So the only explanation possible is...

(An improved and updated version of this essay has appeared in my book Brief Peeks Beyond. The version below is kept for legacy purposes.)

The Sleep of Reason Produces Monsters (etching by Goya, c. 1799). Source: Wikipedia.

In logic, a strong distinction is made between deductive and inductive inferences. Here is an example of a deductive inference:

Amsterdam is the capital of the Netherlands. Therefore, if I go to Amsterdam, I must go to the Netherlands.

Clearly, a deductive inference is necessarily implied by its premise, beyond any doubt. Now consider the following inductive inference:

My house has been broken into and there are unidentified footprints in the backyard. Therefore, the footprints were left by the burglar.

Now the inference cannot be derived with certainty from the premises. There is only a reasonable probability, given the circumstances, that the footprints were made by the burglar. Indeed, they could conceivably have nothing whatsoever to do with the burglary; they could have been made, for instance, by the gardener who came in to collect some forgotten tools while you were not home.

Inductive inferences are entirely dependent on our ability to correctly evaluate probabilities. However, probabilities are notoriously tricky to evaluate without the benefit of statistics based on past empirical observations of analogous situations. For instance, consider this hypothetical situation:

For the past 10 years, 90% of the times the postman came to my house after I was already awake. Therefore, I inductively infer that on Monday the postman will come after I wake up.

Here the probabilities are easy to estimate based on past empirical observations of analogous situations: 10 years of it, to be precise. These previous, empirical observations of the arrival of the postman form a so-called 'reference class' of earlier occurrences. The probability of the inductive inference can then be calculated based on this reference class (in this case, 90% probability that the inference is correct). But what about cases when no proper reference class is available? For instance:

Vicky returned from clinical death claiming to have seen the doctors working on her body as if she stood outside of it. Therefore, Vicky’s story is a post-event confabulation based on earlier memories.

But wait; how many times have similar stories, told in analogous situations in the past, been known to be confabulations? Here is another:

George saw a luminous object in the sky performing maneuvers impossible for any known aircraft. Therefore, George saw an alien spaceship.

How many times have similar observations in the past been known to be caused by spaceships from another planet? A final example:

The fundamental laws of nature have been the same across space since the Big Bang. Period.

Now, where are the reference classes in these cases? There aren’t any. Our estimate of probabilities here is not based on objective statistics of previous empirical observations. Instead, and this is a key point, it is subjective; it is based solely on our paradigm – a set of subjective values, assumptions, and beliefs that inform us of what should be possible or likely. According to this paradigm, consciousness is a by-product of brain activity, so Vicky could only have confabulated her story. According to this paradigm, we already catalogued every observation that can conceivably be produced by the dynamics of our earthly reality, so George could only have seen a spaceship from another planet. And finally, if the laws of nature were changing over time our entire scientific edifice would be foundationless, so they could only have stayed the same.

In all these cases, the form of the thought is this: 'Since all other alternatives allowed by the paradigm can be discarded, then the only alternative left must be true.' In other words, we extract conclusions by elimination of alternatives. The problem here is that, to infer conclusions by elimination, we must know the boundaries of reality. In other words, we must assume that our paradigm is complete; that there is no yet-unknown aspect or facet of reality lying outside our current paradigm. This is a supremely arrogant, naïve, and dangerous assumption on the face of it; one that history shows to be more-than-likely wrong (for this latter inference we do have a solid reference class!). You see, we don’t know what consciousness is or where it comes from, so discounting that it can exist independent of brain activity is precipitated at best. We don’t know all the parameters and dynamics of our earthly reality, so postulating a non-earthly agency to explain certain bizarre observations is hastened. And finally, we just cannot know whether the laws of physics have been the same since the Big Bang; yes, we have models based on this assumption that seem to explain reality, but that’s inverting the argument: these models were built so that they would make sense of the assumption to begin with.

Now here is the problem: a very significant portion of our worldviews, even the most hard-nosed scientific ones, is based just on this type of inductive inferences unsupported by a proper reference class. In fact, science itself is based on this kind of inductive inferences: after all, they are the only way to claim that the same laws and dynamics empirically observed under laboratory conditions apply to reality at large, over time and across space.

Inductive inferences motivated only by paradigms, instead of empirically-derived reference classes, lead to worldviews that are at least as much a reflection of our own thoughts (and limitations of thought) as they are a reflection of a supposedly objective nature. We live in a reality largely defined by a paradigm – a set of beliefs – as opposed to objective, empirical facts. This may reflect a level of unconscious closed-mindedness and sheer naïveté that one day may profoundly surprise us.