Beneficial and Harmful Explanatory Machine Learning

Given the recent successes of Deep Learning in AI there has been increased interest in the role and need for explanations in machine learned theories. A distinct notion in this context is that of Michie's definition of Ultra-Strong Machine Learning (USML). USML is demonstrated by a measurable increase in human performance of a task following provision to the human of a symbolic machine learned theory for task performance. A recent paper demonstrates the beneficial effect of a machine learned logic theory for a classification task, yet no existing work has examined the potential harmfulness of machine's involvement in human learning. This paper investigates the explanatory effects of a machine learned theory in the context of simple two person games and proposes a framework for identifying the harmfulness of machine explanations based on the Cognitive Science literature. The approach involves a cognitive window consisting of two quantifiable bounds and it is supported by empirical evidence collected from human trials. Our quantitative and qualitative results indicate that human learning aided by a symbolic machine learned theory which satisfies a cognitive window has achieved significantly higher performance than human self learning. Results also demonstrate that human learning aided by a symbolic machine learned theory that fails to satisfy this window leads to significantly worse performance than unaided human learning.

[1]  Tim Miller,et al.  Explainable AI: Beware of Inmates Running the Asylum Or: How I Learnt to Stop Worrying and Love the Social and Behavioural Sciences , 2017, ArXiv.

[2]  Ute Schmid,et al.  A Human Like Incremental Decision Tree Algorithm: Combining Rule Learning, Pattern Induction, and Storing Examples , 2017, LWDA.

[3]  Amina Adadi,et al.  Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI) , 2018, IEEE Access.

[4]  Donald Michie,et al.  Cognitive models from subcognitive skills , 1990 .

[5]  Stephen K. Reed,et al.  Selecting analogous problems: Similarity versus inclusiveness , 1990, Memory & cognition.

[6]  Kevin Barraclough,et al.  I and i , 2001, BMJ : British Medical Journal.

[7]  Ute Schmid,et al.  Automatic Generation of Analogous Problems to Help Resolving Misconceptions in an Intelligent Tutor System for Written Subtraction , 2016, ICCBR Workshops.

[8]  J. Ross Quinlan,et al.  Simplifying decision trees , 1987, Int. J. Hum. Comput. Stud..

[9]  Rui Camacho,et al.  Building symbolic representations of intuitive real-time skills from performance data , 1994, Machine Intelligence 13.

[10]  Ehud Shapiro,et al.  Algorithmic Program Debugging , 1983 .

[11]  Vítor Santos Costa,et al.  Inductive Logic Programming , 2013, Lecture Notes in Computer Science.

[12]  Noam Chomsky,et al.  The faculty of language: what is it, who has it, and how did it evolve? , 2002, Science.

[13]  Ben J. A. Kröse,et al.  Learning from delayed rewards , 1995, Robotics Auton. Syst..

[14]  Jerry R. Hobbs.  Abduction in Natural Language Understanding , 2008 .

[15]  Ute Schmid,et al.  Empirical Evidence for Derivational Analogy , 2020, Proceedings of the Twenty First Annual Conference of the Cognitive Science Society.

[16]  J. Marks.  Performance and Competence in Second Language Acquisition , 1998 .

[17]  Ivan Bratko,et al.  Behavioural Cloning: Phenomena, Results and Problems , 1995 .

[18]  J. Gregory Trafton,et al.  Memory for goals: an activation-based model , 2002, Cogn. Sci..

[19]  Stephen Muggleton,et al.  Ultra-Strong Machine Learning: comprehensibility of programs learned with ILP , 2018, Machine Learning.

[20]  Stephen Muggleton,et al.  Learning optimal chess strategies , 1994, Machine Intelligence 13.

[21]  Amit Dhurandhar,et al.  TED: Teaching AI to Explain its Decisions , 2018, AIES.

[22]  G. A. Miller.  THE PSYCHOLOGICAL REVIEW THE MAGICAL NUMBER SEVEN, PLUS OR MINUS TWO: SOME LIMITS ON OUR CAPACITY FOR PROCESSING INFORMATION 1 , 1956 .

[23]  Ute Schmid,et al.  Inductive rule learning on the knowledge level , 2011, Cognitive Systems Research.

[24]  Tom M. Mitchell,et al.  Generalization as Search , 2002 .

[25]  Stephanie M. Stalinski,et al.  Journal of Experimental Psychology: Learning, Memory, and Cognition , 2012 .

[26]  Jason Weston,et al.  Curriculum learning , 2009, ICML '09.

[27]  Tim Miller,et al.  Explanation in Artificial Intelligence: Insights from the Social Sciences , 2017, Artif. Intell..

[28]  John R. Anderson,et al.  Use of analogy in a production system architecture , 1989 .

[29]  Stephen Muggleton,et al.  Meta-interpretive learning of higher-order dyadic datalog: predicate invention revisited , 2013, Machine Learning.

[30]  Stephen Muggleton,et al.  Bias reformulation for one-shot function induction , 2014, ECAI.

[31]  Stephen Muggleton,et al.  Machine Discovery of Comprehensible Strategies for Simple Games Using Meta-interpretive Learning , 2019, New Generation Computing.

[32]  A. Shiryayev.  On Tables of Random Numbers , 1993 .

[33]  J. Ross Quinlan,et al.  Learning Efficient Classification Procedures and Their Application to Chess End Games , 1983 .

[34]  H. Klausmeier,et al.  Relationship of selected cognitive abilities to concept attainment and information processing. , 1967, Journal of educational psychology.

[35]  Andrew Cropper.  Learning efficient logic programs , 2018, Machine Learning.

[36]  Jaime G. Carbonell,et al.  Derivational analogy: a theory of reconstructive problem solving and expertise acquisition , 1993 .

[37]  De,et al.  Relational Reinforcement Learning , 2001, Encyclopedia of Machine Learning and Data Mining.

[38]  K. Holyoak,et al.  Mathematical problem solving by analogy. , 1991, Journal of experimental psychology. Learning, memory, and cognition.

[39]  Herbert A Simon,et al.  The understanding process: Problem isomorphs , 1976, Cognitive Psychology.

[40]  Noam Chomsky,et al.  वाक्यविन्यास का सैद्धान्तिक पक्ष = Aspects of the theory of syntax , 1965 .

[41]  M A Just,et al.  From the SelectedWorks of Marcel Adam Just 1990 What one intelligence test measures : A theoretical account of the processing in the Raven Progressive Matrices Test , 2016 .

[42]  Demis Hassabis,et al.  Mastering the game of Go with deep neural networks and tree search , 2016, Nature.

[43]  José Hernández-Orallo,et al.  The teaching size: computable teachers and learners for universal languages , 2019, Machine Learning.

[44]  Shane Legg,et al.  Human-level control through deep reinforcement learning , 2015, Nature.

[45]  Xiaojin Zhu,et al.  Machine Teaching: An Inverse Problem to Machine Learning and an Approach Toward Optimal Education , 2015, AAAI.

[46]  T. Niblett,et al.  AUTOMATIC INDUCTION OF CLASSIFICATION RULES FOR A CHESS ENDGAME , 1982 .

[47]  Shie Mannor,et al.  Graying the black box: Understanding DQNs , 2016, ICML.

[48]  K. Holyoak,et al.  Surface and structural similarity in analogical transfer , 1987, Memory & cognition.

[49]  Dedre Gentner,et al.  ANALOGICAL REMINDING: A GOOD MATCH IS HARD TO FIND. , 1985 .

[50]  Thomas L. Griffiths,et al.  Faster Teaching via POMDP Planning , 2016, Cogn. Sci..

[51]  Donald Michie.  Experiments on the Mechanization of Game-Learning Part I. Characterization of the Model and its parameters , 1963, Comput. J..

[52]  Michael Kearns,et al.  On the complexity of teaching , 1991, COLT '91.

[53]  Ivan Bratko,et al.  Reconstructing Human Skill with Machine Learning , 1994, ECAI.

[54]  Richard Reviewer-Granger.  Unified Theories of Cognition , 1991, Journal of Cognitive Neuroscience.

[55]  P. Johnson-Laird,et al.  Mental Models: Towards a Cognitive Science of Language, Inference, and Consciousness , 1985 .

[56]  Stephen Muggleton,et al.  Meta-interpretive learning: application to grammatical inference , 2013, Machine Learning.

[57]  John N. Williams,et al.  Performance and Competence in Second Language Acquisition , 1996 .