Defining and Debating Algorithmic Causality

“Algorithmic causality” encapsulates computational, formal, and epistemic aspects of implementing causal learning into computer systems. This chapter delves into the use of algorithms for understanding and learning causal relationships, which are crucial in artificial intelligence (AI) and machine learning. Drawing from Bishop’s theoretical framework, the chapter explores debates around “reasoning by association” vs. “causal reasoning” in AI. It discusses advancements in causal inference, Bayesian networks, and graphical models to deepen AI’s understanding of causality. The chapter also addresses debates on simulating the human mind in AI, touching on Strong AI and the historical evolution of artificial neural networks (ANNs). It highlights the limitations and advantages of ANNs, emphasizing the need for a comprehensive approach in AI development. The discussion extends to unsupervised learning, neural network vulnerability to adversarial examples, and the challenge of operating within non-Euclidean spaces. The latter part of the chapter addresses criticisms of AI and robotics, categorizing them into anti-technological attitudes and humanist views. Critiques based on fear of unpredictable machines (“Frankenbots”) and concerns about job displacement are countered with ethical guidelines and workforce adaptation. Humanist criticisms questioning AI’s ability to emulate human creativity and consciousness are addressed by acknowledging AI’s unique contributions without aiming for direct emulation.