This is Jessica. Religion and spirituality are pretty far from what I usually blog about, but recently several writers (who are also researchers) brought it up in essays and I found it refreshing. Rachael Meager wrote about semantic voids that resist labeling or assignment of meaning, referencing concepts from Buddhist thought and their own practice throughout. Michael Nielsen wrote on sacrifice, also touching on a semantic void of sorts in understanding the experience of a young monk who self-immolated, and later reflecting on what sacrifice for “an enlarged self” can mean.
This reminds me of how taboo religion, in the sense of believing in some higher organizing power or “intelligence” that gives meaning to life, can seem as a topic among scientific academic types. I would say I have been strongly influenced by religious philosophy, or at least always gravitated to it, and I started regularly practising a variant of Kashmir Shaivism years ago in grad school, but I rarely acknowledge this outside of my closest friends (at least until now, I guess). Reading Meager writing about practising Buddhism, and Neilsen acknowledging his aversion to self-indulgence in even discussing sacrifice got me thinking about how odd this is, that I can devote considerable effort to what might be considered religious matters–e.g., trying to discriminate between desires and intentions are driven by ego versus which are aligned with an organizing principle greater than myself, or reflecting on the nature of that organizing principle–but somehow one I feel I must avoid recognizing this as part of my identity in the majority of my interactions. Being private about these things to some degree is natural, since religion is obviously a personal matter, but I think the aversion also stems in large part from a learned tendency to deny having pre-existing faith in anything. Neuroticism about activating doubt about my rationality runs deep, I suppose.
The role of rationality in attitudes and expressions of religion among technical people is really kind of fascinating. I remember in grad school, I was briefly dating a physicist, and religion came up at some point in one of our conversations. He mentioned he was an atheist, and I replied by saying something about how it was hard for me to imagine not at least entertaining the possibility of something god-like or absolute. He responded with something about how the beauty of Maxwell’s equations was like a religious experience to him. Which is of course completely valid and not really my place to judge, but I remember thinking at the time about how cliche it seemed, as if only by putting a scientific enough spin on it does it feel safe to acknowledge believing in anything greater than everyday reality. But it really is hard to separate the two. I can’t seem to do it. I find it amusing how part of why I feel unsettled when I find out that someone I otherwise relate to identifies as an atheist is because by asserting that there is no god, rather than that this is uncertain, atheism strikes me as close to a straw man, point null hypothesis, and I have trouble getting past that part.
The pockets of heavily moralistic, often bordering on religious thinking that have become more visible in tech in recent years are also intriguing. Some writers associated with the rationalist community have consistently discussed religious matters, for example, like Scott Alexander mentioning god in his essays. But there’s also the vision quest-y obsession with meaningfulness that’s become associated with Silicon Valley tech elite, often a weird mix of secular self-help and borrowing specific practices from religious traditions. Vaguely theological references sometimes make it into AI discourse, like Rich Sutton’s teleological view of mankind having the honor of ushering in the “fourth great age of the universe”. Some of this seems genuine to me, but most of the latter variety seems distinct from what I think of as religion, more about aesthetics and power than real belief in something. The association between religion and power is epitomized by Sam Altman musing about how “the most successful founders do not set out to create companies. They are on a mission to create something closer to a religion, and at some point it turns out that forming a company is the easiest way to do so.”
I consider myself a pragmatist in many ways, and so it seems natural to wonder whether one can “have” religion or be a religious person as a result of making a conscious choice after recognizing that acting religious can bring benefits. Is this really religious behavior, or is it a kind of secular mimicry? Is it even possible for us to discriminate between the two and to choose? I imagine if I spent time thinking about it I could probably persuade some part of me to believe that randomness and chaos and individual agency are all that exist, but then I wonder why on earth would I want to do that, since it sounds stressful. And if I did succeed in convincing myself, I wonder if I would somehow end up developing a kind of religious philosophy around that perspective, to fill a psychological need to have faith in something. Nielsen posted a relevant passage from Alan Watts on X this week, in which Watts argues that the benefits of a myth can only come when it is thought to be truth, and that for this reason “most of the current return to orthodoxy in some intellectual circles has a rather hollow ring.” I’m not sure exactly what circles he had in mind, but the quasi-religious discussion in certain tech circles often feels to me like something very different than the kind of trust that I associate with being religious. Can you have religion without vulnerability? Is this the necessary ingredient for distinguishing true religious devotion from more ego-driven pursuits?
I wonder if things would be different if we were more comfortable as scientists with acknowledging that having irrational commitments to certain ideas is part of what we do. We perceive this as conflicting with scientific objectivity, but science is full of leaps of faith we make to locate ourselves within various epistemic-philosphical conditions. Religion and spirituality have high surface friction with our preferred identities, but the strength of the aversion can seem almost pathological, driven by deep discomfort with the idea of a priori belief and the possibility of aspects of reality that resist direct perception or inference.
Quote from the blog post: “I wonder if things would be different if we were more comfortable as scientists with acknowledging that having irrational commitments to certain ideas is part of what we do.”
This sentence reminded me of something I recently came across in a book by Cleckley (1941/1988) concerning the psychopathic personality titled “The Mask of Sanity”. In the beginning of the book Cleckley writes the following:
“Few, if any, who prophesy on the grounds of mystic insight or special revelation come to conclusions more extraordinary than those reached by some who profess, and often firmly believe, they are working within the methods of science.” (p. 5).
When reading these words I was reminded of a paper by Bem (2011) which presents experimental evidence for “(…) the anomalous retroactive influence of some future event on an individual’s current responses (…)” (p. 407). As LeBel and Peters (2011) state concerning these remarkable published findings by Bem (2011): “By using accepted standards for experimental, analytic, and data reporting practices, yet arriving at a fantastic conclusion, Bem has put empirical psychologists in a difficult position (…)” (p. 371). You can even combine this all with the papers by John et al. (2012) concerning questionable research practices and their use, and the “false-positive psychology” paper by Simmons et al. (2011), and subsequently underline the Cleckley quote. Or you can quote it for a second time to emphasize it:
“Few, if any, who prophesy on the grounds of mystic insight or special revelation come to conclusions more extraordinary than those reached by some who profess, and often firmly believe, they are working within the methods of science.” (Cleckley, 1941/1988, p. 5).
haha, I like the quote. Could be applied to some of the beliefs seeping into AI discourse, though that’s more engineering than science in the scientific method sense.
It was an interesting book to read, for several reasons, and I am very happy I came across it when writing a certain manuscript. I ended up using several quotes from the book at (in my view) crucial places in the writing, and I think it added something crucial to the writing that I appreciate a lot but find hard to describe. Perhaps it was the final part of the puzzle or something like that, or perhaps it provided some structure in some way. I don’t know which words to use here.
Anyway, I found an online version which was the 1988 “made available for educational purposes” version if I am not mistaken, so it’s available to read somewhere online. Given your remark about AI, and certain discussions on here about AI writing and poetry, I wanted to share the following quote as well which you might appreciate as well:
“Renowned critics and some professors in our best universities reverently acclaim as the superlative expression of genius James Joyce’s Finnegan’s Wake, a 628-page collection of erudite gibberish indistinguishable to most people from the familiar word salad produced by hebephrenic patients on the back wards of any state hopsital.” (p. 7)
Jessica:
You write, “This reminds me of how taboo religion, in the sense of believing in some higher organizing power or “intelligence” that gives meaning to life, can seem as a topic among scientific academic types.”
I’m surprised you say this. I guess that most of the academic scientists I know are atheists, but I’ve also known many academic scientists who are openly religious. So, in my experience, religion and religious belief are not at all “taboo” among scientists. Science and religion mostly cover non-overlapping topics, so scientists might not put their religious beliefs into their scientific papers, but they’ll talk about it informally.
A related topic is supernatural beliefs that are not religious. This has come up a lot on the blog over the years: many scientists believe in ESP, astrology, UFOs as space aliens, the power of subliminal smiley faces to alter attitudes on immigration, and lots of other claimed “spooky action at a distance” phenomena that are contrary to our understanding of science. Lots of the events in the Bible and other religious texts are contrary to our understanding of science too, but belief in religion doesn’t seem to interfere with science in the way that belief in astrology etc. does. There are exceptions, though, such as the literature on intercessory prayer which is just as bad as the literature on social priming etc.
Huh, I wonder if there’s some selection bias in who gravitates to what areas, or at least in who I talk to, because I only seem to meet atheists. Though like I mention in the post, these things also don’t come up enough to really know how many people feel about them.
>Lots of the events in the Bible and other religious texts are contrary to our understanding of science too, but belief in religion doesn’t seem to interfere with science in the way that belief in astrology etc. does.
This is interesting, and something similar crossed my mind while writing the post on how religious beliefs can be easily compartmentalized, as if they aren’t really opposed to being rational and requiring evidence behind beliefs. Whereas something like astrology is much more taboo to mention, and considered abhorrent to scientific sensibilities. It’s like we accept that it is rational on some level to have some perspective on a higher power, so long as you can avoid letting it interfere too much with your model of how the world works. I find this kind of ironic, because most religion is supposed to be about the nature of the world. But I agree with your point, there’s a distinction.
‘[I]t seems natural to wonder whether one can “have” religion or be a religious person as a result of making a conscious choice after recognizing that acting religious can bring benefits. Is this really religious behavior, or is it a kind of secular mimicry?’
There is an implicit assumption within this question. The assumption is that there is a dichotomy between ‘true believers’ and ‘fake believers’. How can we distinguish between ‘true believers’ and ‘fake believers’?
I often find myself referring to ‘true believers’ as people who never question their religious beliefs. Usually, however, I prefer to refer to such people as fundamentalists. I find fundamentalism in research or religious practice very concerning, as it stifles progress. I often tell people at my church that the day they no longer have any doubts is a sad day, as it means they have abandoned rationality — at least with regard to their religious beliefs. However, this can also ‘infect’ other beliefs of a scientific nature, which is even more troubling when it happens.
I like the perspective that questioning is part of rational people having religious beliefs. It also seems related to accepting vulnerability as being part of religious belief, like without perceiving yourself as somehow vulnerable you have no reason to be questioning.
Having a “religious experience” in Maxwell’s Equations automatically conjures up the classic t-shirt:
https://www.zazzle.com/and_god_said_maxwells_equations_unisex_t_shirt-235152434074421469
Haha, maybe it’s the work of that guy I dated!
Thanks for sharing this! I had a few reactions; I don’t know if/how they connect, so I’ll just jot them down in case they are of interest to anyone else.
* It strikes me that many of the issues you discuss with regard to religion, like the construction of a framework for making moral choices or for understanding what it means to live a meaningful life, are things that I think about quite a lot but not within the framework of “religion” per se. For me, what makes something a “religion” is less about the role of faith, meaning, or morality and more about its social organization. I suspect this perspective is, at least in part, due to being gay and growing up in a time/place (not long ago!) when no mainstream religion would tolerate someone like me. When all the religions you see treat your very existence as “immoral”, you pretty quickly learn to look elsewhere for guidance on issues about faith, meaning, or morality.
* Once I felt comfortable enough (or perhaps stupid enough) to identify with any particular belief system, I decided I was “agnostic” and that’s what I’ve stuck with since. I suppose that could be a cop-out, but I think it is actually a facet of my broader perspective for understanding the world. As I note on the first day of my undergrad intro stats classes, I think the best thing about understanding statistics is that it gives you a framework for dealing with uncertainty gracefully.
* I also think that many scientists fail to appreciate how much of what we do and believe is, ultimately, taken on faith. By that I mean that we adopt many methods and perspectives not based on some kind of “rational” argument, but because of things like elegance/beauty (e.g., many theories in physics) or trust in authority (e.g., Bob’s experiences in biology). I think if scientists acknowledged this more, we might end up in fewer pointless debates between theories/methods that are largely equivalent but which are adhered to as, essentially, matters of blind faith.
Yeah, I agree religion conveys social organization to many people. My husband grew up in Oklahoma and Kansas where there was a lot of fundamentalism and also has trouble not associating religion with intolerance (as well as unwelcome attempts at conversion). When I use the term I don’t so much think of social organization as being necessary, but I can see how its difficult to separate.
> entertaining the possibility
Sure, you can consider the possibility. But, then you have to condition on the data.
Thank you for the thoughtful post! Though raised in a Jewish household and still identify as such, I became an atheist when introduced to the Holocaust in religious school at our synagogue. I’ve been one ever since. I look at the Jewish religion as one of inner reflection that does not require a belief in God. As a choral singer, I’ve sung in both church and synagogue choirs that have included several requiem masses including a Jewish one recently composed by Simon Sargon. Obviously, a religious spirit exists for some people as it is the only explanation for the music of Bach or the paintings of the Italian Renaissance. It is possible to enjoy all of these and still be an atheist. Participation in religious community does not require a belief in God.
It’s interesting just how deeply ingrained the cultural concept of God is… I mean if you went back 3000 years ago it’d be all about “gods” ie. a world of beings that have superpowers. Monotheism is soooo ingrained now that we practically define religion as the belief in a single great creator god who is in charge of everything.
Anyway I agree with you that participation in a religious community does not require a belief in God, or Gods, or the supernatural, or anything like that. It probably helps though, just to avoid others treating you weirdly because you confuse or antagonize them.
Quote from above: “It’s interesting just how deeply ingrained the cultural concept of God is… ”
I thought about your comment, and I wanted to say something about how it could be some sort of by-product of having a relatively over-developed brain (or whatever the correct words are here). In this light, it could be some sort of combination of 1) the need or attempt or want to understand and explain, and 2) the need or attempt or want to control and influence things.
The following sentences of your comment might also relate to this general idea in the sense that 3000 years ago it would be about many gods, which if I am remembering correctly, might all be connected to certain specific aspects (?) of life, such as agriculture whatever. Perhaps this could be seen as a more separated and specific case of the need/attempt/want to understand and explain, and the need/attempt/want to control and influence things for all these things the gods are tied to.
Now this general idea can perhaps also be applied to explain certain certain things and developments concerning views about god or gods. For instance, more knowledge about growing crops and more control to influence things related to that may have led to less sacrifices to the sun gods to produce corn (or whatever).
And more knowledge about things in general may have led to several different gods becoming a singular god over time. And in present day that singular god might relate more to larger, and more complicated issues, for which knowledge and control is limited, like the purpose of life, etc. This might also relate to somewhat more attention for god, or spirituality in general, in present day in certain places when explanation or understanding is all of a sudden needed in times of trouble, confusion, search for meaning, etc.
I guess this has all been worded much more optimally, but as a result of your comment I thought about it and tried to formulate the general reasoning as depicted above. I am not sure if it makes any sense, but from my limited time thinking about this phrasing just now I reason with these two things of 1) needing/attempting/wanting to understand and explain, and 2) needing/attempting/wanting control and influence you could possibly connect several things related to spirituality or views about god or gods in different settings or even time periods.
And these two things might simply be a by-product, or result of, a relatively over-developed human brain, combined with a further possibly particular (?) human feature of language and communicating things and knowledge and stories over generations.
Quote from above: “Now this general idea can perhaps also be applied to explain certain certain things and developments concerning views about god or gods.”
I just looked up some stuff about greek gods and came across the following which I thought fits nicely with the gist of the general idea:
“Beyond origin, the most dramatic god vs demi-god differences manifest in what these beings can and cannot do. A god’s power is often conceptual and limitless within their sphere, while a demigod’s power is typically physical and bound by certain rules.”
Perhaps this is an example of what can be tried to fit in the general idea in the sense that the greek god and demigod difference can be tied to things that are less or more understandable and controllable. For relatively more complex and uncontrollable things there are gods and for relatively less complex and more controllable things there are demigods connected to them.
Anyway, that’s enough for now I think.
The term God and the term AI have both been so anthropomorphized and associated with nonsense that it’s hard to use them, but God much worse so. Monotheism and the idea of religion as belief in a single creator god being ingrained in interpretations of what religion rings very true, and makes it hard to talk about religious belief without firing up a bunch of stereotypes. Many (non-Western) religious philosophies are nothing like the single anthropomorphized creator view.
Even Western religious ideas like classical theism and divine simplicity are quite far from an anthropomorphic (vs. anthropomorphized) understanding of God.
“It’s interesting just how deeply ingrained the cultural concept of God is…”
Amusingly, the two main religions here, Shinto and Buddhism, don’t have a “concept of god”. Buddhism doesn’t have any god** and Shinto has an infinite number of them.
Which probably explains why most Japanese don’t “get” Christianity.
Thus Japan’s a great place for atheists. Since it has the lowest percentage of Christians of most countries I’ve ever heard of, you don’t get Christian dogma shoved down your throat 24/7. Pretty much the only bad news here is that a big chunk of the literary world is, not unreasonably, seriously taken with French literature. The problem is that a few of them buy into Catholic angst big time. Uh, guys. The Japanese don’t do Christianity, let alone Catholic angst. I’m here to read about your insights into Japan, not western religions. If I want Catholic angst, I’ll reread Graham Greene.
*: If you take it seriously. But if you listen to what the priests say to their parishoners at funerals, it sounds rather indistinguishable from Catholicism, which I found rather weird. (The priests’ problem is that Buddhism officially doesn’t buy into the afterlife thing. Reincarnation, yes, afterlife, no. But you can’t tell a grieving family that their aunt is flat out gone unless she’s either been really good or really bad, in which case she’ll be back as a goose or a fox or a slug.)
>the two main religions here, Shinto and Buddhism, don’t have a “concept of god”… Thus Japan’s a great place for atheists.
I find this confusing. How is it useful to try to distinguish these as not having a concept of god when talking about atheism? E.g., Shinto involves belief in “spirits”. Traditionally, Buddhists believe in karma and reincarnation. Many other non-Western religions are like this, i.e., there is no “god” in the way you are using the term, but they are clearly not secular philosophies. Is there really such an important distinction betwen theism and these when discussing atheist vs religious views?
> Many other non-Western religions are like this, i.e., there is no “god” in the way you are using the term, but they are clearly not secular philosophies.
Religions are non-secular by definition.
That was a reply to a comment about how “Monotheism is soooo ingrained now that we practically define religion as the belief in a single great creator god who is in charge of everything.”
Anonymous:
>Religions are non-secular by definition.
Obviously. I’m asking why one would want to define atheism as non-belief in only a creator god but not in other types of obviously religious phenomena like spirits or reincarnation. Daniel Lakeland points out that a single creator god is often the assumption when the term religion comes up, and I agree that seems to be the case. But David in Tokyo’s comment about Japan implies that this is somehow a meaningful distinction, which I don’t get. It would seem more coherent for an atheist perspective to be about non-belief in any spiritual phenomena or “truth”.
Jessica doesn’t get “atheism” as a term for how we “what you see is what you get atheists” see religion.
English doesn’t have a term for it, so we borrow it. And it’s not wrong, because religions have saints and spirits, and the spirits are just mini-versions of gods. So as a term, atheism is fine for we folks who think religion is problematical at every level.
The bottom line is that there’s no there there: physical reality is all you get. As soon as you look for “something else”, you are in trouble. You are hypothesizing objects/phenomena that don’t exists, and and since gods are one typical example of those nonexistent things, referring to people who see those nonexistent things as theists seems pretty reasonable. That leaves the rest of us as atheists. This is normal language use as described by the generative linguists, such a Lakov and Johnson.
If you don’t like being called a theist, don’t be one. Reality is deep and subtle and beautiful. Looking for “something more” is missing/ignoring/denying that deep subtle beauty.
I think we’re talking past each other. If atheism is generally referring to non-belief in any spiritual philosophy, then that makes sense to me – that is what I assumed the term meant in common usage. It was the implication that it’s primarily rejecting the single creator god view, but is not so much opposed to something like Buddhism, that was confusing me.
If anything, writing this blog post has convinced me that the lines between atheism and agnosticism and some religious views can be very blurry. Which I guess is where I started from, but I have a much better sense of atheist arguments now!
Much of Buddhism is compatible with athiesm. The stuff about reincarnation and soforth isn’t, but the stuff where you spend time introspecting the way your own mind works is totally compatible with atheism. That’s the bulk of Zen teachings. You can be a Zen Buddhist and an atheist in my opinion.
Under that view, my orientation to religion is likely compatible with atheism. Which works out well actually, given that I tend to prefer non-dualist perspectives!
To me “theism” means belief in some causes of the physical world which are “extra-material” and sentient… ie. “gods” of some sort. You can’t see them, you can’t know anything directly about them, but they nevertheless act to change the way the world works. Saints perform miracles, Thor decides to strike people down with lightning, Judaic god floods the earth after warning Noah, or spirits come and steal your underwear out of the laundry etc etc.
Buddhism is mostly nothing like that. Zen Buddhism doubly so (based on my limited study for a few years). The teachings of Siddhartha included a bunch of stuff about reincarnation, but my impression is that was a widely accepted “fact” about the world, and that he was attempting to explain how it worked rather than that he invented that stuff. In any case, his primary teaching is related to how our minds work, and what it means to be a good human… that we experience suffering in certain ways, that we think in certain ways, that certain things are good and moral, other things are to be avoided… all of which is informed by actual experience of being a human… ie. with evidence. There is no real appeal to extra-material sentient beings who influence the world.
The atheists I know object to the idea of extra-material sentient beings, not to the idea that we can alleviate suffering by changing our thought processes or that we should be kind to others or avoid addictive substances or whatever.
I don’t have much to add, but did intensively study (and practice!) Buddhism for a hot minute back in my early 20s. I would say it is clearly a set of technically “atheist” religions. Any study of the source material will find references to a creator God or Gods conspicuously absent. However, it’s success in far-flung cultures owes to its syncretic blending with indigenous and other worldviews – especially prominent in Tibetan flavors of Buddhism, but visible everywhere from India through Japan. What then happened is that many of the more contemporary Western assimilations of Buddhism shed some or all of the ‘cultural baggage’ (…some would argue essential features were thereby lost…) and emphasized the philosophical and pragmatic aspects that can be considered maximally compatible with secular humanism. In the extreme, this has lead to people like Sam Harris promulgating Buddhist meditation techniques but explicitly within a worldview that is dogmatically hostile to organized religion and attendant belief structures. The wisdom of this will only be known in the fullness of time, but whether it can or still should be called “Buddhism” is probably debatable. There’s a whole other conversation to be had about the perspective of the “Perennial Philosophy” or the universality of certain ethical precepts or metaphysical intuitions about the world…
Agree with Chris, some people have taken aspects of Buddhism and sort of surgically removed it from Buddhism. But I don’t think that’s what is in for example Zen Mind, Beginner’s Mind. If you started with that, you would not find “theism” but you would find relatively unmodified Zen. The SF Zen center had some considerable controversy after the death of Shunryu Suzuki Roshi but the book is pretty much his teachings and I think relatively un-westernized Zen.
I’m not so sure what I think of the westernized surgically removed stuff… I just don’t know anything about it really.
I agree Daniel. I think we’re getting into garbage time now, but to be sure, many teachers and lineages of Buddhism in the West are relatively continuous with their source cultures (little modified), while some have been extensively modified and repackaged into something basically unrecognizable as Buddhism but more as ‘self help’…
Same predicament for yoga, which essentially became an exercise program in the West. Commericialization appears to favor dropping the actual spiritual content but keeping the spiritual vibes.
I think about this a lot too. A lot of scientists have been openly religious or a least spiritual of some sort until fairly recently.
Interesting take on this from Christian author Marilynne Robinson:
https://archive.is/59pEb
”… by asserting that there is no god, rather than that this is uncertain, atheism strikes me as close to a straw man, point null hypothesis, and I have trouble getting past that part.”
Sorry, but I think this statement is the straw man. I don’t believe many atheists would say that it is literally impossible that any version of any god could exist. Just that it is sufficiently unlikely for them to say that they don’t believe it.
Agreed. I would appeal to Carl Sagan’s invisible dragon.
I think the issue I have is that if you define atheism as disbelief in the existence of something like a god, and agnosticism as uncertainty about the existence of something like a god, which both seem like common understandings of what these terms mean, then it’s hard to see how atheism doesn’t imply more information than agnosticism. Is this a misintepretation of what atheism means?
I also wonder how much gets confused by the many ways people can interpret what exactly is being rejected by an atheist. As other commenters have mentioned, if the question is do you believe in God or more broadly something god-like, this often interpreted in a specific way, as a single creator of things who is in charge, often anthropomorphized, and so I also wonder when people say they are atheists, are they disagreeing with that, or are they asserting that there is nothing other than what we can know to be true through science and direct evidence and therefore there can be nothing that might be considered a basis for religion? I assume the latter, but maybe the disbelief is about a narrow category than I am imagining.
Jessica, I think the issue comes down to the surrounding social context of the question being asked. To be tedious, the exact same sequence of words can have very different meaning in different situations, and I think there’s too much confusion over which of those meanings is being employed. Indeed, sometimes it seems the confusion is a deliberate tactic of evangelical proselytizers.
That is, is something like “Do you believe in God?” a philosophical inquiry into the nature of the Universe, or a political pitch to vote Republican?
Some disbelievers don’t want to engage in either Big Philosophical Debates, or contentious beliefs, and they tend to shrug off the whole thing, calling themselves agnostics.
Other disbelievers are more activist about either or both of these topics, and they are more likely to use the label of atheists.
Basically, I’d contend it there’s a strong aspect of how someone regards the social issue of religious belief, rather than just their own personal belief.
Going from ”point null hypothesis” to ”imply more information than agnosticism” is a considerable weakening of the claim, which I agree is appropriate.
Any category in natural language is a vague, fuzzy set dependent on inferred context (as Seth suggests). This shouldn’t be acknowledged for theism and neglected for atheism.
As you say, there are many versions of theism and I suspect most atheists would agree that they aren’t all equally reasonable. My suspicion is that many atheists would say that the versions of theism they are aware of are either unreasonable or too vague/poor in their theistic content. For example, many atheists would disagree with this: ”that there is nothing other than what we can know to be true through science and direct evidence” (i.e. scientism). But I’m not sure what the theistic content is there?
I wouldn’t want to put words in your mouth, but I’m wondering whether you might actually be taking issue with poor ontology (e.g. scientism) rather than atheism?
> I think the issue I have is that if you define atheism as disbelief in
> the existence of something like a god, and agnosticism as uncertainty
> about the existence of something like a god
Define atheism as “not believing in things without evidence”. Religion is “believing in things without evidence”. (Of course, that’s religion as belief, not religion as culture.)
Agnosticism seems to be “not knowing how to evaluate the evidence”.
> are they asserting that there is nothing other than what we can know to
> be true through science and direct evidence
I don’t know what you mean by “direct”. Obviously, there might be things that exist for which we don’t have evidence. But, why would you believe they exist before there is any evidence?
However, I think there is a confusion here. The issue is not that there is no evidence regarding religion. There is overwhelming evidence that religious beliefs are not true.
The logic will be clearer if we replace the ambiguous, vague word “god” with something clear, e.g., “If I drink a glass of orange juice every day, in a year the proof that P NP will come to me in a dream, and I will win the million dollars.”
My not-equal sign disappeared: P not-equal NP.
I do think there is a confusion in the terms here. I consider atheism not to be a belief against the existence of a deity but simply an absence of belief. In the same way ‘asexual’ is an absence of sexual desires and not strictly any kind of stance ‘against’. You could just call this absence of belief ‘agnostic’ or ‘uncertain’ but I think in this context saying one is ‘uncertain’ is more like suggesting one considers some specific deity or god as having some quite sizable probability. If one really just meant uncertain as in, does not know for sure, then I think that is perfectly consistent with atheism too. I would say I am uncertain in the sense of not being absolutely sure, but seeing as I have no logical or empirical reason to believe or even postulate a deity of any satisfactorily theistic kind, then I simply do not do so (I am without theism, or ‘a’theistic)
Jamie:
This is how Bertrand Russell put it:
> I consider atheism not to be a belief against the existence of a deity
> but simply an absence of belief.
This seems to be hair splitting. E.g., I don’t believe I will win the lottery today. Doesn’t that also mean that I believe I won’t win the lottery today? Maybe you want to say that a probability of 10^-5 is different from a probability of 0. But, no one would make such a distinction in any area other than religion.
Andrew: Yes, I was going to refer to something presumably derived from the same quote where I’ve heard it said that theists are all atheists with respect to every other god beisdes their own, and atheists just go that last tiny step.
David Marcus: You could see the distinction as hairsplitting but I think it is possibly meaningful/useful to make it in this context. If the position is staked out as being actively ‘disbelief’ as opposed to simply not believing then very quickly one finds different theists expect the atheist to engage and be well versed in the very specific aspects of each of their own religion that they are convinced make it true and for the atheist to disprove it. And if humoring them in this one often finds oneself then being accused of being a ‘militant atheist’. I think on paper I am an atheist but again it is not even part of how I would identify myself in the same way I would not identify as either an a- or non- or anti-bigfootist or pick any other less offensive seeming reference that just doesn’t feature into how people think of the world. Certainly I could answer that indeed, I do NOT believe in a deity/think there is no deity, but again this comes more from an absence of seeing any reason to believe and not from feeling a need to come up with specific arguments debunking theism, if that makes any sense to you.
Jamie:
It might not be the same quote. Bertrand Russell was notorious for repeating himself, and I think he used this line about the Olympian gods many times. It’s a good line, and it’s related to the point that I often make about open-mindedness, which is that you still have to choose what you decide to be open to.
I agree, there is confusion in terms. From comments here it seems that “atheist” can mean about as many things as “religious” can mean. My difficulty understanding atheism that I joke about in the post was about atheism as a lack of uncertainty about whether there could be something like god, not atheism as absence of belief with the possibility of uncertainty.
Since Bertrand Russell came up, I feel like someone should mention this BR anecdote. (I guess most of us know it, but there’s a first time for everyone):
Russell supposedly wrote that when he was admitted to Brixton Prison in 1918 for opposition to WWI, “I was much cheered, on my arrival, by the warder at the gate, who had to take particulars about me. He asked my religion and I replied ‘agnostic’. He asked how to spell it, and remarked with a sigh: ‘Well, there are many religions, but I suppose they all worship the same God.’ This remark kept me cheerful for about a week.”
I say “supposedly” because I haven’t read The Autobiography of Bertrand Russell, which is said to be the source for this.
Phil:
I did read Bertrand Russell’s autobiography–all 3 volumes of it!–and that story is indeed there. I posted on it back in 2011. The autobiography had some good stuff but it was also disturbing in that he left out huge chunks of his life. It read like a movie that had been drastically cut by the studio.
If you want to learn about all the missing parts, I strongly recommend Ray Monk’s biography of Russell which I briefly mentioned here.
I’m definitely not reading three big volumes about anyone, but I guess I could handle 750 pages, so, OK, sounds good!
I looked back at your previous post and was struck by how long I’ve been away of Douthat’s douchebagism. I think of it as something I only recognized within the past eight or ten years but, although I don’t recall that particular post, having re-read it it now seems likely that my negative feelings about Douthat go back at least fifteen years.
Phil:
Bertrand Russell’s autobiography is 3 small volumes. As I said, it’s readable but it’s really weird, which I guess is part of the charm. Ray Monk’s biography of Russell is 2 large volumes. There were some places where he went into more detail than I thought was absolutely necessary, but after reading it I got a much better sense of Russell’s life and work.
Jamie,
> if that makes any sense to you.
If you mean that you can’t on the spur of the moment explain to theists where their error is, that’s fine. If you mean you don’t believe in god because your parents didn’t (as one friend sort of told me once), that’s more like saying you aren’t sure whether the sky is really blue because you never bothered to look.
Hi Jessica,
Long-time lurker, first-time commenter. I’m a reformed clergyman (40 years now) who has always enjoyed math—especially probability. To the excellent recommendation of Marilynne Robinson, I’d add Freeman Dyson’s Templeton Lectures, collected in the book Infinite in All Directions.
Also interesting, in the world of statistics, is the relationship between Pascal (a man of deep faith) and the gambling savants of his time. There’s a good chapter on that in Ian Hacking’s The Emergence of Probability.
On the issue of selection bias: I’ve always had scientists and academics—PhD-level engineers, chemists, etc.—in both urban and rural congregations. But the scientifically trained professionals most represented in my churches have been physicians and surgeons.
I think the core insight underlying Feuerbach, if I remember him correctly, is right. We have feelings, intimations, longings that we are uncomfortable acknowledging, so we project them onto an external force. Or maybe we acknowledge them but claim that they originate from the part of us that is transpersonal, perhaps universal. How much of religion would remain if we fully owned all that we are and aspire to be or fear?
As for atheism per se, I think it cedes too much. You tell me enough about a proposed god that I can have a basis for believing or not believing, and I’ll give you an answer. Religious gods, for me, are way too undefined.
Finally, what gec called the social organization of religion brings up Durkheim. I suspect, from an evolutionary point of view, religion started out they way he characterized it and only in the last 3000 years or so did it try to use the framework for broader moral or political purposes.
Re: “I feel unsettled when I find out that someone I otherwise relate to identifies as an atheist is because by asserting that there is no god, rather than that this is uncertain, atheism strikes me as close to a straw man, point null hypothesis, and I have trouble getting past that part.”
I agree w/ Mattias’ comment above, that it is this (Jessica’s) statement that is the straw man. It’s not the “point null” that atheists are rejecting, but rather a connection between evidence and belief that hasn’t been met.
I wonder if one substitutes “intelligent beings running a computer simulation that we’re all in” for “god,” would Jessica’s unease still exist. If not, why not?
I disagree strongly with several parts of this post. Despite this, I like it and I’m glad you (Jessica) posted it, as it touches on several interesting and important topics. The question of whether participating in the rituals of belief requires actual belief is a fascinating one, and I think the answer has always been very different in different traditions. About finding meaning even though “randomness and chaos and individual agency are all that exist,” isn’t that the whole point of existential philosophy? These questions aren’t at all new.
> but rather a connection between evidence and belief that hasn’t been met.
Agree.
I assume the use of “point null” is to indicate that the claim is unreasonable. Let’s ignore whether point nulls are always unreasonable. It is wrong to say atheism is a point null because once you rule out god, you are left with all the various possibilities that we put under the label of “science”. As should be obvious, there isn’t just one possible scientific theory. If anything, belief in god is closer to being a point hypothesis than is science.
There is a quote that struck me many years ago that I am about to horribly butcher but goes something like this…
Biologist see the world and can’t believe there is a God, astrophysicists see the world and can’t believe there isn’t.
As a constructivist I think that things (and whatever) exist by being constructed, for the persons who construct them. Now construction is pretty complicated. It has social and individual elements, many of which are unconscious. Also that there is construction doesn’t rule out that there is an external world that constrains construction, or that constructions may be well aligned with something that truly exists in the external world. But construction happens when and arguably even before we consciously perceive anything, so the external world is inaccessible to us without construction, and I’m generally agnostic about it, even though I accept that there are some very solid widely shared constructs that make it very tempting to say “these things really exist and nobody can deny it” (at least hardly anybody actually denies them), and I’m fine to do science to figure out which constructs are stronger and more stable, and which are less strong and fall down if we explore them.
So I’m not saying we can consciously set out to create a god out of nothing, and then we have one. What I’m saying is that religious belief, and particularly religious experiences connected to religious practice, can be seen as results of construction processes, and what they refer to can be seen as existent by means of being constructed, for those who run these (at least partly unconscious and social and constrained) processes. (One can even go as far to say that “god” “exists” in a sense for those who don’t believe in god, namely as something other people believe in, at least to the extent that the non-believers are affected by it.) An atheist may say that this does not mean that there is any real and objective truth to the beliefs, and this kind of construction may very well work without any such truth, and yeah, that’s true. Actually certain religious persons may have difficulties with my view for the same reason, only under the opposite premise – nothing in what I write above makes sure that their beliefs are as true as they might think for any other person, other than social processes). However as a constructivist I generally think that construction is the major mode how “things” exist for us, so existence by construction is about as good as existence gets. At least this gives religion a justified claim to refer to something that exists.
My father is an evangelic pastor, but he isn’t very missionary. I have talked with him a lot about his beliefs, which I don’t share, but I have a lot of respect for them. Particularly I have learnt that the whole issue whether god “really” exists or not makes very little difference regarding his religious experiences and what they mean to him. His sermons are for people who are open to explore what texts and messages from the bible could mean to them and their lives, their experiences and practices, They are not about whether anybody should believe or whether belief is true or whether god really exists. An atheist who says that god doesn’t really exist and the texts and messages in the bible are man-made has nothing to say that would be relevant to the religious practices and experiences of my father. The question whether atheism is right or not just passes him by. Constructivism gives me a positive and respectful understanding for this without the need to share the belief (and I’m not proud to not share it, and neither do I think I’m right and he is wrong – my construction processes just have led me somewhere else), and make the exchanges with my father an enriching experience.
I should probably add, as I haven’t explained this above, that I use construction in the sense of “constructing” our perceptions and thoughts by means of brain functions, body processes etc. I’m not talking about building something physical. (If that wasn’t clear anyway.)
Thanks for sharing, this resonates with me. I don’t really feel the need to question whether a basic sense of faith I feel is “objectively” true, because that seems like the wrong question to apply; that it might be a mental construct isn’t really relevant to the truth of it. This doesn’t seem like a violation of science to me, but perhaps to some people it does.
If what is in your head is a statement about what is not in your head, then the mere fact that it is in your head does not make the statement true.
There is nothing wrong with fiction. I enjoy all sorts of fiction. But, I know it isn’t true.
Science is the process we use to determine what is true.
External reality is inaccessible without the mediation through mental constructs, and every truth statement that can be verified refers to such constructs rather than external reality itself. Technically you are right that the mental construction of one person doesn’t make anything true for another person, but a religious person may not bother. The own experience and sharing with a like-minded community may just be good enough. And nobody “knows it isn’t true”, at least not beyond their own constructs.
The correct statement is that the mental construction of one person doesn’t make the external world do that. Is that what you meant to write? Sure, you may be able to find other people who believe the same false things you believe. But, no matter how many people believe a false thing, it still is false.
I agree with what you call “correct statement”, however I don’t have the privileged access to the external world that would be required to know whether certain religious beliefs are “false” there. And as far as religion is about personal and social experience and practice, it is not that relevant what is true or false in the external world.
You don’t need “privileged access”. You can use science (and statistics).
“asserting that there is no god, rather than that this is uncertain, atheism strikes me as close to a straw man, point null hypothesis, and I have trouble getting past that part.”
Most atheists, I think, are technically “agnostic atheists”; they are uncertain about a god but see no evidence for it.
Ah, well that would help resolve some of my confusion about how they know there is nothing like a god.
> Most atheists, I think, are technically “agnostic atheists”; they are
> uncertain about a god but see no evidence for it.
That’s wrong. Why should we be uncertain if there is no evidence? There are oodles of things for which there is no evidence. Do we have to be uncertain about all of them? Or, just about god?
But, this misses an important point: There is lots of evidence that there are no gods.
There are oodles of things for which there is no evidence, and, yes, I think we should be uncertain about all of them. I’m a bit surprised by your implied disagreement with this stance. Do you really that we should be certain of things for which there is no evidence?
It depends. You need to consider your prior and all of the data (not just some of the data). There is overwhelming evidence against there being gods. That’s why we developed science: because the god method didn’t work.
Kashmir Shaivism? That’s interesting! How did you end up there?
Technically I think its also influenced by Vedanta. I had a friend in Ann Arbor where I was living at the time who took me to a meditation center there, and I found it compelling, perhaps in part because it was so different from what I’d experienced of Western religion. I think a lot of the reason it stuck was timing. I’d had this intuition on and off throughout my teens that there was more to reality than just the physical world, but then in the second half of my twenties it started to feel like I needed to try to understand what it was. So I was on some level on the lookout for a more coherent philosophy.
I was put off by religion when I learned that despite being the right age to be chosen to be the Dalai Lama, but for other reasons, did not make the cut.
And, recall that messing around with the Dalai Lama can be dangerous as witnessed by this Sgt. Bilko T-shirt incident of 1987:
https://www.upi.com/Archives/1987/11/14/Sgt-Bilko-mistaken-for-the-Dalai-Lama/9590563864400/
Hardcore atheist here. Let me try and clarify things from this side of the fence.
> I find it amusing how part of why I feel unsettled when I find out that someone I otherwise relate to identifies as an atheist is because by asserting that there is no god, rather than that this is uncertain, atheism strikes me as close to a straw man, point null hypothesis, and I have trouble getting past that part.
Others have pointed out the atheist/agnostic divide, but I think some elaboration is useful. Pro-religious arguments usually avoid empiricism and instead argue from first principles: Descartes’ “I think therefore I am” was followed by “and God exists,” via an ontological argument that does not invoke observation at all; the “first mover/creation” argument, a form of ontological argument, have been popular for millennia; and another popular strain argues that morality or rationality cannot exist without a God. The biggest exception that comes to mind is Swinburne’s “The Existence of God,” which mixes logical arguments with Bayesian empiricism. Since counter-arguments follow the form of the associated argument, they rarely need to invoke evidence at all and thus don’t have a “null hypothesis.”
An agnostic, then, is best thought of as the peer reviewer of a paper. The paper author/religious person is making an argument, be it logical or empirical, and the task before is not to pose your own counter-argument but to evaluate how the other person’s stands up. They may have those arguments, but they don’t have to share them. They can simply respond back “not good enough, methodology is flawed” and not be compelled to endorse or reject the paper’s conclusions.
Most atheists can be thought of as authors of a commentary on a paper. They didn’t come up with the original hypothesis, but unlike the agnostic they are making some sort of counter-argument, in public, and with full expectation those arguments will be scrutinized. This doesn’t necessarily take the form of a hypothesis, but the end goal is an explicit rejection of the original paper’s conclusions rather than a technical objection. When a hypothesis is invoked it’s usually against a specific God. For instance William Rowe’s “argument from evil” is a hardened version of “why does suffering exist?,” but that presupposes that God is opposed to suffering and willing/able to intervene to prevent that suffering.
A few “hardcore” atheists are just straight-up authoring their own papers. They are trying to create original arguments or hypotheses that apply to many or even all gods, without explicitly referencing other arguments/papers. I consider David Hume’s argument against miracles a good example, with little effort you can elevate it to an argument against divine revelation that works against a large number of gods.
> The pockets of heavily moralistic, often bordering on religious thinking that have become more visible in tech in recent years are also intriguing. Some writers associated with the rationalist community have consistently discussed religious matters, for example, like Scott Alexander mentioning god in his essays.
There’s a reason for that. Take this example: why is The Bible a holy text? Because it is infallible, and it claims to be a holy text. There’s an obvious circularity there. Now compare it to this: how can we be sure LLMs are capable of intelligent thought? Because beings capable of intelligent thought are capable of explaining their reasoning, and LLMs are capable of explaining their reasoning. The circularity isn’t as obvious, but it’s there: how do we know their explanation of their reasoning was their actual reasoning? You can’t even prove I’m thinking, how on Earth can you prove something structured entirely unlike me is engaged in thought?!
In my opinion, both the rationalists and religious apologists are invoking the same logical fallacies to justify their beliefs. This leads to a sense of familiarity and “sameness” between them. It doesn’t help that our culture is dominated by religious language, thus the closest metaphor is often a religious one. Speaking of:
> I imagine if I spent time thinking about it I could probably persuade some part of me to believe that randomness and chaos and individual agency are all that exist, but then I wonder why on earth would I want to do that, since it sounds stressful.
I’m the exact opposite. I suppose you might be able to convince me of the existence of some higher power, but only if I sacrifice a lot of my knowledge and critical thinking, and even then I’d likely be miserable. The religious worldview seems perfectly designed to invoke fear and anxiety, and there’s nothing comparable on the atheist side of things. There’s a reason for our divergence: as a second or third-generation atheist, I only got a light sampling of religious thought via culture. Our different upbringings exposed us to different arguments, and so what seems alien and foreign to you is homely and comfortable to me.
> I wonder if things would be different if we were more comfortable as scientists with acknowledging that having irrational commitments to certain ideas is part of what we do. We perceive this as conflicting with scientific objectivity, but science is full of leaps of faith we make to locate ourselves within various epistemic-philosphical conditions.
I think of this mathematically. Euclidean geometry rests on five axioms, and four of them are interdependent. After many centuries of trying to show the same for the fifth postulate, the parallel line one, mathematicians instead swapped that postulate out for others and invented new forms of geometry. I’d say the same about the philosophy of science: we don’t accept things on faith, but instead declare axioms that define our epistemology.
As humans, though, we’ve personalized those axioms and tend to cling to them too tightly, to the point of staking our careers on a particular set. I’m firmly on the Bayesian side of things, but I’ve made an explicit effort to practice thinking in terms of the axioms of frequentism. When I get to the point of being able to convince someone I’m a frequentist and not a Bayesian, I’ll count that as a success. The goal here is to depersonalize these axioms and think about them critically, the opposite of taking any of them on faith.
> how can we be sure LLMs are capable of intelligent thought?
I’m certainly not sure of that. Recently I’ve been thinking that maybe we’ve misinterpreted the Turing Test. I think all it shows is that the machine is (in some ways) as intelligent as the person. Maybe many people are just running an LLM type algorithm in their heads. I think I run such an algorithm in my head, but I also think I run other types of algorithms too.
The Turing Test’s biggest flaw is that it doesn’t account for pareidolia. Some people thought Eliza, a sixty-year-old chatbot, was human! Conversely, I have yet to find a “foundational” LLM that can reliably count. While I don’t think that issue makes the test useless, it does highlight that we put far too much emphasis on surface details and presentation when it comes to evaluating these LLMs.
As for “LLM-type algorithms,” you’ve got the chain of causality reversed. LLMs are based on CNNs, which in turn are based on neural networks, which were simplifications of biological studies of real-life neurons. I wish I’d saved the citation, but I once stumbled on a paper which compared a neural network against human volunteers at an image recognition task, and found that (if you deliberately limit the human side) the successes and failures are eerily similar between both. Their imitation of our biology, and the resulting flashes of familiarity, only feeds into our pareidolia.
One way to think of LLMs is like a virtual earthworm: start with a primitive nervous system, but rather than have it drive a complicated body through a tricky environment, simplify everything else down so that does nothing but take in and excrete words written by humans. Let that evolve for the equivalent of hundreds of millions of years, and you’ll get back something supremely efficient at constructing sentences that look quite human, but which nonetheless lacks anywhere near our brainpower.
As I’ve said before, Turing’s Turing Test, as described in his paper, doesn’t have that problem, since it doesn’t ask the machine to fool people, it asks the machine to be sensitive and thoughtful, and compares it’s performance in a task that requires sensitivity and thoughtfulness with human performance on that task. Nowhere does he talk about fooling users. That’s not what the test is about.
“Conversely, I have yet to find a “foundational” LLM that can reliably count.”
Yep. They can’t do multiplication, either. The idea that LLMs can “solve math problems” is seriously insane.
A better way of thinking of LLMs is that they are random next word generators that don’t have an internal model of the world (or anything else) to check their output against. So the language they output is most similar to the speech of Wernike’s Aphasia patients. Lots of words, no meaning.
That LLMs generate interesting-looking text is truly amazing, impressive, seriously kewl. And intellectually vacuous.
Of course we’ve been developing algorithms based on our knowledge of humans. My point is that when you have such an algorithm, it can offer a model or suggestion for how the algorithm in our heads works. But, I think it is a mistake to assume that all humans are running the same algorithm in their heads.
I just read Turing’s paper. I hadn’t read it before. It is worth reading.
What are, according to you, the “axioms of frequentism”?
I was wondering what that meant, too.
Hopefully Hornbeck will reply. But if I were to try to describe something like that I’d say that:
1) For statistical purposes we replace unknown mechanism with the output of high kolmogorov complexity random sequences.
2) Probability is defined as the long run limit of the frequency of observed behavior in those sequence outputs.
That doesn’t seem to be enough to derive frequentist statistics.
Daniel Lakeland is pretty close, but I’d put the fundamental axioms as:
1. Probability represents long-term behaviour.
2. The expected contribution of measurement error comes arbitrarily close to zero as the sample size grows.
3. Hypotheses, or potential quantified explanations of observations, cannot have probabilities assigned to them.
David J. Marcus is correct, that’s not enough to actually reconstruct frequentism. Part of the reason is that I’m leaving out the low-level axioms about probability, since by invoking “probability” I’m implicitly suggesting the Kolmogorov axioms or something equivalent are in play. The second axiom also depends on the Central Limit Theorem, so any axioms needed to construct that also must be included.
But part of the reason is that frequentism doesn’t have a consistent set of axioms. Fisher, for instance, would add:
4. Hypotheses can either be unworthy of belief, and thus safely ignored, or otherwise.
5. If a hypothesis is sufficiently inconsistent with a summary statistic, we are allowed to reject/ignore it.
And based on those, would invoke Karl Pearson’s p-values to reject or fail to reject hypotheses in isolation. Jerzy Neyman and Egon Pearson would reject those axioms, and replace them with:
4. One hypothesis can be unworthy of belief relative to another, or otherwise.
5. If one hypothesis is sufficiently more consistent with a summary statistic than another, we are allowed to accept the former and reject/ignore the latter.
And spin up likelihood ratios or confidence intervals instead. If you look at how frequentism is actually practiced, though, you’re dealing with a Frankenstein’s monster that tries to invoke all of those axioms simultaneously. Hubbard et. al lay this out quite nicely:
> In the merging of the two schools of thought, it is often taken that Fisher’s significance testing implies an alternative hypothesis which is simply the complement of the null, but this is difficult to formalize in general. For example, what is the complement of a N(0, 1) model? Is it the mean differing from 0, the variance differing from 1, the model not being Normal? Formally, Fisher only had the null model in mind and wanted to check if the data were compatible with it.
> In Neyman-Pearson theory, therefore, the researcher chooses a (usually point) null hypothesis and tests it against the alternative hypothesis. Their framework introduced the probabilities of committing two kinds of errors based on considerations regarding the decision criterion, sample size, and effect size.
> Raymond Hubbard and M. J Bayarri, “Confusion Over Measures of Evidence (p ’s) Versus Errors (α ’s) in Classical Statistical Testing,” The American Statistician 57, no. 3 (2003): 171–178, https://www.tandfonline.com/doi/abs/10.1198/0003130031856.
Ironically, they are guilty of doing the opposite, “Fisherizing” Neyman and Pearson’s ideas. Flipping to Neyman-Pearson, we find:
> It is clear that besides H_0 in which we are particularly interested, there will exist certain admissible alternative hypotheses. Denote by Ω the set of all simple hypotheses, which in a particular problem we consider as admissible. If H_0 is a simple hypothesis, it will clearly belong to Ω. If H_0 is a composite hypothesis, then it will be possible to specify a part of the set Ω, say ω, such that every simple hypothesis belonging to the sub-set to will be a particular case of the composite hypothesis H_0. […]
> Suppose further that besides the hypothesis H_0 to be tested, there are only two alternatives H_1 and H_2. … Looking at the diagram we see that, if the process of sampling was repeated many times, then, were the hypothesis H_0 true, most sample points would lie somewhere near the point O. On the contrary, if H_1 or H_2 were true, the sample points would be close to O in comparatively rare cases only. … It is clear that in the situation presented in the diagram the best critical region with regard to H_1 will not be the best critical region with regard to H_2. While the first may be ω_2, the second may be ω_3. But it will be shown below that in certain problems there is a common family of best critical regions for H_0 with regard to the whole class of admissible alternative hypotheses Ω.
> Jerzy Neyman, Egon Sharpe Pearson, and Karl Pearson, “IX. On the Problem of the Most Efficient Tests of Statistical Hypotheses,” Philosophical Transactions of the Royal Society of London. Series A, Containing Papers of a Mathematical or Physical Character 231, no. 694–706 (February 16, 1933): 289–337, https://royalsocietypublishing.org/doi/abs/10.1098/rsta.1933.0009.
Both Neyman and Pearson didn’t limit the number of alternative hypotheses, I doubt they even believed the number had to be finite (an earlier section mentions “an infinite set of population distributions” can be drawn from a Gaussian distribution). Nonetheless, their likelihood ratio only allowed one-to-one comparisons, so even they tended towards a null/alternative hypothesis duality that made life simpler. Confidence intervals are an explicit endorsement of this, as stating that X% of the summary statistics you calculate will fall in the interval implies (100-X)% of them will not.
The result is an utter mess. Turning back to Hubbard et. al:
> As a prime example of the bewilderment arising from the mixing of Fisher’s views on inductive inference with the Neyman-Pearson principle of inductive behavior, consider the widely unappreciated fact that the former’s p-value is incompatible with the Neyman-Pearson hypothesis test in which it has become embedded (Goodman 1993). Despite this incompatibility, the upshot of this merger is that the p-value is now inextricably entangled with the Type I error rate, α.
You cannot generate a full set of axioms for a system that is self-contradictory. At best, you can skim off some “fundamental” ones everyone agrees on, but something will always be missing. On the plus side, though, this indifference to contradiction makes it easier to convince people I understand frequentism!
Go back to the early discussions of probability (de Moivre, etc) or even look at the basic gambling examples still used today.
It is so obvious probability is combinatorics, but quickly becomes intractable for all but the simplest problems. Observing frequencies is a way to estimate the underlying “true” (but unknowable) combinatoric probability. It is not probability itself.
https://en.wikipedia.org/wiki/The_Doctrine_of_Chances
An easy test to see if you are dealing with actual probabilities is checking whether they are limited to rational numbers (ratio of two integers). If not, then you aren’t working with the real thing, its an approximation/estimation.
For a great example of what ensues from confusing the estimation with a bona fide probability, check out all the attempts at binomial proportion confidence intervals. None behave correctly.
Thanks HJ Hornbeck, this is very interesting, but potentially opens a can of worms for which this place (discussion of a rather unrelated posting) is not the right one in all likelihood. Also thanks Daniel, that’s another interesting shot at it.
I’m not sure whether frequentism can be set up by a system of axioms, so I agree with your (HJ Hornbeck) critical remark in that respect, but it may be worthwhile and instructive to try. In any case I wouldn’t agree with your axiom 2, because there is nothing non-frequentist about accepting that measurements can be systematically biased. This is an issue of measurement theory and the specific measurements we’re considering, not about frequentism. Also 3 doesn’t hold in general, as there can be situations in which certain hypotheses of interest can be modelled as themselves drawn from a frequentist probability distribution (for example recognition of a letter from handwriting), in which case a Bayesian analysis is fine for a frequentist (also I’d always separate a frequentist concept of probability from the application of “classical frequentist inference” such as hypothesis tests and confidence intervals; one can be a frequentist about the meaning of probability without being committed to tests and CIs). I may have more to say but maybe I already said enough to open the can of worms even further… ;-)
> Christian Hennig: Thanks HJ Hornbeck, this is very interesting, but potentially opens a can of worms for which this place (discussion of a rather unrelated posting) is not the right one in all likelihood.
True true, we’re starting to overlap with what I’ve said in a comment thread a year and a half ago, so I’ll keep things brief. Speaking of:
> Anoneuoid: Go back to the early discussions of probability (de Moivre, etc) or even look at the basic gambling examples still used today. It is so obvious probability is combinatorics, but quickly becomes intractable for all but the simplest problems.
You might want to skim over that comment thread: https://statmodeling.stat.columbia.edu/2024/01/08/bayesians-are-frequentists-2/#comment-2303378
> Christian Hennig: In any case I wouldn’t agree with your axiom 2, because there is nothing non-frequentist about accepting that measurements can be systematically biased.
Careful, I’m not saying that non-frequentist statistics must admit systematic bias/measurement error is possible, I’m only saying frequentist statistics assumes any bias or error tends towards zero as sample sizes increase. This comes from the mandated use of “summary statistics” rather than the observed data. I keep forgetting how deep this goes, I only just realized that what I thought was a “parameter space” in Neyman/Pearson (1933) is actually a “summary statistic space.” Now, that doesn’t stop a frequentist from coping with bias/error by “de-biasing” their samples, but by the time you turn the frequentist crank you’ve got to be absolutely certain there’s no long-term bias/error.
> Also 3 doesn’t hold in general, as there can be situations in which certain hypotheses of interest can be modelled as themselves drawn from a frequentist probability distribution …
Then you’re not doing frequentism. Check out the linked comment above, I include multiple quotes from both Fisher and Neyman/Pearson explicitly stating that hypotheses cannot have probabilities assigned to them, or implying the same must be true as a consequence. Nor are they alone.
> Because uncertainty about parameters is epistemic, frequentist inference does not allow probability statements about the parameters of a statistical process. For instance, the fact that a frequentist 95% confidence interval for the normal mean μ is [−0.5, 1.0] does not mean that there is a 95% probability that μ is in [−0.5, 1.0]. Instead, what it means is that if the same procedure to construct confidence intervals was repeated very many times, for all kinds of different datasets, then in 95% of the cases would the true μ lie in the 95% confidence interval.
> Wagenmakers, EJ., Lee, M., Lodewyckx, T., Iverson, G.J. (2008). Bayesian Versus Frequentist Inference. In: Hoijtink, H., Klugkist, I., Boelen, P.A. (eds) Bayesian Evaluation of Informative Hypotheses. Statistics for Social and Behavioral Sciences. Springer, New York, NY.
My only hesitation over axiom three was whether it was an implied consequence of axiom one, or whether I should be explicit because so many frequentists are “accidental Bayesians,” including Fisher himself. The second option won out, for obvious reasons.
> I mentioned earlier that Fisher both rejected the Bayesian cake and wanted to eat it, too: He spoke of the level of significance as a measure of the degree of confidence in a hypothesis. In the minds of many researchers and textbook writers, however, the level of significance virtually turned into a Bayesian posterior probability.
> What I call the Bayesian Id’s wishful thinking is the belief that the level of significance, say .01, is the probability that the null hypothesis is correct, or that 1 — .01 is the probability that the alternative hypothesis is correct. In various linguistic versions, this wishful thinking was taught in textbooks from the very beginning. Early examples are Anastasi (1958, p. 11), Ferguson (1959, p. 133), Guilford (1942, pp. 156-166), and Lindquist (1940, p. 14). But the belief has persisted over decades of teaching hybrid logic, for instance in Miller and Buckhout (1973, statistical appendix by Brown, p. 523), Nunnally (1975, pp. 194—196), and the examples collected by Bakan (1966) and Pollard and Richardson (1987). Oakes (1986, p. 82) reported that 96% of academic psychologists erroneously believed that the level of significance specifies the probability that the hypothesis under question is true or false.
> Gerd Gigerenzer, “The Superego, the Ego, and the Id in Statistical Reasoning,” A handbook for data analysis in the behavioral sciences: Methodological issues (1993): 311–339.
@HJ Hornbeck: Well, I beg to differ. Of course you are right that much frequentist analysis assumes that measurement error has mean 0 and averages out in the long run (which is by the way also implicitly assumed in much standard Bayesian analysis), but there is nothing in frequentist philosophy that says that this assumption is fulfilled automatically. It may well not be, and a frequentist shouldn’t have difficulties admitting this (in which case of course a straight textbook analysis will be problematic, but neither is it part of frequentism that all analyses should be straight textbook analyses).
And re item 3: Von Mises, Neyman, and other frequentists were fine with doing Bayesian calculations in setups where outcomes would be generated at various levels from processes to which frequentist probabilities apply (which is arguably the case for letter recognition from handwriting). Of course frequentists accept Bayes’ theorem and are happy to use it where the involved probabilities are frequentist. We may actually agree about this if I don’t call the “higher level outcomes” (i.e., those that are unobserved and about which inference is required) “parameters” or “hypotheses” in case they are actually random variables. Then indeed, frequentists don’t attach probabilities to parameters (although this is a bit circular).
We’re just not going to agree on axiom two, and since things are so far off topic I’m happy to leave it there.
Your comments related to axiom three intrigue me, though. Do you have some citations to drop? After some searching, I found an instance where Neyman did in fact endorse a limited use of Bayes’ Theorem, much to my surprise:
> Neyman, Jerzy. “Basic ideas and some recent results of the theory of testing statistical hypotheses.” Journal of the Royal statistical society 105.4 (1942): 292-327.
After reading it over a few times, I’m stunned. Neyman leaves out some critical steps, makes some beginner-level mistakes, and if I understood the first problem correctly then he doesn’t seem to realize he’s just shown frequentism to be much less useful than Bayesian statistics. I’ll need some time to verify that, though, my skill level with frequentism is obviously nowhere near Neyman’s and I could easily be overlooking something.
I’m curious if Von Mises fares any better.
HJ:
Mises botched a simple applied statistics problem–see pages 84-85 of this article.
This is not to say that Mises was always wrong about statistics, just that even simple applied statistics can be hard. The problem that he got wrong could actually have been analyzed just fine using classical hypothesis testing. Mises just didn’t apply it right.
HJ: See von Mises’ discussion of Inference and Bayes’s Problem from p.116 of “Probability, Statistics, and Truth”, 1928 version, vivble here:
https://archive.org/details/in.ernet.dli.2015.189506/page/n129/mode/2up
Of course whether Neyman or von Mises got some calculations wrong has no bearing on their philosophical views.
> Mises botched a simple applied statistics problem–see pages 84-85 of this article.
This bit is confusing: “it is unnecessary to search for an explanation for the discrepancy, especially given that, as von Mises notes, birth numbers of boys or girls are among the rare data that actually do generally follow the binomial distribution.”
It’s unnecessary to search for an explanation for the slightly subnormal dispersion because we already expect the dispersion to be lower than for a binomial distribution and we already have an explanation for that?
(Also, it seems that the “fairly healthy p-value of 0.10” – when the dispersion is correctly computed – is the sort of binary thinking that may prevent people from noticing interesting facts like that one.)
Carlos:
There are a few things going on here. First, almost any possible explanation involving variation would give you overdispersion in sex ratios rather than underdispersion. Second, data will never be exactly at their expected value, and the underdispersion in these data is well within what could occur by chance. Third, the sample size is very small so that any plausible variation would be undetectable from these data. Fourth, his explanation for the underdispersion makes no mathematical sense. Fifth, had he seen the equivalent amount of overdispersion he could’ve explained it pretty much the exact same way. Sixth, he summarizes in terms of the variance rather than the standard deviation, thus depriving himself of any intuition or background knowledge regarding the actual scale of variation.
I’m pretty sure that what was going on was that he wasn’t particularly interested in sex ratios; he just used this set of available data to demonstrate a calculation. The trouble is that it’s bad statistical practice, and bad scientific practice, to spin a web of complicated theory to explain a data pattern that falls well into what would be expected from random variation.
Nobody’s perfect, and maybe this is just one silly mistake he made. But it’s still a mistake. As with the other mistakes made in that article, it’s a mistake that I think is fair to call sloppy. He didn’t think things through. In this case it’s ironic because a simple hypothesis test could’ve revealed the problem. But, back then, the idea of a hypothesis test for variation (rather than just a hypothesis test for the mean) may have been kind of obscure, so he didn’t think of seeing whether the pattern which he was so elaborately explaining could’ve been much more simply and plausibly be explained as just a bit of chance variation in 24 data points.
Once you insert infinity you are not dealing with probability, its an approximation. The simple fact is, there is no infinitely precise dataset to plug in.
Also, Andrew, you write the dispersion “is not at all significantly less”. You are simply using a more stringent significance cutoff than von Mises. There is no principled way to choose that value, which is why people use the arbitrary number 0.05.
The correct approach is use Bayes rule to compare the underdispersed models (eg, I’d say a “streaky” model is plausible due to seasonal infectious disease) with the binomial model, and any other models of births too. The more the merrier.
Anon:
My best analysis here is not based on hypothesis testing. Any reasonable Bayesian analysis for this example would conclude that there’s no evidence for Mises’s ridiculous theory. I just used the p-value when writing up the example to demonstrate that even using crude statistical methods it should be clear that such a result could easily happen by chance.
My guess is that Mises had no intuition about the variability of the chi-squared distribution with 23 data points. He was just overreacting to a random pattern in a small dataset. People do that all the time; it’s just kind of embarrassing for that to happen in a book with the grandiose title, “Probability, Statistics and Truth.” A bit more understanding of probability and statistics would’ve got him closer to the truth on that one!
Thanks for the citations! Von Mises is going to take some time to dig into, but on a preliminary reading he shares one of the misunderstandings that Neyman makes as well.
> Christian Hennig: Of course whether Neyman or von Mises got some calculations wrong has no bearing on their philosophical views.
Agreed, and I have little interest in that sort of mistake. But the Neyman citation I gave is relevant to this post:
> The situation is different if we adopt the subjective point of view. There is considerable variation of, say, radicalism, among the present proponents of this theory. The most radical seem to be the writings of Jeffreys, for whom to any proposition, on any amount of information, there corresponds a perfectly determined probability, which has nothing to do with frequencies. However, I could not quite follow the method of actually determining those probabilities. […]
> As I have already mentioned, I cannot very well follow the writings of Jeffreys, …
> Neyman [1942]
Neyman does not fully understand the Bayesian interpretation of probability. That’s easy to empathize with, he was heavily influenced by Fisher’s theories of probability and by this point has probably spent a good decade thinking in probability-as-frequency terms and surrounded by others who did the same. Rather than reach out and ask for help understanding the opposite view, though, he instead critiques it based on what even he acknowledges as an incomplete understanding. Unsurprisingly, the result is deeply flawed. Had Neyman instead approached the topic from an exploratory angle, instead of a more declarative one, it would have resulted in a much better discussion.
I do want to provide a brief sketch of what I consider one of Neyman’s mistakes, though, on the off chance I’m being too declarative myself.
> Therefore, if one is interested only in the “intensity of belief” … he may find it useful to measure it by means of the Bayes’ formula. If, however, he wishes to know something more definite about the relative frequencies, say of successes in selection and breeding, and the necessary a priori probabilities are missing, then the formula of Bayes will not help him very much, in fact it will not help him at all.
> Neyman [1942]
As a Bayesian who knows the concepts of the strength of a prior and the posterior predictive distribution, this is obvious nonsense to me. But while the date of publication here is 1942, Neyman delivered this paper as a lecture in 1939. Numeric sampling of probability distributions didn’t exist back then, and barely existed outside the physics community even in the 1950’s. I’m privileged with knowledge and computational power Neyman didn’t have, so I do have to grade him on a curve. But even given that concession, I find it hard to believe he was unaware of Bernoulli processes. All the pieces were there, and he had the brains to fit them together, but he didn’t.
> There are a few things going on here.
Andrew, I have to start by saying that I completely misread your comment as “do NOT generally follow the binomial distribution” and misunderstood the “as von Mises notes” reference.
I agree with some of your points, but why does his explanation for the underdispersion make no mathematical sense? If the births come from an inhomogeneous population the expected dispersion is indeed smaller. I agree that it may not make biological sense, because to have an effect larger than 10% one would need something that “half the population has a boy:girl ratio of 2:3 and the other half a ratio of 3:2”.
I also agree that he didn’t really need to look for an explanation of the slightly subnormal dispersion found. It’s not clear why one would need to do so when the ratio is 0.869 but in a previous example he says that 0.928 is not very different from 1 and that “according to Lexis, this result affords a confirmation of the hypothesis […]” without further commment. The number of observations is around 20 in both cases, hardly a large number. While it’s kind of ironic to see prominent critic of significance testing suggesting that another critic of significance testing should have seen that the difference is not statistically significance it’s true that he was probably just looking for an example of underdispersion.
> half the population has a boy:girl ratio of 2:3 and the other half a ratio of 3:2
I don’t think anyone was ever going to check that but I intended to write 2:1 and 1:2.
Wouldn’t the posterior be about equal (or say factor of 2 or 3) to that of the binomial model?
Without looking in detail, Its got more free parameters. So it would have flatter likelihood (ie, is consistent with a wider set of observations) but it would also be more accurate.
HJH> 3. Hypotheses, or potential quantified explanations of observations, cannot have probabilities assigned to them.
CH> Also 3 doesn’t hold in general, as there can be situations in which certain hypotheses of interest can be modelled as themselves drawn from a frequentist probability distribution (for example recognition of a letter from handwriting), in which case a Bayesian analysis is fine for a frequentist (also I’d always separate a frequentist concept of probability from the application of “classical frequentist inference” such as hypothesis tests and confidence intervals; one can be a frequentist about the meaning of probability without being committed to tests and CIs).
Given that item 3 was originally about assigning probabilities to hypothesis it clearly was about “frequentist inference” not about “frequentist probability”.
CH> And re item 3: Von Mises, Neyman, and other frequentists were fine with doing Bayesian calculations in setups where outcomes would be generated at various levels from processes to which frequentist probabilities apply (which is arguably the case for letter recognition from handwriting).
Von Mises was of course frequentist about probability but definitely not about inference. His book doesn’t mention confidence intervals, statistical significance or Student’s t or chi-squared distributions. He makes no reference to Neyman or E. Pearson and he’s not kind to Fisher.
“From our point of view, there is no doubt that the likelihood is a correct measure of the probability of inference in two cases. First, when we have reason to assume that the unknown initial probabilities are uniformly distributed over all the positive values of p […] Second, if we know nothing about the a priori probabilities but, if the number of experiments n is very large, then we know that the influence of the initial probabilities is not very considerable […] What meaning the likelihood could have in any case other than the two just described is inconceivable to me. I do not understand the many beautiful words used by Fisher and his followers in support of the likelihood theory. The main argument, namely, that p is not a variable but an ‘unknown constant’, does not mean anything to me.”
“In fact, unless one of the two cases just described should prevail […], we would expect that no one would even think of applying the theory. Unfortunately, this is not the case. Fisher emphatically avoids all reference to Bayes’s solution of the problem of inference; this is for him a matter of principle. Therefore he cannot admit that, unless we have some information concerning the initial distribution, his notion of likelihood is applicable only in cases of a large number of experiments. […] We can only hope that statisticians will return to the use of the simple, lucid reasoning of Bayes’s conceptions, and accord to the likelihood theory its proper role.”
“The Anglo-American statistical theory based on Fisher’s notion of ‘likelihood’, which rejects Bayes’s solution of the problem of inference provides meaningful results in two instances […] The so-called small-sample theory is to be completely rejected.”
I’d just like to emphasise once more the distinction between interpretations of probability and approaches to doing inference. For me the term “frequentism” is best used regarding interpretation of probability. This is what the term refers to, this is what the old foundational texts by Venn and von Mises are about. Also, I distinguish epistemic probability from Bayesian calculations (Bayes himself was ambiguous about the interpretation of probability and the strong connection between epistemic probability and Bayesian reasoning was established later).
If anyone wants to set up a system of axioms for frequentism, for me it’d be essential that this preserves the distinction between interpretation of probability and inference (as correctly noted, von Mises, who is a frequentist by anyone’s standards, didn’t promote Neyman-Pearson “classical frequentist inference”). Setting up the axioms in such a way that this distinction is blurred can’t lead to anything good in my view.
Also, the discussion whether von Mises and Neyman “get Bayesian inference wrong” (meaning connected to epistemic probability) is irrelevant to whether they were fine to use Bayesian calculus for problems where there is a two-stage probability setup with both stages modelled by frequentist probabilities.
HJ – Thanks for the explanation. What made me say I have trouble not thinking about atheism as like assuming a point null hypothesis is the interpretation of it as an argument against all gods — in your analogy, the atheists that are straight up authoring their own papers claiming no version of a God exists. To me, the notion of “God” has never seemed easy to pin down, i.e., different people can have very different beliefs about what that might mean, from the single creator god to eastern philosophies that emphasize no separation between the divine and the non-divine. So to imply that one has amassed evidence against all of them strikes me as about as realistic as expecting two differents things in the world to be exactly equal. However, it sounds like many atheists are opposing specific narrower conceptions, in which case atheism is not so much asserting there is nothing like a god, it is only asserting a lack of evidence to convince the atheist that there is.
I agree, the goal in science is to depersonalize axioms and think about them critically, the opposite of taking any of them on faith. But I don’t believe one is ever absolved of making any commitments in science. There is always ontological uncertainty about the method by which we should be trying to understand the world. To some, scientific methods can be applied to matters of faith to prove them wrong. To others, that kind of thinking itself implies a kind of omniscience they don’t agree with.
> To me, the notion of “God” has never seemed easy to pin down
You are playing whack a mole: We say, What exactly do you mean by god? You say X. We say the evidence is overwhelming against X. You say, what about Y? We say the evidence is overwhelming against Y. You say, what about Z? I’m sorry, but the evidence is overwhelming against all of these.
We didn’t give up gods and develop science just for fun. Science works. The god hypothesis doesn’t. So, we are stuck with science. You can’t change reality just by wishing it were different.
Thanks for the reply back! Nowadays, it’s rare that I get to cut loose about atheism and religion.
> However, it sounds like many atheists are opposing specific narrower conceptions, in which case atheism is not so much asserting there is nothing like a god, it is only asserting a lack of evidence to convince the atheist that there is. … To some, scientific methods can be applied to matters of faith to prove them wrong. To others, that kind of thinking itself implies a kind of omniscience they don’t agree with.
It’s more sophisticated than that. Michael Martin’s “Santa Claus Argument” states that you’re justified in believing something doesn’t exist based on a number of criteria, the three most important of which are that insufficient evidence exists to prove that thing exists, that if a thing exists there should be evidence to prove it exists, and that I posses sufficient evidence to prove it exists if it does exist. There’s no shortage of religious people who believe in the existence of a god, and the vast majority would say they have solid justification for that belief, so I should have abundant evidence or arguments to demonstrate a god’s existence, and those arguments should be of sufficient quality to be convincing. And yet despite actively seeking out these justifications, from people who claim to be experts in them, every single one has failed to merely increase my belief in a god. I’m more justified in saying “there’s no further evidence out there that could convince me” than “well, maybe I should remain agnostic because the next person could give me a convincing argument.”
Science rests on the same inductive bedrock. You can drive the p-value or likelihood ratio or Bayes factor of observing a black swan past any arbitrary threshold, provided you don’t go near Australia, thus you can always “reject” the hypothesis that some swans are black. But once you visit there and start counting, your “rejection” is rescinded. That “rejection” was never of an absolute or objective truth, but instead a “rejection” of justified belief, a subjective conclusion about the world. The only claims to scientific objectivity that are even plausible are over epistemology, or the method you use to determine if a belief is justified, and in my opinion Lindley’s 1957 paper “A statistical paradox” shattered even those.
As further proof of the equivalence, try turning the “Santa Claus Argument” against a cancer treatment. If it reduces the odds of people dying from cancer, you should be able to show that fewer people die from cancer when given the treatment than when they haven’t. If you can’t show that, given ample opportunity, then I’m justified in thinking the treatment doesn’t work. The “Argument” is really just a subset of the scientific process, so if you think the scientific process allows me to justify belief then I can apply the “Argument” to conclude I’m not justified in believing in God. And if you think I can invoke the “Argument” against a god, then I’m justified in invoking the scientific process against claims about a specific god, absent a strong counter-argument that distinguishes between the two and shows the portion absent from the “Argument” cannot be used against such claims.
(Or, if you want the short version, what David J. Marcus said.)
> To me, the notion of “God” has never seemed easy to pin down, i.e., different people can have very different beliefs about what that might mean, from the single creator god to eastern philosophies that emphasize no separation between the divine and the non-divine. So to imply that one has amassed evidence against all of them strikes me as about as realistic as expecting two differents things in the world to be exactly equal.
Ironically, one of the atheist arguments I’ve been developing hinges on religious disagreement. The short version goes like this: suppose a friend casually mentions that Sue has adopted a dog, specifically a large, ailing German Shepard. In another conversation, another friend mentions that Sue has adopted a tiny Chihuahua puppy. A third friend mentions Sue has adopted a middle-aged Bulldog. And so on. Now, what is most likely: 1. Sue has adopted a German Shepard, 2. Sue has adopted a dog, or 3. there is no dog?
If two or more people’s claims about a property that something possesses clash, then at minimum it is justified to be skeptical of those claims. Since the only consensus here is that Sue has adopted a dog, option 2 seem like the most justified option. But the term “dog” is defined by the properties all dogs posses (eg. barking, wagging tails), and if Sue did adopt a dog it should have a consistent set of properties. The lack of consistency thus makes option 3 the most justified, despite all your friends asserting the dog exists.
It’s still a work in progress, but you get the gist.
>If two or more people’s claims about a property that something possesses clash, then at minimum it is justified to be skeptical of those claims.
I have a balloon. One person says its red, another green, and another blue. Do you doubt the existence of the balloon? Perhaps the people are colorblind. Perhaps the people are unfamiliar with dog breeds. There is still a balloon. There might still be a dog. Should I doubt the something, or should I doubt the ability with which people can detect or describe its properties?
Fascinating! The original metaphor invoked dog breeds, which consist of (at minimum) fur colour, pattern, and character; limb, face, and ear morphology; as well as behaviour and likelihood to certain diseases. Your counterexample swaps out all those dimensions for one, colour, which is also notorious for being unreliably perceived. Yet despite the attempt to vastly water down the scale of disagreement, it still proves my point. You can invoke colourblindness or an optical illusion (think “the dress”), but both are quite rare. Colourblindness is rarely total, as well, and the most common type tends to confuse red and green but has no problem differentiating either from blue. Three people could disagree over whether a balloon was Brick Red, Mahogany, or Wild Watermelon, which are all similar shades of red, but three people disagreeing over the three very different colours that form the foundation of many colour systems? That most people have been exposed to thousands of times? We’re still justified in thinking something funny is going on here, and that there is no balloon.
You might want to consider going in the opposite direction, instead: suppose a friend casually mentions Sue has adopted a dog, specifically a large German Shepard that’s had a rough upbringing and is now dealing with diabetes late in life, as well as missing canine teeth and failing vision that makes it fearful of strangers but it nonetheless still manages to rouse itself and nuzzle Sue when it hears her come home from a long day at work. Rather than strip away or water down dimensions of disagreement, pile them on and see if you start to doubt whether or not there’s a dog.
This is interesting. It makes me wonder what you’d say about the existence of feelings, which seem high dimensional. If we hadn’t developed language for talking about sadness, for example, perhaps everyone would describe it differently. Even with established language, there are still zillions of ways to describe what sadness feels like, and perhaps there are zillions of ways to feel sadness. No one doubts the existence of their sadness when they can’t show that others experience sadness in exactly the same way. Does this mean that sadness doesn’t exist?
Colorblindness is not quite rare, at around 8% in men….or perhaps we would not use the same verbal description upon seeing the “8% in men” figure, which could lead a third person to doubt the condition’s existence? ;) I actually have a mild red-green colorblindness. If the balloon was green, and you had a person with red-green colorblindness, one with blue-green colorblindness, and one without, then you might plausibly get 3 colors as stories. But in any case, it doesn’t matter. It seems that your point is that people’s descriptions of an object that are not precisely the same give evidence that the said object does not exist. I provide a pretty mundane counter-example to this with a very plausible reason as the lack of precision in description.
Let’s pile on the descriptions of the dog into a story about the dog. We hear different versions from three friends. If the versions are slightly different, we might suppose that not all friends heard it from Sue directly but heard it second, third, or fourth hand. Have you ever played the telephone game? Or if we knew that they all visited Sue at the same time, then perhaps they all remembered different things, or if the stories fantastically differed, perhaps they were playing a joke on us and then we might be skeptical. But it all depends. Only certain scenarios and information would tend toward our believing that the dog simply didn’t exist at all.
I agree that a lack of precise agreement in the descriptions could well make one skeptical about every detail of each description, but it would seem to only justify skepticism about the existence of the object itself under certain other conditions.
> jd: If the versions are slightly different, we might suppose that not all friends heard it from Sue directly but heard it second, third, or fourth hand.
Very true! My argument does fall apart under those conditions. But remember, this is really an argument against the existence of any god. Many people don’t justify their faith because they heard about God from a friend of a friend of a friend, they instead have felt or experienced the effects of God directly. I have plenty of arguments that work against the former, but as the existence of agnostics implies the latter is far more difficult to argue against. Likewise, when you argue that people could be mistaken, that drags us back to empiric counter-arguments for the gods, where I’m most comfortable arguing from.
So within this analogy, we’re assuming your friends met with Sue and interacted with the dog directly. We’re also assuming they come to their beliefs honestly, and are unlikely to claim something they don’t believe to be true (though we leave open the possibility of honest misinterpretation of their first-person experience). This rules out pranks, as well.
> I agree that a lack of precise agreement in the descriptions could well make one skeptical about every detail of each description, but it would seem to only justify skepticism about the existence of the object itself under certain other conditions.
It’s those “certain other conditions” that I’m trying to nail down. I’ve outlined when I think they occur, and I’d be curious about when you think they do.
> Jessica: No one doubts the existence of their sadness when they can’t show that others experience sadness in exactly the same way. Does this mean that sadness doesn’t exist?
Not under this argument, no. What differs between the dog and emotions is that the dog is external to and separate from any one person. My “sadness” may be divergent from or even contradictory to your “sadness,” but all the evidence for what I mean by that term comes directly from me. There is no secondary source, and thus no room for contradiction via honest direct experience.
We could rescue the gods from this argument by pulling the same maneuver, but not only are we invoking solipsism we’re directly contradicting many faith traditions, most sects of Christianity included. Your confusion over how anyone could be an atheist wouldn’t make much sense, either, as my lack of faith simply means God doesn’t exist for me but does exist for you. There’s nothing contradictory about that, under this rescue.
in the Bush years, sophisticated coastal elite types tended to have negative views of religion especially Christianity. Eastern spirituality was more OK but people often really wanted to emphasize it was non-theistic.
nowadays when someone is very anti-religious it comes across as Midwest/suburban/transplant/Reddit. but if you’re in NYC or SF, Western religion is one among the other pleasant and acceptable forms of spirituality, even if there are relatively few orthodox believers.
I think this probably drives the larger openness to spirituality, Christian or not: in a lot of places the forces pushing people to embrace an atheist identity aren’t there anymore.
of course part of the reason for this split is that if you’re in a Midwestern suburb and move to a big city, you might be a type of person who was a big or small problem for evangelical Christians in that suburb. I expect if the national political power of the Christian right returns, the negative attitudes will also return.
I grew up in a largely religious country (India) so I took all that stuff seriously at first, but later in life came to think that religion is basically a result of people trying to attribute causality and agency to randomness in data. You look up at the clouds and see a face, it’s tempting to think that there’s meaning there. One can then start to build up an increasingly elaborate theory around external agency to explain what is basically noise. Add to that the terror of inevitable death, and the lucrative money-making opportunity by separating people from their money, and you have religion. Not to mention the opportunity to sexually exploit helpless children.
One irony is that in science we stick to beliefs (a particular theoretical position) out of some kind of misplaced loyalty–to an idea one had had, to one’s former advisor’s beliefs, etc. It’s rare to see a scientist change their belief when confronted with overwhelming data that is inconsistent with their current belief. Just pick some random scientist and study their career trajectory–in most cases, you will find that they magically find support for their beliefs, decade after decade. So in a way scientists adopt a religious attitude even when trying to be rational.
Hi Jessica, thanks for the interesting post.
As a hardcore Bayesianist and hardcore atheist (born in USSR) let me suggest some comments
1) I do acknowledge what I’m not making rational, weighed decisions all the time. Moreover, I deliberately avoid rational cost-benefit analysis for the most important decision in my life. I got married because emotions of love, not because careful analysis in Bayesian framework. I got into science because I like science, not because minimizing the loss function told me to do so.
Original meaning of the word “Religio” mean (as far as we can tell) human relations to things this human consider important (not necessary supernatural). One can feel “religio” towards Gods, but also toward family, friends, neighbours, pets, art, state, Lord of the Rings, Maxwell equation etc..
I certainly feel that some things are more persistent / important than me, and that some things are better to be observed with closed eyes rather than through the microscope of rationality. Art is one of these things. But (a) I just dont have any need to assign agency or consciousness to these important things (b) I dont claim that these things are fundamentally mysterious and cannot be analyses through rationality at all. Other people can assess my marriage or my career path or my taste in music with analytical toolkit. I just refuse to do so out of respect.
This is interesting. As I mentioned in replying to HJ above, my comment about the point null hypothesis was based on an understanding of atheism in the narrow sense of asserting that there can be no God.
I like the original meaning of “religio”, because I think of religion as closer to a personal value system than some of the more canonical associations that word carries.
After reading some of the descriptions of atheism that commenters have posted, I am thinking that at least for me, the question of what stance to take on religion is a matter of no free lunch: there will be circumstances where calling myself religious seems unduly restrictive or creates a false sense of disagreement between me and those who consider themselves non-religious, and situations where calling myself uncertain (whether you want to label that atheism, agnosticism, or something else) seems unduly restrictive. I’ll probably continue to default to considering myself religious because I value my own religious practice. And maybe because in the circles I hang out, identifying as religious is less expected
It helps to say what your words mean, if you don’t want to use their standard meanings. A belief in god is separate from a personal value system, according to the usual meanings of the words.
“Atheist” = “a person who does not believe in the existence of a god or any gods” according to Merriam-Webster. If that wasn’t what you were discussing, you should have used a different word.
The evidence that gods don’t exist is overwhelming. So, it isn’t surprising that someone you met described themselves as an atheist. Of course, religious belief is not necessary for moral philosophy.
Of course, one can enjoy the rituals and practice of a religion or culture without believing things that aren’t true. But, if you believe things that aren’t true, that is a delusion.
“religion as closer to a personal value system than ”
Religion has a horrendous history as a value system, personal or otherwise. From Buddhists* attacking the Tokyo commuter system with nerve gas to Catholic priests abusing children (and being protected by their higher-ups). Heck, even the Quakers got in on the game with their indigenous boarding schools**. My family was also in on the game, being on the prosecutorial side in the Salem witch trials (my mom claimed; don’t have a reference). Did I mention the Crusades? Or the “troubles” in Northern Ireland.
Seriously, the idea that religion is moral in any sense is seriously contraindicated by a lot of painful reality.
*: Maybe that was an exception, you’re thinking. Wrong. Many Buddhist temples were well-armed military organizations, some terrorized their local comminities. See Houjouki’s description of maurading Buddhists (well, you may need to read the original, because that passage was not included in the one English translation I saw, and not mentioned in the Wiki article). To say nothing of the Chinese Shao Lin movies.
**: https://www.fcnl.org/updates/2021-10/quakers-grapple-legacy-indian-boarding-schools
There is lots of nasty stuff in the bible. Perhaps the fact that many religious people believe things that are false (e.g., gods exist) contributes to their having bad morals. Perhaps the belief that god is on your side can reduce your empathy for those who do not share your false belief. Maybe not. But, I think a firm grasp of reality is a better basis for a moral system.
(Continuing)
2) Andrei “the maker of chains” Markov was a famous atheist, who suggested that principles of probability should be applied to religion and God as well. He was later excommunicated from Russian Orthodox church due to his support of Leo Tolstoy.
So lets continue this (Makovian) like of reasoning by applying Bayesian toolkit to the questions of religion. Not necessary the path I would take, but I was not the one who started by comparing atheism to a point null hypothesis =)
First of all, if you accept Cox theorem, then all need to accept that all your believes can be expressed as probabilities. Does this mean than one’s degree of faith can be expressed as p(God exists)? So if your p(God exists) is less that a certain threshold (say 0.05, the forever damned number) your are an atheist, if it is >0.95 you are believer, and your are agnostic otherwise.
But of course, “God exists” is extremely context-dependent statement. And it only make sense at all in a very narrow modern western post-christian world. What do you mean by God? What does you mean by exists? Does believing in the existence of the particular God is that defines religion, or is it about the way these believes are expressed?
Ok, lets try another way. Let \Omega be the possible set of all ontological statements about the world (Christianity, Gnosticism, Kabbala, Pantheism, atheistic big bang etc etc). Let \Omega_r be the subset of \Omega, the set of ontological statements considered religious (so Gnosticism is in \Omega_r but Big Bang is out, some statement would be borderline and could be discussed). So one’s religiosity is better expressed as
\sum_{w \in \Omega_r) p(w)
It is clear, however, that this expression would be intractable since \Omega is very hard to define, and assigning p(w) for each w \in \Omega would be an impossible task.
For the sake of computation, we may approximate \Omega with \Omega_a, the set of all ontological statements one have consistently encountered to form an opinion about them. E.g. for a general lay westerner, \Omega_a may consist of {Christianity, Islam, Judaism, Buddhism, Atheism}. We can then define a proper probability distribution on this reduced space, but the value of such approximation is highly dubious.
***
One think that is interesting to analyze in Cox’ framework is “Credo quia absurdum”, Tertullian’s “I believe because it is absurd”, a phrase widely used by christian as a defense against rational nitpicking.
At first it seems that “Credo quia absurdum” goes directly against the Cox theorem. But actually they can be accorded.
Indeed, if one starts with a prior p(X) = 1, one will end with p(X| data) = 1, and no matter of evidence against X will change this. If one define faith as p(x)=1, than it should remain unchanged againt all evidence, which separate it from mere p(x)=0.99999 believe.
Jessica: How did you maintain your incorrect belief growing up and going to school and college? Did you avoid taking science
classes? When the physicist you were dating politely pointed out your error, did it occur to you that you might have made a
mistake?
> I wonder if things would be different if we were more comfortable as
> scientists with acknowledging that having irrational commitments to
> certain ideas is part of what we do.
Of course, that is not what good scientists do.
David:
That’s some Dawkins-level obnoxiousness right there!
lol, it sure is.
You must be using a different meaning of the word “obnoxious” than the dictionary gives.
You didn’t answer my questions. But, maybe you already gave the answer when you wrote, “having irrational commitments to certain ideas is part of what we do.” I can imagine how this belief could lead to other irrational beliefs.
David:
Regarding whether your comment is obnoxious, it depends on what definition you use!
From Merriam-Webster: “odiously or disgustingly objectionable : highly offensive.” I don’t think your comment is obnoxious under that definition.
From Cambridge: “very unpleasant or rude.” I’d say that your personalization of Jessica’s remarks was rude, also it seems very unpleasant to say that Jessica is not a good scientist.
From American Heritage: “Very annoying or objectionable; offensive or odious.” Was your comment very annoying? I’m not sure. It seemed annoying, but maybe not very much so. Was it objectionable? Kinda, yeah, I think it’s objectionable to take a philosophical discussion and move it to personal level.
But I don’t want to overstate this. There are worse things than being obnoxious, and just about all of us are obnoxious from time to time–I know I am! Overall I appreciate your comments.
Perhaps it comes down to whether you consider zealotory obnoxious. When I can no longer distinguish someone’s argumentation from this, I generally stop engaging.