Back in April 2017, famed atheist and freethinker Michael Shermer wrote an entry for Scientific American entitled, “What Would It Take to Prove the Resurrection?” In it he examines the lengths required to ‘prove’ that Jesus’ Resurrection took place as a matter of history.

Before we begin, it’s important to note that “proof” is the wrong term to use in this context. We can’t *prove* with statistical or mathematical certainty that Abraham Lincoln even existed. We don’t have irrefutable evidence of his existence. *Proof* is therefore not what we’re interested in when it comes to historical inquiry.

## Valid Knowledge?

At one point in the article Shermer says this:

Interestingly, the term “valid knowledge” doesn’t appear in the Stanford Encyclopedia of Philosophy entry that introduces key concepts in Epistemology (see here). This tells us that Shermer is using idiosyncratic, and therefore unhelpful, terminology. Moreover, it’s not clear the concept of “invalid knowledge” is even coherent; a*ll* knowledge (if it’s actually knowledge) is *valid*.

This issue aside, here’s what I think Shermer is trying to say: The proposition that Jesus died for our sins (a) lacks sufficient evidence (is “faith-based”), and (b) lacks *warrant*. We’ll take a closer look at both of these claims.

## Principle of Proportionality?

In defense of claim (a), Shermer says the following:

Here again we have some very odd and peculiar terminology. As it happens, there is no “principle of proportionality” that demands extraordinary evidence for extraordinary claims. What’s really going on here is that Shermer is proffering the all-too-familiar atheistic slogan ‘extraordinary claims require extraordinary evidence.’ For a good response to this, check out this lecture by my friend Tim McGrew; it’s one of the best I’ve seen.

Shermer goes on to say:

What this translates to, in the language of probability, is this: “Given that billions of people haven’t risen from the dead, the probability that Jesus rose is extremely low.” [1] Does the consequent follow from the antecedent? Actually, it doesn’t. That is, it doesn’t, *unless* we assume a frequentist interpretation of probability.

## Frequentism

Here’s a simple version of frequentism:

*The probability of an attribute A in a finite reference class B is the relative frequency of actual occurrences of A within B.*

Basically the idea behind Shermer’s argument is this. Jesus is part of the reference class called humans. Of that class, no one has come back from the dead (so says Shermer). Since there are no other occurrences of Resurrection within the reference class of humans, it follows that the probability that one of them did is extremely close to zero [1]. This is how the argument works. And it’s clear it only works given *frequentism*.

Here’s the problem: Even if we grant that no one else in human history has come back from the dead, frequentism itself is not the only interpretation of probability theory. This highlights the fact that arguments are only as good as the assumptions they make. Challenge the assumptions and the argument falls apart.

We can take this a step further and show that, even if we assume frequentism, it doesn’t necessarily follow that the prior probability of Jesus coming back from the dead is low. This is due to the so-called “reference class problem” for frequentism.

## Reference Class Problem

On frequentism, we assign probabilities by looking at how many occurrences of an event happen within a given reference class. The problem arises when we attempt to *define* our reference class. The more detailed we get, the less trials we find.

Shermer uses “humans” as his reference class. He says that since no *other* human has ever risen from the dead, it’s highly unlikely Jesus did. But what happens when we get more specific? For instance, what happens when we ask: How many Jewish carpenters named Jesus that were killed under the reign of Pontius Pilot by crucifixion stayed dead in the first century? On frequentism, since there’s only 1 trial in this reference class, this probability is either 0 or 1.

The reason this is such a problem is that *no reference class is any more valid than any other*. Thus, on frequentism, probabilities are rendered *relative*. This is all very counterintuitive. Intuitively, including more information in our analysis should render a more accurate probability; it shouldn’t result in relativization.

## God & Freedom

If that weren’t enough, there’s another problem lurking nearby for frequentism. Namely, the Resurrection of Jesus involves the free action of God. On the Resurrection hypothesis, God *freely* raises Jesus from the dead. A frequency interpretation of the probabilities involved doesn’t seem appropriate. Consider the following scenario:

Mark, a 30 year old, *loves* Twix. He loves Twix so much that since turning 18, he’s eaten a Twix bar every single day (luckily for Mark, he has the metabolism of a hummingbird). This means he’s chosen a Twix bar over, say, a Snickers bar, 4,380 times. Frequentism says that since it’s never happened in the past, the probability Mark freely chooses to eat a Snickers bar for his next snack is 1/4380. But suppose we learn some new information: Mark recently fell head-over-heels for a girl that loves *Snickers*. Is the probability the same or has it changed? On frequentism, the probability doesn’t change at all. But intuitively we know that the chances Mark buys a Snickers is at least moderate.

The scenario above illustrates the point that frequencies don’t really work with the actions of free agents. I can be a hardcore carnivore all my life and then suddenly choose to become an insufferable vegan (no offense to my vegan friends).

Shermer’s claim that the prior probability of Jesus coming back from the dead is low turns out to be highly dubious in at least 3 respects.

## Warrant

What about Shermer’s second claim? We’ve already seen that Shermer’s first claim is unsupported, but what about his second claim that Christian belief lacks warrant? Here again it appears that Shermer is assuming that warrant requires evidential or probabilistic justification. This is an undefended assumption.

The Christian philosopher Alvin Plantinga has argued at length that Christian belief needn’t be supported by argument and evidence if it is to be warranted. This is because, if Christianity is true, beliefs about the truth of Christianity are not brought about by way of any natural process. They involve the work of the Holy Spirit. Given this, if Christianity is true, Christian belief will meet the conditions of warrant: They are produced by a belief producing process that is functioning properly in an appropriate environment according to a design plan successfully aimed at the production of true beliefs (for more on this, check out my 4 part series).

When Shermer says that Christian belief lacks warrant, he’s actually making the claim that Christianity is false. But he’s far from demonstrated that in this brief little article.

## Conclusion

While much more could be said, it is clear that Shermer’s claims include a number of dubious assumptions. Pithy one-liners are no substitute for philosophical rigor.

[1] This can be cashed out in terms of prior probabilities. The same concerns apply.