Two years ago, I described a view I was developing that science is grounded in an implicit monotheism: On God: Science and Modernity. This essay is a short exposition of a more developed version of this view: A Monotheistic Foundation for the Epistemology of Science.

Two Axioms

The epistemology of science can be formulated as having two axioms:

  • The axiom of an universal objectivity: the universe (everything that can be observed) is objective
  • The axiom of natural induction: natural patterns in some observations are some evidence that those patterns also (will) hold in other observations. (The predicate “natural” here is left as a definition that must be accounted for my particular accounts of natural induction.)

Note that the applicability of Bayesian reasoning requires universal objectivity since Bayesian reasoning relies on the fact that each observation is made in the same universe – even if the observations happen at different times, places, and observers.

The axiom of natural induction is required for any reasoning beyond the purely deductive (i.e. logic and, arguably, mathematics), and the natural restriction ensures that not all possible inductions are considered valid (i.e. they must be natural inductions). The question of which conditions qualify an induction as natural is specified by a particular account of natural induction, which should be resolved by an empirical investigation into how natural induction is actually used (i.e. “natural induction” definitionally refers to whatever we are using when we agree with some inductive arguments and disagree with others).

These two are in fact axioms because they are assumed in the epistemology of science and are not derivable.

  • Universal objectivity is not derivable since, without it, it is consistent to assume that universal subjectivity is false.
  • The axiom of natural induction is not derivable since it makes no deductive claims.
    • Note that natural induction is commonly “derived” invalidly by appealing to itself (i.e. “natural induction is true because it’s made true predictions in the past”), but circular arguments are inconsistent.


Universal objectivity implies that there is exactly one universe, as the universe is everything that can be observed and all observers observe the same universe.

In the universe, natural induction implies that there are patterns that have no further explanation (as they appeal to only induction). (Claim #1) This arises from the fact that there must be finitely many downward levels of explanation in any explanation hierarchy. (Claim #2) Patterns in the base (lowest level) of an explanation hierarchy only appeal to induction. Note that the choice of hierarchy here is irrelevant as long as the choice is consistent. Typically scientists use a roughly standardized hierarchy of explanations, for example in order from higher-to-lower level:

… > Biology > Chemistry > Physics

Biological patterns can be explained in terms of biological, chemical, or physical patterns. Physical patterns can be explained in terms of physical patterns but not in terms of biological patterns.

All this goes to show that the epistemology of science postulates a single objective universe in which basic and unquestionable patterns (which we may be unable to access the truth of) that ultimately account for everything that we observe.

A simple way of describing this set of beliefs is that the epistemology of science postulates some inaccessible source that accounts for why there are the basic patterns in the universe that there are.

TODO: Make the end of this section more clear and better argued, right now its a little incoherent

This inaccessible source is outside of the universe, as it is postulated as unobservable (otherwise induction would not be an axiom). The scientific project is to describe patterns at different levels of abstraction above the base level, which ultimately is the standard by which the accuracy of all of these descriptions is measured. In other words, everything that science can/attempts to describe – that is, everything is the universe – is accounted for by appeal to that inaccessible source. I call this source God.

This God may not have many of the aspects of the God of many religions but, certainly in common with those ideas, this idea of God describes Him as omnipotent and omniscient. I do not attempt to take a stance here on the omnibenevolence of this God, but I think it can be argued.

If you don’t want to call this source God, that’s fine. But I think “God” is a fitting name for such an extra-universal entity that is the basis for understanding reality.


Proof of claim #1. Cyclical hierarchies of explanations are inconsistent because an explanation cannot be in terms of itself. Hierarchies of explanations that, given a particular level, have an infinite depth are viscous regresses because some explanations at this level can only be accounted for after infinitely many steps. So, the only kind of hierarchy of explanations left is of finite hierarchies.

Proof of claim #2. Induction is required to establish non-deductive facts and there are no lower layers to appeal to in the explanation, and induction is an axiom and so as an explanation for these patterns it is not derivable.