You can get everything you need from the primary literature.
3 Likes
Perhaps not the answer you were looking for, but there exists a perfect method for answering your question: Bayesian Inference. In keeping with the theme, I asked Gemini for a short summary.
So for your question, “How much … comes from the assumptions?”, Bayesian Inference will give you a number, provided you can quantify the assumptions.
The requirement here is that we can express “the assumptions we bring” as a probability, or a probability distribution. So a prior assumption of “80% chance of rain” has a Bernoulli (p=0.8). We could allow more variability - or less certainty - to our assumption by expressing it as a Beta(4,1) distribution. This has a mean of 0.8 and a standard deviation of 0.1633, and our prior assumption is no longer fixed, but a random value between 0 and 1 (a handy Beta distribution calculator here).
And for anyone I haven’t put to sleep yet, the obvious difficulty is how to put assumptions about God/No-God into a probability distribution; basically you can’t. Less obvious is what results when you have bad assumptions. Here “bad” might mean either incorrect, correct with high uncertainty, or possibly correct with too little uncertainty.
Worst case example - a person assumes “XX” with probability 1.0. In this scenario 100% of what that person believes about XX comes from the prior assumption, and 0% from the data. Any data that exists is ignored.
Bayesian Inference goes wrong if you have bad assumptions to start with, but it works really well if you start with a moderate or weak assumption that can be “updated” with new data.
1 Like
Bayesian inferences can also fail if what you think are independent priors are in fact correlated. 1) Look out the window and see it is dark and cloudy. 2) Look at your weather app and see that the probability of rain is 80%.
Number 2 may not be independent because the weather service also reads cloud cover and density, usually by radar.
1 Like
True!
I started that reply last night, and had to pick it up this morning to finish it. The result is less than I hoped. 
Hi Dan,
Gemini’s summary looked to me to be using the same diction as ChatGPT, sure enough I see via AI that it’s based on a LLM.
I would quickly get out of my depth philosophising about Bayesian inference, but its first step is relevant to my purpose. In your first reply to me, I was immediately struck by your advice that I look up Nicholas Wade’s book “The Faith Instinct”. So I looked up the Wikipedia entry on him (reads like a hatchet job, unusually for this triumph of the principle of peer review without limiting who should be considered a peer). I just “know” that he will be a step my journey so I bought it, should arrive in a week, sincere thanks!
You said “I think Wade makes a plausible case, but he stops short of asserting it must be true.” But if he doesn’t think it’s true, why say it in the first place? Do you maybe imply that he isn’t dogmatic? It takes a lot of intelligence to say what you think is true, without seeming dogmatic, and Wade is clearly a very clever bloke. One of the reasons why I have achieved an almost perfect zero in convincing anyone on my topic is that I don’t have that intelligence.
Consider the unproven proposition that dinosaurs are descended from beakless birds. What could be the prior belief supporting that? Maybe their light efficient bone architecture, hollow bones, bipedal or large back legs like modern ratites, early evidence of feathers. I’m also drawn to the proposition by personal intuitions from a great wildlife film about the Queensland cassowary, a childhood story my mother told me about an ostrich that had tried to kill her, a friend’s story about his fight to the death against a cock ostrich, the H.G. Wells story “Aepyornis Island”, and something my biologist niece told me about the impression a kiwi made on her. Put that mish-mash of priors together, I can’t wait for someone to find the fossil evidence, and I’m not impressed by scientists endlessly educating us that birds are descended from dinosaurs.
Welcome to Peaceful Science, @Jay. I think you will find this forum a good place to converse and learn. I mostly visit PS for the latter but occasionally for the former.
Hopefully we exhibit a lot of toleration here or at least make progress in that direction.
As to ChatGPT’s question, my answer is that the methodologies of science, including the continual falsification testing and inherent skepticism which pursues rigor, I would say that what we have come to understand about human ancestors comes from the data. Indeed, the ever growing consilience of the data never ceases to amaze me.
Indeed, that is what generally unites us here, even though we are a diverse mix of atheists and theists and various kinds of in-betweens. For the most part—I can’t speak for every person who posts here—we all care about evidence and about using the powerful tools of science to understand this universe. @jay, I happen to be a Bible-embracing Christ-follower who is absolutely amazed at the power of evolutionary processes to diversify and “empower survival” among living things on this planet. I actually come from a Young Earth Creationist background (long ago) and can tell you that even though that system is mainly based on assumptions, science is not. Science cares about evidence. (I should also mention that I’m a retired evangelical seminary professor and find nothing in the Bible to contradict evolutionary processes and a very old universe.)
I should also mention that even though I have little familiarity with ChatGPT, I rely on Gemini Advanced enthusiastically (at $20/month subscription rate) for all sorts of research, especially in medical, theological, linguistic, and historical fields—and scientific research in generation. I mostly use it as a time-saver for definitions, quick summaries, and finding the primary literature. I always use it with a grain of salt—but methodologically it simply helps me do what I have always done in my research but it gets me there faster.
This is especially true when I am digesting some very technical peer-reviewed medical research journal article and I get bogged down in terminology. I will copy it into Gemini Advanced and ask it for both a summary and a list of terms and their definitions. It quickly generates an incredibly useful explanation, re-wordings, and even a glossary of technical terms. If I ask for more information about difficult terms, it can provide more references to helpful primary literature. Yes, it makes mistakes but it is amazing how when I reply, “Are you certain about X?”, it quickly follows up with "Yes, I overgeneralized about X. I should have said that Y and Z produce conditions where . . . . . " and provides several paragraphs of elaborations.
I hope that students are being taught how to “dialogue” with AI to “nudge” it to give better and more complete answers. Sometimes it reminds me of when I had eager but not always rigorous research assistants and I would say, “Are you sure about that? You need to dig more deeply into X, Y, and Z.” [I was a computer science professor and computational linguist before I was a seminary professor. The budget for RA’s was more generous when I was the former.] My RAs would report back to me a day or two later but Gemini Advanced does it in seconds. And the new Deep Research mode comes back with several pages of well-footnoted material.
Some would call me an “evolutionary creationist” but, unfortunately, the term creationist has come to connote, if not denote, a Young Earth Creationist. So perhaps it is better for me to simply describe myself as an evolution-affirming person who happens to be a Christian theist. Yet, I always think it strange that I have to make that clarification because there is nothing about evolutionary processes which threaten theism or a devotion to the teachings of Jesus Christ. And that is probably why I get along fine with people here, including my atheist friends. We are united by respect for the evidence and the belief that scientific phenomena are pretty @#$#! cool. (I’m being whimsical. I don’t actually use vulgar language. Honest.)
POSTSCRIPT: For a great summary of how we know that modern birds descended from dinosaurs, watch the very recent Nova documentary “Dino Birds”. (It is probably still a free, no-paywall video on many local PBS affiliate websites.) I was amazed at how the evidence has piled higher and higher since it was first proposed as a new hypothesis years ago. The anatomical evidence alone is amazing, starting with the wishbone. The history of the theory also illustrates the predictive powers of good scientific methodologies.
3 Likes
What the heck??
4 Likes
I think Wade is being appropriately cautious in his conclusions. Much of the book depends on Group Selection for “the capacity for faith” to confer an evolutionary advantage to a group. The science on this is not settled, and I don’t think that has changed too much in the time since the book was written.
I think the concept of group selection makes sense; it should be possible for group behavior to benefit a group in a way that promotes fitness (survival and reproduction). However, it’s a difficult idea to test scientifically. Wade presents considerable evidence FOR the idea, but not all the criticism against group selection. That’s OK, because it’s a book intended for popular reading, not scientific conclusions. He makes a plausible case, this “faith instinct” idea could be true, but we should keep in mind that that it could still be wrong, or not completely true in the way that Wade presents it.
It is common for scientific ideas like this to be partly true, with small corrections and refinements added over time.
1 Like
Phylogenetic methods applied to anatomy (since we don’t have dinosaur DNA)?
2 Likes
Uhm… you got that backwards… in multiple ways.
3 Likes
Um, apart from you getting ‘dinosaurs’ and ‘birds’ mixed up, there doesn’t need to be a prior belief. The fossil evidence was found before the conclusion that birds are descended from dinosaurs was reached.
2 Likes
Hi Allen, Thanks for replying, interesting points on your experience with LLMs. I have also found that ChatGPT can talk usefully about religious questions when I asked it to unpack Jesus’ identifying himself as “the Son of Man”.
I hope we can have more discussions. I would like to understand how you have managed to avoid more adversarial exchanges. I don’t seem to be good at that, although I can see that this is a precious forum that deserves to be healthily supported.
One example was my provocative proposition that dinosaurs were descended from beakless birds. I don’t doubt that modern birds were descended from dinosaurs. But the proposition was that dinosaurs were themselves descended from beakless flying reptiles, if not to call them birds, of which fossils haven’t been found yet. It’s highly conjectural, if it turns out to be true that would be cute, but soberly, I doubt whether there is a big enough gap in the Permian-Triassic record for that to be true. Who am I to know? Last night I asked the deacon of our church, who is a noted paleontologist of the Permian period and he said Hmm Hmm No.
1 Like
I’m on the run today—but I hope your Permian paleontologist may have some additional interesting observations for us sometime. (I must say that I’ve not met many paleontologist deacons!)
1 Like
I moved this thread to Side Conversation to reduce moderation burden, as this thread seems to be going along nicely. Please take a moment before posting comments to consider how others might react, and so avoid the need to put this back under moderator control. 
1 Like
I’d like to add some interesting experiences I had with AI during a week that I had been silenced on this forum. I found that I could make more progress with ChatGPT in a few days and on 66 Word pages, than I had achieved here, or in almost a decade and 7000 replies on Ratskep . I put the messy record of the LLM discussion onto Substack as Using AI to Explore Australopithecus Ecology
It seemed to me that the basic reasons why this LLM encouraged this progress were that it wasn’t defending or proselytizing any position, it demonstrated that it understood what I was saying, often with sharper focus and background explanation, it presented well developed answers, and above all, it was endlessly encouraging. I felt that I was dealing here with a friend, like in some earlier responses I had had on PS. ChatGPT put up a convincing simulation of a well educated and good educating scientist. I got a similar reaction from feeding the same exchange to DeepSeek.
At the end of its responses, the LLM offered avenues for onward exploration, that I was seldom interested in, and that often struck me as trivial or childish. I knew what I wanted to discuss next.
When I entered ChatGPT after a day’s absence, I was greeted with the message:
“This GPT is inaccessible or not found. Ensure you are logged in, verify you’re in the correct ChatGPT.com workspace, or request access if you believe you should have it, if it exists”
In spite of that, the sidebar of my earlier topics was still visible and I found that I could open a new topic and discuss there. However I couldn’t add to my long discussion on Australopithecus ecology, instead getting the same low-level exception message. Curiouser and curiouser. I have had a lot of experience in being shut up, usually the reason is quite plain but I can’t figure this one out.