Loading 4 Votes - +

What Is The Supernatural


I saw a question today on Slashdot asking about Running Your Own Ghost Investigation appealing to nerds asking them what tools and methods they would use for doing so. Despite the Slashdot forum slamming the guy pretty hard, it did seem an interesting question from a scientific method point of view for a repost here considering all the raging faith debates and my own posting on the Ouija Board experiment.

I see the primary question as first being What Is Supernatural? since that would drive the metric follow-up question What Evidence Would You Require to Change Belief or Disbelief? The keyword change is important here … no need to address whether you do or don’t currently believe. Simply, what type of equipment and procedure could produce the evidence it would take to change your status quo opinion on the truth of the supernatural definition in either direction.

For instance, I’m always bugged by the SyFy show GhostHunters (while I find it entertaining) because more often than not, the captured evidence is fleeting and only exists on one piece of equipment. You never have the EVP time sync’d to both a thermal and IR capture, etc.

Similarly tagged OmniNerd content:

Thread parent sort order:
Thread verbosity:

The keyword change is important here … no need to address whether you do or don’t currently believe.

I don’t think that’s right. The proper way to update one’s beliefs in light of new evidence is given by Bayes’ Theorem. Ultimately, evidence increases or decreases one’s level of belief, but accepting or rejecting a hypothesis requires that the level cross some threshold.

If your degree of belief, prior to observing the evidence, is extremely low, then it requires a great deal more evidence than if your prior degree of belief is higher. If you’re fairly certain there are no such things as ghosts (say, P(ghosts) < 0.05) you need lots more evidence than somebody who’s on the fence (say, P(ghosts) ≅ 0.5).

We can get even more concrete. Let O(x) = P(x) / P(¬x) — the “odds” of x being true, and let e(x) = 10log10(O(x)) — the “evidence for x” (measured in decibels). Bayes’ Theorem then says:

  • e(ghosts|evidence) = e(ghosts) + 10log10[P(evidence|ghosts) / P(evidence|¬ghosts)]

The second term on the right hand side is the strength of the evidence, in decibels. According to Jaynes, one decibel of evidence is about the smallest difference a person can perceive. Three decibels roughly doubles the probability of P(ghosts|evidence), and ten decibels is a tenfold increase.

At zero evidence, P(ghosts) and P(¬ghosts) are equally likely, with probability 0.5. There’s some level of evidence — that is, some (positive) level of e(ghosts|evidence) at which it becomes reasonable to accept the hypothesis, and some (negative) level of evidence at which it becomes reasonable to reject the hypothesis. Let’s say P(x) < 1/20 and we reject, and P(x) > 19/20 to accept, which correspond to roughly e(x) < -12.8 and e(x) > 12.8

So, if your evidence is only three or four decibels strong, it could (maybe) move a non-believer into the “agnostic” zone, and it could (maybe) move an agnostic into the believer zone, but it can’t turn a non-believer into a believer. It all depends on where you start.

You’d need at least 25dB of evidence to convert a non-believer into a believer. That means P(evidence|ghosts) has to be over 300 times more likely than P(evidence|¬ghosts), so even if there were some piece of evidence that we were certain we’d see given the existence of ghosts (I’m not aware of any), it would still have to be something that had a less than 0.0032 probability of occurring if ghosts didn’t exist.

The kinds of evidence typically on offer — EVP, thermal noise, and so on, are simply not that unlikely, so they’re unconvincing.

0 Votes  - +
Your first question by scottb

I came across this blog post a year or so ago in which Richard Carrier offers some, I think, very cogent observations on the meanings of the terms “natural”, “supernatural”, and “paranormal”.

Share & Socialize

What is OmniNerd?

Omninerd_icon Welcome! OmniNerd's content is generated by nerds like you. Learn more.

Voting Booth

Can Trump make America great again?

14 votes, 1 comment