Dignum and Tina Eliassi-Rad and Fosca Giannotti and János Kertész and Alistair Knott and Yannis Ioannidis and Paul Lukowicz and Andrea Passarella and Alex Sandy Pentland and John Shawe-Taylor and Alessandro Vespignani ...

    work page doi:10.1016/j.artint.2024.104244 2025

  70. [70]

    Davani, A., Díaz, M., Baker, D., and Prabhakaran, V

    Zezulka, Sebastian and Genin, Konstantin , title =. Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency , pages =. 2024 , isbn =. doi:10.1145/3630106.3659020 , abstract =

    work page doi:10.1145/3630106.3659020 2024

  71. [71]

    Explanatory Unification and the causal strcuture of the world , booktitle=

    Philip Kitcher , editor=. Explanatory Unification and the causal strcuture of the world , booktitle=. 1989 , pages=

    1989

  72. [72]

    Understanding Sharpness Dynamics in

    Geonhui Yoo and Minhak Song and Chulhee Yun , booktitle=. Understanding Sharpness Dynamics in. 2025 , url=

    2025

  73. [73]

    Philosophy of Science , author=

    Dimensions of Scientific Law , volume=. Philosophy of Science , author=. 2000 , pages=. doi:10.1086/392774 , number=

    work page doi:10.1086/392774 2000

  74. [74]

    Explanatory Unification , urldate =

    Philip Kitcher , journal =. Explanatory Unification , urldate =

  75. [75]

    Explanation and Scientific Understanding , urldate =

    Michael Friedman , journal =. Explanation and Scientific Understanding , urldate =

  76. [76]

    Structural Representation and Surrogative Reasoning , volume =

    Chris Swoyer , doi =. Structural Representation and Surrogative Reasoning , volume =. Synthese , number =

  77. [77]

    , title =

    Duran, Juan M. , title =. The Routledge Handbook of Philosophy of Scientific Modeling , publisher =. 2024 , doi =

    2024

  78. [78]

    Bayesianism vs

    Sprenger, Jan , isbn =. Bayesianism vs. Frequentism in Statistical Inference , booktitle =. 2016 , pages =

    2016

  79. [79]

    Null hypothesis significance tests

    Schneider, Jesper , year =. Null hypothesis significance tests. A mix-up of two different theories: The basis for widespread confusion and numerous misinterpretations , volume =. Scientometrics , doi =

  80. [80]

    Models as autonomous agents , booktitle=

    Morrison, Margaret , editor=. Models as autonomous agents , booktitle=. 1999 , pages=

    1999

Showing first 80 references.