Self-explaining AI as an Alternative to Interpretable AI

Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 12177)


The ability to explain decisions made by AI systems is highly sought after, especially in domains where human lives are at stake such as medicine or autonomous vehicles. While it is often possible to approximate the input-output relations of deep neural networks with a few human-understandable rules, the discovery of the double descent phenomena suggests that such approximations do not accurately capture the mechanism by which deep neural networks work. Double descent indicates that deep neural networks typically operate by smoothly interpolating between data points rather than by extracting a few high level rules. As a result, neural networks trained on complex real world data are inherently hard to interpret and prone to failure if asked to extrapolate. To show how we might be able to trust AI despite these problems we introduce the concept of self-explaining AI. Self-explaining AIs are capable of providing a human-understandable explanation of each decision along with confidence levels for both the decision and explanation. Some difficulties with this approach along with possible solutions are sketched. Finally, we argue it is important that deep learning based systems include a “warning light” based on techniques from applicability domain analysis to warn the user if a model is asked to extrapolate outside its training distribution.


Interpretability Explainability Explainable artificial intelligence XAI Trust Deep learning 


Funding and Disclaimer

No funding sources were used in the creation of this work. The author (Dr. Daniel C. Elton) wrote this article in his personal capacity. The opinions expressed in this article are the author’s own and do not reflect the view of the National Institutes of Health, the Department of Health and Human Services, or the United States government.


  1. 1.
    Adebayo, J., Gilmer, J., Muelly, M., Goodfellow, I., Hardt, M., Kim, B.: Sanity checks for saliency maps. In: Proceedings of the 32nd International Conference on Neural Information Processing Systems NIPS 2018, pp. 9525–9536. Curran Associates Inc., Red Hook (2018)Google Scholar
  2. 2.
    Ahmad, M.A., Eckert, C., Teredesai, A.: Interpretable machine learning in healthcare. In: Proceedings of the 2018 ACM International Conference on Bioinformatics, Computational Biology, and Health Informatics - BCB 2018. ACM Press (2018)Google Scholar
  3. 3.
    Aliman, N.-M., Kester, L.: Hybrid strategies towards safe “Self-Aware” superintelligent systems. In: Iklé, M., Franz, A., Rzepka, R., Goertzel, B. (eds.) AGI 2018. LNCS (LNAI), vol. 10999, pp. 1–11. Springer, Cham (2018). Scholar
  4. 4.
    Alvarez-Melis, D., Jaakkola, T.S.: Towards robust interpretability with self-explaining neural networks. In: Proceedings of the 32nd International Conference on Neural Information Processing Systems NIPS 2018, pp. 7786–7795. Curran Associates Inc., Red Hook (2018)Google Scholar
  5. 5.
    Arya, V., et al.: One explanation does not fit all: a toolkit and taxonomy of AI explainability techniques. arXiv eprints: 1909.03012 (2019)Google Scholar
  6. 6.
    Ashby, W.R.: An Introduction to Cybernetics. Chapman & Hall, London (1956)CrossRefGoogle Scholar
  7. 7.
    Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.R., Samek, W.: On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation. PLoS ONE 10(7), e0130140 (2015)CrossRefGoogle Scholar
  8. 8.
    Barnes, B.C., et al.: Machine learning of energetic material properties. arXiv eprints: 1807.06156 (2018)Google Scholar
  9. 9.
    Beede, E., et al.: A human-centered evaluation of a deep learning system deployed in clinics for the detection of diabetic retinopathy. In: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems CHI 2020, pp. 1–12. Association for Computing Machinery, New York (2020)Google Scholar
  10. 10.
    Belkin, M., Hsu, D., Ma, S., Mandal, S.: Reconciling modern machine-learning practice and the classical bias–variance trade-off. Proc. Natl. Acad. Sci. 116(32), 15849–15854 (2019)MathSciNetCrossRefGoogle Scholar
  11. 11.
    Belkin, M., Hsu, D., Xu, J.: Two models of double descent for weak features. arXiv eprints: 1903.07571 (2019)Google Scholar
  12. 12.
    Bordes, F., Berthier, T., Jorio, L.D., Vincent, P., Bengio, Y.: Iteratively unveiling new regions of interest in deep learning models. In: Medical Imaging with Deep Learning (MIDL) (2018)Google Scholar
  13. 13.
    Bostrom, N.: Superintelligence: Paths, Dangers, Strategies, 1st edn. Oxford University Press Inc., Oxford (2014)Google Scholar
  14. 14.
    Breiman, L.: Statistical modeling: the two cultures (with comments and a rejoinder by the author). Stat. Sci. 16(3), 199–231 (2001)CrossRefGoogle Scholar
  15. 15.
    Chen, C., Li, O., Tao, D., Barnett, A., Rudin, C., Su, J.: This looks like that: deep learning for interpretable image recognition. In: Wallach, H.M., Larochelle, H., Beygelzimer, A., d’Alché-Buc, F., Fox, E.B., Garnett, R. (eds.) Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, 8–14 December 2019, Canada, Vancouver, BC, pp. 8928–8939 (2019)Google Scholar
  16. 16.
    Dombrowski, A.K., Alber, M., Anders, C.J., Ackermann, M., Müller, K.R., Kessel, P.: Explanations can be manipulated and geometry is to blame (2019)Google Scholar
  17. 17.
    Elton, D., Sandfort, V., Pickhardt, P.J., Summers, R.M.: Accurately identifying vertebral levels in large datasets. In: Hahn, H.K., Mazurowski, M.A. (eds.) Medical Imaging 2020: Computer-Aided Diagnosis. SPIE, March 2020Google Scholar
  18. 18.
    Erhan, D., Bengio, Y., Courville, A., Vincent, P.: Visualizing higher-layer features of a deep network. Technical report 1341, University of Montreal: also presented at the ICML 2009 Workshop on Learning Feature Hierarchies. Montréal, Canada (2009)Google Scholar
  19. 19.
    Frosst, N., Hinton, G.: Distilling a neural network into a soft decision tree. arXiv eprintss: 1711.09784 (2017)Google Scholar
  20. 20.
    Gal, Y., Ghahramani, Z.: Dropout as a Bayesian approximation: representing model uncertainty in deep learning. In: Balcan, M.F., Weinberger, K.Q. (eds.) Proceedings of The 33rd International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 48, pp. 1050–1059. PMLR, New York, 20–22 June 2016Google Scholar
  21. 21.
    Goertzel, B.: Are there deep reasons underlying the pathologies of today’s deep learning algorithms? In: Bieger, J., Goertzel, B., Potapov, A. (eds.) AGI 2015. LNCS (LNAI), vol. 9205, pp. 70–79. Springer, Cham (2015). Scholar
  22. 22.
    Goldfeld, Z., et al.: Estimating information flow in deep neural networks. In: Chaudhuri, K., Salakhutdinov, R. (eds.) Proceedings of the 36th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 97, pp. 2299–2308. PMLR, Long Beach, 09–15 June 2019Google Scholar
  23. 23.
    Goodfellow, I.J., Shlens, J., Szegedy, C.: Explaining and harnessing adversarial examples. arXiv eprintss: 1412.6572 (2014)Google Scholar
  24. 24.
    Hasson, U., Nastase, S.A., Goldstein, A.: Direct fit to nature: an evolutionary perspective on biological and artificial neural networks. Neuron 105(3), 416–434 (2020)CrossRefGoogle Scholar
  25. 25.
    Koh, P.W., Liang, P.: Understanding black-box predictions via influence functions. In: Proceedings of the 34th International Conference on Machine Learning ICML 2017, vol. 70, pp. 1885–1894. (2017)Google Scholar
  26. 26.
    Kulesza, T., Burnett, M., Wong, W.K., Stumpf, S.: Principles of explanatory debugging to personalize interactive machine learning. In: Proceedings of the 20th International Conference on Intelligent User Interfaces - IUI 2015. ACM Press (2015)Google Scholar
  27. 27.
    LaLonde, R., Torigian, D., Bagci, U.: Encoding visual attributes in capsules for explainable medical diagnoses. arXiv e-prints: 1909.05926, September 2019Google Scholar
  28. 28.
    Lie, C.: Relevance in the eye of the beholder: diagnosing classifications based on visualised layerwise relevance propagation. Master’s thesis, Lund University, Sweden (2019)Google Scholar
  29. 29.
    Lillicrap, T.P., Kording, K.P.: What does it mean to understand a neural network? arXiv eprints: 1907.06374 (2019)Google Scholar
  30. 30.
    Linfoot, E.: An informational measure of correlation. Inf. Control 1(1), 85–89 (1957)MathSciNetCrossRefGoogle Scholar
  31. 31.
    Lipton, Z.C.: The mythos of model interpretability. arXiv eprints: 1606.03490 (2016)Google Scholar
  32. 32.
    Lundberg, S.M., Lee, S.I.: A unified approach to interpreting model predictions. In: Guyon, I., et al. (eds.) Advances in Neural Information Processing Systems, vol. 30, pp. 4765–4774. Curran Associates, Inc. (2017)Google Scholar
  33. 33.
    McClure, P., et al.: Knowing what you know in brain segmentation using bayesian deep neural networks. Front. Neuroinform. 13, 67 (2019)CrossRefGoogle Scholar
  34. 34.
    Murdoch, W.J., Singh, C., Kumbier, K., Abbasi-Asl, R., Yu, B.: Definitions, methods, and applications in interpretable machine learning. Proc. Natl. Acad. Sci. 116(44), 22071–22080 (2019)MathSciNetCrossRefGoogle Scholar
  35. 35.
    Nakkiran, P., Kaplun, G., Bansal, Y., Yang, T., Barak, B., Sutskever, I.: Deep double descent: where bigger models and more data hurt. arXiv eprints: 1912.02292 (2019)Google Scholar
  36. 36.
    Ribeiro, M.T., Singh, S., Guestrin, C.: Why should I trust you? In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining - KDD. ACM Press (2016)Google Scholar
  37. 37.
    Richards, B.A., et al.: A deep learning framework for neuroscience. Nat. Neurosci. 22(11), 1761–1770 (2019)CrossRefGoogle Scholar
  38. 38.
    Rolnick, D., Kording, K.P.: Identifying weights and architectures of unknown ReLU networks. arXiv eprintss: 1910.00744 (2019)Google Scholar
  39. 39.
    Rudin, C.: Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat. Mach. Intell. 1(5), 206–215 (2019)CrossRefGoogle Scholar
  40. 40.
    Sahigara, F., Mansouri, K., Ballabio, D., Mauri, A., Consonni, V., Todeschini, R.: Comparison of different approaches to define the applicability domain of QSAR models. Molecules 17(5), 4791–4810 (2012)CrossRefGoogle Scholar
  41. 41.
    Shen, S., Han, S.X., Aberle, D.R., Bui, A.A., Hsu, W.: An interpretable deep hierarchical semantic convolutional neural network for lung nodule malignancy classification. Expert Syst. Appl. 128, 84–95 (2019)CrossRefGoogle Scholar
  42. 42.
    Shrikumar, A., Greenside, P., Kundaje, A.: Learning important features through propagating activation differences. arXiv eprintss: 1704.02685 (2017)Google Scholar
  43. 43.
    Spigler, S., Geiger, M., d’Ascoli, S., Sagun, L., Biroli, G., Wyart, M.: A jamming transition from under- to over-parametrization affects generalization in deep learning. J. Phys. A: Math. Theor. 52(47), 474001 (2019)MathSciNetCrossRefGoogle Scholar
  44. 44.
    Sutre, E.T., Colliot, O., Dormont, D., Burgos, N.: Visualization approach to assess the robustness of neural networks for medical image classification. In: Proceedings of the SPIE: Medical Imaging (2020)Google Scholar
  45. 45.
    Swartout, W.R.: XPLAIN: a system for creating and explaining expert consulting programs. Artif. Intell. 21(3), 285–325 (1983)CrossRefGoogle Scholar
  46. 46.
    Torkkola, K.: Feature extraction by non-parametric mutual information maximization. J. Mach. Learn. Res. 3, 1415–1438 (2003)MathSciNetzbMATHGoogle Scholar
  47. 47.
    Yeh, C.K., Hsieh, C.Y., Suggala, A.S., Inouye, D.I., Ravikumar, P.: On the (in)fidelity and sensitivity for explanations. arXiv eprints: 1901.09392 (2019)Google Scholar
  48. 48.
    Zeiler, M.D., Fergus, R.: Visualizing and understanding convolutional networks. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014. LNCS, vol. 8689, pp. 818–833. Springer, Cham (2014). Scholar
  49. 49.
    Zhang, C., Bengio, S., Hardt, M., Recht, B., Vinyals, O.: Understanding deep learning requires rethinking generalization. arXiv eprints: 1611.03530 (2016)Google Scholar
  50. 50.
    Zhang, Q., Cao, R., Shi, F., Wu, Y.N., Zhu, S.C.: Interpreting CNN knowledge via an explanatory graph. In: McIlraith, S.A., Weinberger, K.Q. (eds.) AAAI, pp. 4454–4463. AAAI Press (2018)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.Radiology and Imaging SciencesNational Institutes of Health Clinical CenterBethesdaUSA

Personalised recommendations