Dividing logically into two parts, Causation, Backward and Forward presents an original approach to reasoning about causation.
The primary unit, which makes up the bulk of the book, is concerned with the automation of causal reasoning, and describes a deep-reasoning, domain- independent causal model based upon regularity and causal counterfactual necessity. A reasoning device equipped with this particular causal model can make sophisticated discriminations that are too challenging for most currently documented causal AI models. For example, a properly equipped device using defeasible causal determinations could contend with single-event causation, time-varying causal structure, cause/effect interchange, accidental universals, fields (standing conditions), branched causal patterns, and epiphenomena.
Causal model implementations assemble causal information on-line, then construct (rather than select) solutions. Accordingly, they exhibit deeper reasoning than do causal nets, and display a concomitant greater capacity to assess situations not explicitly anticipated. Causal model implementations retrieve implicit causal information from a database of hypotheticals (DBH), possibly constructed from functional flow diagrams or associated failure space analyses. Reconstructing such a DBH to accommodate system configuration changes is often faster than redesigning an equivalent set of shallow rules. Locality considerations, though not intrinsic to the model, can guide a standard procedure for pruning the search space used in truth-evaluating counterfactuals.
Causal model implementations are maximally useful when a system’s causal structure itself varies with time. In such cases causal model implementations permit, in effect, the generation at an epoch of relevant fragments of a causal net with dynamic structure. Several example paper implementations are presented in detail in Causation, Backward and Forward’s Appendix B.
The second unit seeks to demonstrate that logical fallacies dominate the philosophical literature on (1) causation under time reversal and (2) backward causation. A quest for a deeper understanding of backward causation is warranted, given key advances in physics which suggest the possibility of retroactive influences.
These advances include Dirac’s approach to electrodynamics, developing theories of quantum gravity and wormholes, Cramer’s transactional formulation of generalized Wheeler-Feynman absorber theory, and Lorentz-invariant theories of tachyons.
The second unit begins by showing that backward causation is ubiquitous in reverse time provided it manifests itself in a certain manner profoundly different from standard descriptions. The new formulation of backward causation (1) withstands bilking paradoxes, and (2) accounts for the fact that, in forward time, causes later than their effects virtually defy detection.
A first attempted application of revisionist backward causation suggests a resolution of the celebrated quantum mechanical EPR problem (for one typical experimental arrangement) that (1) seems to render intuitively transparent violation of the Bell inequality in just the quantitative manner predicted by quantum mechanics and (2) eliminates the need to contemplate instantaneous action at a distance. Another application serves to explain the pattern of concentric rings in electron diffraction experiments in a manner suggesting that backward causation, properly analyzed, is central to conceptual resolution of wave/particle duality!
The closing argument proffers the idea that the book’s backward causation and EPR treatments, by highlighting logical and empirical evidence for the nonmonotonicity of time, should stimulate further research in theories of knowledge and belief. Therefore, as a parting bonus, a preliminary framework is offered for an appropriately structured new theory of knowledge.