Reasoning with Plausible Causal Knowledge

Abstract

This dissertation initiates a new direction in causal non-monotonic reasoning. We begin by reviewing commonly proposed and adopted principles of causal non-monotonic reasoning, and determine their suitability for characterising causal non-monotonic reasoning in a general sense. From this review, a significant gap in the literature is identified: research in causal non-monotonic reasoning has so far been focused on the re-use of familiar mathematics, and not concerned with the intuition that underpins human causal reasoning. To fill this gap, we propose a new approach to causal non-monotonic reasoning: one that is motivated primarily by the accurate representation of how humans typically reason with cause and effect in day-to-day life. We call this new approach Plausible Causal Reasoning (PCR). We propose some new principles that help characterise PCR, and define a new formalism for causal non-monotonic reasoning. The new formalism is shown to: satisfy each of our new principles; reason correctly with some important examples; and be well-behaved with respect to the standard properties used to determine the acceptability of logical reasoning mechanisms. We also indicate possible avenues of future work that would further develop this new approach to causal non-monotonic reasoning.