Synthese 198 (10):1–32 (2020)
AbstractWe propose a formal framework for interpretable machine learning. Combining elements from statistical learning, causal interventionism, and decision theory, we design an idealised explanation game in which players collaborate to find the best explanation for a given algorithmic prediction. Through an iterative procedure of questions and answers, the players establish a three-dimensional Pareto frontier that describes the optimal trade-offs between explanatory accuracy, simplicity, and relevance. Multiple rounds are played at different levels of abstraction, allowing the players to explore overlapping causal patterns of variable granularity and scope. We characterise the conditions under which such a game is almost surely guaranteed to converge on a optimal explanation surface in polynomial time, and highlight obstacles that will tend to prevent the players from advancing beyond certain explanatory thresholds. The game serves a descriptive and a normative function, establishing a conceptual space in which to analyse and compare existing proposals, as well as design new and improved solutions.
Similar books and articles
Learning to Signal: Analysis of a Micro-Level Reinforcement Model.Brian Skyrms, Raffaele Argiento, Robin Pemantle & and Stanislav Volkov - manuscript
Mindreading and Endogenous Beliefs in Games.Lauren Larrouy & Guilhem Lecouteux - 2017 - Journal of Economic Methodology 24 (3):318-343.
Exploring Sociality and Engagement in Play Through Game-Control Distribution.Marco C. Rozendaal, Bram A. L. Braat & Stephan A. G. Wensveen - 2010 - AI and Society 25 (2):193-201.
On Salience and Signaling in Sender–Receiver Games: Partial Pooling, Learning, and Focal Points.Travis LaCroix - 2020 - Synthese 197 (4):1725-1747.
Information-Driven Coordination: Experimental Results with Heterogeneous Individuals. [REVIEW]Viktoriya Semeshenko, Alexis Garapin, Bernard Ruffieux & Mirta B. Gordon - 2010 - Theory and Decision 69 (1):119-142.
Solving the Black Box Problem: A Normative Framework for Explainable Artificial Intelligence.Carlos Zednik - 2019 - Philosophy and Technology 34 (2):265-288.
Undecidability in the Imitation Game.Y. Sato & T. Ikegami - 2004 - Minds and Machines 14 (2):133-43.
On the Narrow Epistemology of Game Theoretic Agents.Boudewijn de Bruin - 2009 - In Ondrej Majer, Ahti-Veikko Pietarinen & Tero Tulenheimo (eds.), Games: Unifying Logic, Language, and Philosophy. Springer.
On the Explanatory Depth and Pragmatic Value of Coarse-Grained, Probabilistic, Causal Explanations.David Kinney - 2018 - Philosophy of Science (1):145-167.
The Pragmatic Turn in Explainable Artificial Intelligence (XAI).Andrés Páez - 2019 - Minds and Machines 29 (3):441-459.
Solving the Black Box Problem: A Normative Framework for Explainable Artificial Intelligence.Carlos Zednik - 2021 - Philosophy and Technology 34 (2):265-288.
Added to PP
Historical graph of downloads
Citations of this work
The Ethics of Algorithms: Key Problems and Solutions.Andreas Tsamados, Nikita Aggarwal, Josh Cowls, Jessica Morley, Huw Roberts, Mariarosaria Taddeo & Luciano Floridi - 2021 - AI and Society.
The ethics of algorithms: key problems and solutions.Andreas Tsamados, Nikita Aggarwal, Josh Cowls, Jessica Morley, Huw Roberts, Mariarosaria Taddeo & Luciano Floridi - 2022 - AI and Society 37 (1):215-230.
AI and its New Winter: From Myths to Realities.Luciano Floridi - 2020 - Philosophy and Technology 33 (1):1-3.
What is Interpretability?Adrian Erasmus, Tyler D. P. Brunet & Eyal Fisher - 2021 - Philosophy and Technology 34:833–862.
Local Explanations via Necessity and Sufficiency: Unifying Theory and Practice.David S. Watson, Limor Gultchin, Ankur Taly & Luciano Floridi - 2022 - Minds and Machines 32 (1):185-218.
References found in this work
Making Things Happen: A Theory of Causal Explanation.James Woodward - 2003 - Oxford University Press.