This article is a little long, so I’ve tried to make it easier to scroll through by breaking it up into sections. Here is the table of contents:
I promise I’ll try to make it entertaining. Let’s dig in.
I. Null Hypothesis and Satanism
A. The appeal to science
The fifth tenet of The Satanic Temple’s Seven Fundamental Tenets is:
Beliefs should conform to one’s best scientific understanding of the world. One should take care never to distort scientific facts to fit one’s beliefs.
I love how meticulous the phrasing is here! The tenet isn’t that we should “believe science” or “agree with scientists”. As nouns, “science” and “scientist” don’t appear at all. The focus of this tenet is the word scientific. It is a modifier: it is a way of understanding the world, and a way of coming to conclusions about facts. The appeal to science isn’t an appeal to a “what”; it is an appeal to “how”.
This resonates deeply with the elements of Satanism I am most passionate about. I love Satanism because I see it as explicitly and overtly adaptive. This makes it different from every theistic religious tradition I know. The core is not a set of beliefs about the world, but a set of beliefs about how to gain knowledge about the world. Implicit in that is an assumption: knowledge itself might grow and change.
The term “null hypothesis” started out as a bit of technical jargon from experimental science and inferential statistics. Over time its use has grown. This isn’t uncommon. A very specific term in academia gets shared around in popular culture and used by people who only partially understand its context or meaning. As a result, the non-technical “common usage” meaning of the term morphs and twists. Eventually, even the professionals (in this case: professional scientists) end up using the term in the non-technical way. This can make things very confusing.
B. Atheists, please stop saying this.
This makes me cringe the most: “Atheism is the null hypothesis.”
No.
No, no.
No. The entire rest of this article will literally be about why this is wrong. Before I get into that, however, I want to say to any atheists in the audience who have used this phrase: I get it.
I know what you mean, I know where you’re coming from, and I know why you are saying it. That statement (“atheism is the null hypothesis”) is a step the follows after a longer chain of prior conversation. That prior conversation may not appear in every single Twitter thread or comments section, but it is part of the larger discourse in our culture between theists and atheists. That conversation goes like this:
Atheist: "I don't believe in God because I don't see any evidence for God."
Theist: "Why do you need evidence?"
Atheist: "I like to justify my beliefs using scientific method. I need evidence to belief a hypothesis."
Theist: "Well, isn't it also a hypothesis to not believe in God?"
Atheist: "You're the one making the assertion God exists. That's the hypothesis. The negation of that is called the null hypothesis. The null hypothesis is what you should believe by default."
Theist: "Well, why can't the null hypothesis be that God exists?"
In this kind of dialogue, calling atheism the “null hypothesis” is shorthand for a lot of things. It’s a quick way of expressing some very complex ideas about why evidence isn’t required to not believe in the very complex concept of a (usually, specifically Abrahamic) God.
I have no problem with people using verbal shorthand in casual conversation…. normally.
In this case, however, I think the shorthand creates confusion. It perpetuates misunderstandings of what “null hypothesis” means.
Then these misunderstandings end up everywhere.
You get think-pieces that say things like:
In my graduate studies, I learned that every time you formed a hypothesis (God is), you were also required to develop a null hypothesis that says the opposite of your hypothesis (God isn’t). Keep in mind that there are no “facts” in science, but rather hypotheses (educated guesses) and theories (hypotheses that have been supported by science, but that may ultimately be disproved). Now, I’m not a scientist, but it makes perfect sense within this model to have the “null hypothesis” that God doesn’t exist. However, to leap from that to certitude of God’s non-existence is to violate the principles of the scientific method, isn’t it?”
And comments on discussion boards that say things like:
The real distinction between the null hypothesis and the alternative hypothesis is that the null hypothesis is assumed true until rejected due to contradictory data. If the null hypothesis is rejected, and the alternative hypothesis is not, the alternative hypothesis can become the next accepted theory, which becomes the null hypothesis for future tests. It’s sort of a “king of the hill” approach, with the current champion serving as the null hypothesis taking on challengers.
Comments like these make me die a little inside every time I read them. They are completely wrong! But it’s difficult to explain in a simple way exactly why they are wrong. To really understand the idea of the null hypothesis, we have to understand more about where it comes from: scientific method, and specifically experimental hypothesis-testing.
II. Three Ideas in Scientific Method
Opening Disclaimer: I’m not going to cover all of scientific method in this blog article. I expect that’s obvious, but it’s 2020 and nothing is obvious so here we are. The topic is huge: there are books and courses and fields of study dedicated to it. For now, I’m going to focus on three concepts that I think are critical for understanding the technical meaning of “null hypothesis”. This discussion won’t be complete, but I hope it will give you enough information so that you can use the idea of “null hypothesis” correctly in your own life, and talk about it intelligently in conversations about theism and atheism.
A. Setting the stage
Let’s start with the important distinction between three ideas: a hypothesis, a prediction, and a measurement. (I’m going to avoid using the word “theory” in this article, because that’s a whole other ball of worms.)
A hypothesis is a statement about some relationship in the world. It can be abstract or concrete. It can be a causal relationship, or just a systematic pattern in the universe. Some examples of good hypotheses include:
- Smoking causes cancer
- The speed of light is constant regardless of the speed you are traveling
- Thinking religious thoughts makes people more relaxed
- Cats are smarter than dogs
Now hang on a minute! When I say “good hypothesis” I don’t mean that it’s correct, or even that it’s interesting. “Cats are smarter than dogs” may be false (it’s not). But it’s a statement about a relationship between things in the universe. As such, it can lead to a prediction.
A prediction is a statement about a measurable relationship between two variables. For something to be a good prediction, it must be about variables that can be measured.
A measurement is the data that you actually record. The measurement gives you the numbers, which you then use to check your prediction.
If your hypothesis is “smoking causes cancer”, one possible prediction could be: “If you take two groups of people who are identical except that group A smokes and group B doesn’t, and you track them from age 20 to age 50, you will see more occurrences of cancer in group A than B.”
You can go out and measure that. Of course, it can still be complicated. For example, what precisely does it mean for two groups of people to be “identical”? How can we tell if the specific people we have chosen for our group A and group B are similar enough to qualify for saying the groups are identical? This is part of why experimental design can be tricky, and why it’s so important for any experiment to be done over and over again by different people.
Nonetheless, your hypothesis is a statement about a relationship between things (“smoking causes cancer”), and you have a prediction about a relationship between two variables (“rate of cancer in group A will be higher than rate of cancer in group B”). If you are able to measure that predicted relationship between the variables, you have support for the hypothesis. You have evidence for the hypothesis. You have some reason to believe the hypothesis.
Of course, there are other predictions that can be made from the same hypothesis as well. For example, another prediction might be: “The average number of cigarettes a person smokes per week has a positive correlation with the person’s probability of getting cancer.”
Now we’re getting a little more into statistics, and we have to think about things like how correlation is measured and how “probability of cancer” is determined. This can get into the weeds, as the conversation turns to “sampling” and population assumptions. So just like with the previous example, the details of using this prediction to test a hypothesis can be tricky.
However, the basic principle of the reasoning process is the same. Your hypothesis is a statement of a relationship between things. Your prediction is a statement of a relationship between variables. You measure the variables to see if you can find that relationship. If you do, then that is evidence for the hypothesis. In other words, it gives you some reason to believe the hypothesis.
B. Enter, our hero: the null hypothesis
I keep on hammering on the idea that a hypothesis is a relationship.
I’m doing that because, in the context of scientific method and experimental testing, the technical jargon null hypothesis simply means: no relationship.
If your hypothesis is that smoking causes cancer, then the null hypothesis is that there is no relationship between smoking and cancer.
If your hypothesis is that cats are smarter than dogs, then the null hypothesis is that there is no relationship between the intelligence of cats and the intelligence of dogs.
It is crucial to understand what this means in its technical sense. If we breeze too quickly past the technical meaning, it can get misinterpreted in so many ways.
For example, when discussing the hypothesis “cats are smarter than dogs”, the null hypothesis isn’t “dogs are smarter than cats.” The null hypothesis isn’t “there do not exist any cats who are smarter than some dogs.” The null hypothesis isn’t “every cat has the same intelligence as every dog.”
No, no, no.
The null hypothesis is no relationship between overall cat smartness and overall dog smartness.
How can you tell if this “smartness” relationship exists? You come up with some kind of prediction.
For example, you might think: on average, the time it takes for a cat to get out of a maze will be shorter than the time it takes a dog to get out of the same maze. That’s a relationship between two variables. You can measure that! Good prediction.
So, you put a bunch of cats and dogs into a maze, and measure how long it takes them to get out. You’ll hit some interesting details, like: what do you do with the ones who never get out, but just sit around licking themselves until you fetch them?
But once you’ve figured that out, you have a bunch of individual measurements of individual “escape times” for cats and dogs. You can calculate the average escape time for the cats, and the average escape time for the dogs. You have to determine (using statistics) whether there is a reliable difference between these averages.
(Statistical reliability, also called statistical significance, is a whole big thing that we don’t have time for here. Suffice it to say that there are some smart cats and some derpy cats, and there are some smart dogs and some derpy dogs, so you need to make sure that any difference you measure between the average dog time and average cat time isn’t just because you happened to get a bunch of derpy dogs or derpy cats by accident. That’s all I’ll say about that here.)
When you measure a statistically reliable difference between your variables, that gives you a reason to believe your hypothesis.
If your measurement does not show you a statistically reliable difference between your variables, then you have not obtained a reason to believe your hypothesis.
This gets phrased as: “You do not have evidence for the hypothesis.”
This is not the same as: “You have evidence that the hypothesis is false.”
But it is the same as: “This experiment hasn’t given you any reason to believe the hypothesis.”
Which is a brilliant segue into the next topic, which is: Why do you need a reason to believe things?
III. Justified Belief
A. Pink Dragons
Credit where credit is due: The following fictional dialogue was inspired by and is based in part on a similar dialogue in the science fiction novel Anathem by Neil Stephenson.
"Why shouldn't I believe that there is a pink dragon that farts nerve gas living on Jupiter?"
"Do you have any evidence that there is a pink dragon that farts nerve gas living on Jupiter?"
"No, but I don't have any evidence that it doesn't exist either. If I don't have evidence either for or against, then why should I not belief in it?"
"What about a blue dragon?"
"I was discussing a pink dragon."
"Yes, but if you have the same amount of evidence for both a blue dragon and a pink dragon, then why wouldn't you believe that there is a blue dragon on Jupiter?"
"Maybe there is. Maybe there are both pink and blue dragons on Jupiter."
"OK, then how about plaid dragons? And striped dragons? Are all of those dragons up there as well?"
"Maybe... I don't have any reason to think otherwise."
"And what if they don't fart nerve gas, but they fart jet fuel instead? Or maybe they fart napalm? And not just dragons... I assume you also believe that there are pink nifflers that fart nerve gas living on Jupiter."
"What's a niffler?"
"Does it matter? You have the same amount of evidence for pink nifflers on Jupiter that you have for pink dragons on Jupiter. So if you believe in the pink dragons, then you certainly have no reason not to also believe in pink nifflers. And of course blue nifflers, and striped nifflers, and transparent nifflers...."
"I suppose they could all be up there..."
"That's a very crowded planet. Do you believe they are all up there?"
"No, I guess I can't say that I do."
"Why not?"
"I don't see why I should."
Everyone has reasons they believe things. Sometimes good reasons, sometimes bad reason. And sometimes the reason is very hard to pin down. Why do you believe that your best friend is your best friend? Why do you believe that your name is what you think it to be? Why do you believe that Antarctica exists?
(I actually used the Antarctica example in an analysis of an Ayn Rand quote that I did back in 2013. There is even a diagram.)
Scientific method, and specifically the process of experimental hypothesis-testing, is a framework for answering the question: “Do I have a reason to believe this?” It’s not the only framework, but it’s a good one.
B. Bad Romance
You use it all the time in your day-to-day life! Well, maybe not exactly. You may not be scribbling down numbers and calculating statistics. But when a new potential romantic partner says “I love you!” (hypothesis), you expect them to behave in a certain way (prediction) and if they do not behave that way (measurement) then your feelings tell you… you do not have reason to believe the hypothesis.
Now, you confront them! You say: “HEY! You sure don’t act like you love me!”
And they say: “You can’t expect to see my love in the way I act! It’s just a feeling I have!”
In the language of scientific method, they have just told you: The hypothesis “I love you” yields no measurable predictions.
Do you believe them?
“You have no reason not to believe me,” they whisper in your ear.
Are you convinced?
Maybe you are. Most people are not. Maybe it depends on how you feel about the consequences of being wrong.
C. Big ball of wibbly-wobbly god
“God exists” isn’t a hypothesis, because it isn’t a statement about a relationship between things. For most people, “God exists” is a collection of hypotheses.
This isn’t particularly about “God”, as it happens. If you say “Antarctica exists” to someone, that automatically unpacks in most people’s minds into a whole set of beliefs that you have. For example, when you say “Antarctica exists” that means you believe:
- The south pole of the earth is covered by land
- That land is cold and covered by ice
- It is also surrounded by water.
- Penguins live there.
- It’s warmer in the winter than in the summer.
And so on and so on. If someone says that they believe that Antarctica exists, but it turns out they don’t believe one or two of these specific things…. that’s fine, they just got some facts wrong about Antarctica.
But if someone says they believe Antarctica exists, but they think that Antarctica is a tropical island off the coast of mainland Florida… then you would tell them: No, you’re not talking about Antarctica. That’s not what Antarctica is.
Believing that “Antarctica exists” isn’t a single belief: it’s an interconnected set of beliefs about Antarctica.
The same is true with God: Believing in God isn’t a single thought, proposition or hypothesis. It is an interconnected set of beliefs about the nature of God and God’s relation to the universe.
And some of those beliefs might be hypotheses. What are some of the common hypotheses included in that big ball of wibbly-wobbly beliefs that people call “God”? They might be things like:
- Asking for God’s forgiveness leads to salvation
- Virtue is rewarded in the afterlife
- Gay sex makes God mad
And so on. These are statements about relationships between things. These are hypotheses. But of course not everyone has the same list of hypotheses in their heads when they use the word “God”. So when you hear someone say “God exists”, you should check which specific hypotheses they are talking about… just to make sure.
Maybe get them to tell you a couple, or even just one. Just one hypothesis they have about God.
Then, see if there are any predictions that can be made from it.
Can you measure “salvation”? Can you get two groups of people who are otherwise identical but where Group A has asked for forgiveness and Group B has not, and at some appropriate later point in time measure the percentage of each group that has achieved “salvation”? Because if you do that and you find that Group A has more people who have achieved salvation than Group B, then you have a reason to believe the hypothesis that “asking for God’s forgiveness leads to salvation”! Yay!
But if you cannot measure such a difference, then you have no reason to believe the hypothesis. That doesn’t mean it’s wrong. It doesn’t mean there is “proof” that asking for God’s forgiveness won’t lead to salvation. It only means there is no reason to believe it. So instead of the hypothesis (“Asking for God’s forgiveness leads to salvation”), you end up with the null hypothesis (“There is no relationship between asking for God’s forgiveness and salvation”).
Let’s try again.
Is there a way you can measure someone’s virtue? Is there a way you can measure a person’s level of reward in the afterlife? If so, you’re all set! Conduct a study: measure the level of virtue of a random sample of people who are about to die, then after they die measure their level of afterlife reward, and see if there is a statistically reliable correlation between the two!
Hmmm. You can’t measure that? Shucks. Well, that’s fine. The fact that you can’t measure those things doesn’t mean the hypothesis is wrong. It just means you don’t have any reason to believe it. Instead of believing the hypothesis (“virtue is rewarded in the afterlife”), you end up with the null hypothesis (“there is no relationship between virtue and reward in the afterlife”).
And why can’t you just decide to believe the hypothesis anyway?
Well… there just isn’t enough room on Jupiter for all of those dragons.
Thank you for reading. Comment, like, subscribe.