This quote is from John Loftus’ recent post about his counter-apologetical tactics. Here, he writes a short version of an argument he presented in his longer book Why I Became An Atheist: A Former Preacher Rejects Christianity:
As a former student of James D. Strauss at Lincoln Christian Seminary in Lincoln, Illinois, I credit much of my approach to Christianity to three things that Strauss drilled into us as students, but in reverse. When doing apologetics, he said, “if you don’t start with God, you’ll never get to God.” Strauss is not a Van Tillian presuppositionalist because he doesn’t start with the Bible as God’s revelation, but he does start “from above” by presupposing that God exists and then argues that God’s existence makes better sense of the Bible and the world than the alternatives. Again, “if you don’t start with God, you’ll never get to God.” Since this is such an important, central issue, I’ll focus on why we should not start “from above” with a belief in God, but rather “from below” beginning with the world in which we find ourselves. If successful, my argument should lead us to reject the existence of the sort of God thought to confirm the biblical revelation.
(BTW, you should get the book. I enjoyed it)
Since I’m trying to think more like a Bayesian, I thought I would attempt to formulate this in probability theory.
At first glance, it seems like a fallacious argument; it reminds me of the Prosecutor’s Fallacy. Of course, whenever I see a statement similar to “makes better sense of the [evidence] than the alternatives” it sets off little alarms in my head. What does this person mean by makes better sense than the alternatives? Are they only going by P(E | H), the success rate? Or are they utilizing the entirety of Bayes, plugging in not only the success rate, but the base rate and false positive rate? To see where my reasoning is going, I’ll copy a bit of the above quote but substitute “winning the lottery” with the Bible and the world and “cheating” with “God”:
“if you don’t start with [cheating], you’ll never get to [cheating].” Strauss … start[s] “from above” by presupposing that [you’re a cheater] and then argues that [you being a cheater] makes better sense of [winning the lottery] than the alternatives. Again, “if you don’t start with [cheating], you’ll never get to [cheating].”
This is actually “valid” in a way. It’s not a Prosecutor’s Fallacy because it is explicitly saying that you have to start with the assumption of a high prior probability of cheating to end up with cheating as being the better explanation than the alternatives. Of course, taking reality into account, you can’t start with a high probability of cheating because you need more than the current evidence to support such a high prior. How many times have you seen someone cheat to win the lottery? Exactly. Similarly, what is the prior probability of god’s existence? What is the probability of god’s existence before looking at “the Bible and the world”? Exactly; you can’t use the Bible and the world to argue for the prior, the prior has to be established before that. Duh, that’s why it’s prior probability!
So John Loftus’ prior seems to be more grounded in reality; a more reasonably justified prior. You can’t start “from above” because we have no warrant to. We always have to start “from below” with all of our priors. If we were allowed to start with any old priors we grab from our nether regions, we can skew any argument in our favor. So we should always start at the bottom, from below because that’s where we’re all at.
Where our priors come from is simple induction. How many gods have people believed in that are actually real? How many perpetual motion machines do you know about? (every description of supernatural beings I’ve heard are nothing less than perpetual motion and/or cold fusion machines). These are our priors. The probability we assign before looking at the particular evidence.