What do you think?
Rate this book
God from the Machine: Artificial Intelligence Models of Religious Cognition
"God from the machine" (deus ex machina) refers to an ancient dramatic device where a god was mechanically brought onto the stage to save the hero from a difficult situation. But here, William Sims Bainbridge uses the term in a strikingly different way. Instead of looking to a machine to deliver an already known god, he asks what a computing machine and its simulations might teach us about how religion and religious beliefs come to being. Bainbridge posits the virtual town of Cyburg, population 44,100. Then, using rules for individual and social behavior taken from the social sciences, he models a complex community where residents form groups, learn to trust or distrust each other, and develop religious faith. Bainbridge's straightforward arguments point to many more applications of computer simulation in the study of religion. God from the Machine will serve as an important text in any class with a social scientific approach to religion.
188 pages, Paperback
First published January 1, 2006
30 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 3 of 3 reviews
May 15, 2024
super fucking strange
//////
Here is someone who did a paper on the book that's basically an extended review
//////
Artificial Intelligence & Religion: A Review of Bill Bainbridge’s God from the Machine: Artificial Intelligence Models of Religious Cognition
Cognitive Systems Research 9(3):232-235
June 2008
M. Afzal Upal
James Madison University
//////
Perhaps, wanting to save the best for the end, Bainbridge leaves belief formation and communication for the last two chapters of the book. Chapter 7 is a modified version of Bainbridge (1995) which describes an agent-based simulation with each agent powered by a simple neuralnet.
Each agent desires five commodities (called energy,water, food, oxygen, and life). Agents are randomly assigned these resources at the start and lose 1 unit of each resource in every simulation round.
Agents are divided into four different groups depending on the kind of commodity they can produce.
Water and food producers need 1 unit of energy to make 2 units of water or food.
Oxygen producers need both 1 unit of water and energy to make 2 units of Oxygen.
Energy producers are the only ones who do not need anything to producer energy.
No one can produce life.
Agents decide what to seek in each round by seeing what they lack the most. Having decided what to seek, they also decide what group of agents to seek it from. The learning decision is to learn how many groups there are (num_groups) and to learn the associations between resources and their producer groups.
However, even when some agents come to believe that there are less than four groups (i.e., group_num = 1, 2, 3) they are still programmed to try and trade with agents from groups 4 4–num_groups.
It turns out that since no actual agent can produce life, life production comes to be associated with these phantom groups which Bainbridge labels supernatural.
This simulation according to Bainbridge explains how people come to form religious beliefs.
Religious beliefs are special algorithms for obtaining rewards such as eternal life which cannot be obtained.
What makes these algorithms special is that they cannot be confirmed or disconfirmed, e.g., one would have to die to confirm/disconfirm the belief in life after death.
This theory, originally laid out in Stark and Bainbridge (1987), may or may not be correct. The problem is that the simulation does not seem to tell us anything we did not know before the simulation.
***** (1)
Agent-based social simulations are useful tools for social theory development when they result in the emergence of results that could not have been foreseen by the researchers before spending considerable time and effort to design and run simulations and to analyze the results.
The problem with all the simulations presented here is that they do not tell us anything we did not know before.
Agents in simulations of Chapter 8 were specifically programmed to favor the theories about the presence of phantom groups and the fact that they exhibit this bias does not tell us anything new about Bainbridge’s theory of religion.
****** (2)
Similarly, relabeling red and green agents Catholics and Protestants and rerunning Schelling’s simulations does not tell us anything that we haven’t known for 40 years.
There is however, another deeper problem with using traditional agent-based social simulation systems to simulate religious phenomena.
Traditional agent-based social simulation systems are designed based on the keep-it-as-simple-as-possible principle.
*****
The idea is that if complex social patterns can emerge from a simulation employing agents with simple decision making and agent-interaction rules and extremely limited memory (e.g., 1 or 2 bits) then it is easy to compute the causal links between the micro-level cognitive processes and macro-level social patterns.
***** (3)
The problem is that what makes religious beliefs interesting and religious is the very fact that they are richly connected with other religious and non-religious beliefs.
Such richly connected beliefs cannot emerge from a society of agent whose memory capacity is limited to one-bit (Doran, 1998; Epstein, 2001).
***** (4)
A reformulation of traditional agent-based social simulation approaches is needed to allow us to model complex cultural phenomena such as217 the formation and propagation of religious beliefs (Sun,218 2006; Upal, 2007; Upal & Sun, 2006, 2007).
In order to have complex shared beliefs emerge at the societal level, individual agents need to be able to represent such beliefs and be able to acquire and modify them
*****
To design predictive computational models we need to design agents that can models cognitive processes of information comprehension, information integration/belief revision, and communication.
My students and I have designed one such multiagent society called CCI (Communicating, Comprehending, and Integrating agents) and embedded it into a multiagent version of Rusell and Norvig (1995) Wumpus World Domain (MWW).
As shown in Fig. 1, MWW is an NNboard game with a number of wumpuses and treasures that are randomly placed in various cells.
Wumpuses emit stench and treasures glitter. Stench and glitter can be sensed in the horizontal and vertical neighbors of the cell containing a wumpus or a treasure.
Once the world is created, its configuration remains unchanged i.e., the wumpuses and treasures remain where they are throughout the duration of the game.
MWW is inhabited by a number of agents randomly placed in various cells at the start of the simulation.
*****
*****
Even so our experiments with a version of the society where we disabled communication revealed that patterns of false beliefs emergent in such a society have a particular structure to them; agents are more likely to have false beliefs about wumpuses than about treasures (Upal, 2007).
The reason appears to be that while hypotheses about the presence and absence of wumpus are harder to confirm and disconfirm for the agents than the hypotheses about the presence and absence of treasures.
This is because agents seek the cells where they believe treasures lie but avoid cells where they believe wumpuses live.
*****
This is exactly what Stark and Bainbridge (1987) argued, hypothesis which are harder to confirm and disconfirm are more likely to be believed by believers.
Furthermore, our subsequent experiments have shown that when agents are allowed to communicate with other agents this pattern continues to hold.
///////
fuck me, i actually appreciate someone talking about Hunt the Wumpus
which was one of the best maze games of the 1970s
And why on earth this is being babbled about in a philosophico-religio-sociological-bunko squad of Religion and Artificial Intelligence
is amusing and maddenningly crazy at the same time
///////
//////
you got it
i got zero faith in reductionist sociologists
not much hope in people or religion
and less than zero faith in computer programmers and non-artificial intelligence
maybe David H. Ahl was right, computers are all about fun, and it can be mildly educational.
I think simulations and games are excellent learning tools
but it's strange how i see the sociologists as peculiar as the religious oddballs
//////
Here is someone who did a paper on the book that's basically an extended review
//////
Artificial Intelligence & Religion: A Review of Bill Bainbridge’s God from the Machine: Artificial Intelligence Models of Religious Cognition
Cognitive Systems Research 9(3):232-235
June 2008
M. Afzal Upal
James Madison University
//////
Perhaps, wanting to save the best for the end, Bainbridge leaves belief formation and communication for the last two chapters of the book. Chapter 7 is a modified version of Bainbridge (1995) which describes an agent-based simulation with each agent powered by a simple neuralnet.
Each agent desires five commodities (called energy,water, food, oxygen, and life). Agents are randomly assigned these resources at the start and lose 1 unit of each resource in every simulation round.
Agents are divided into four different groups depending on the kind of commodity they can produce.
Water and food producers need 1 unit of energy to make 2 units of water or food.
Oxygen producers need both 1 unit of water and energy to make 2 units of Oxygen.
Energy producers are the only ones who do not need anything to producer energy.
No one can produce life.
Agents decide what to seek in each round by seeing what they lack the most. Having decided what to seek, they also decide what group of agents to seek it from. The learning decision is to learn how many groups there are (num_groups) and to learn the associations between resources and their producer groups.
However, even when some agents come to believe that there are less than four groups (i.e., group_num = 1, 2, 3) they are still programmed to try and trade with agents from groups 4 4–num_groups.
It turns out that since no actual agent can produce life, life production comes to be associated with these phantom groups which Bainbridge labels supernatural.
This simulation according to Bainbridge explains how people come to form religious beliefs.
Religious beliefs are special algorithms for obtaining rewards such as eternal life which cannot be obtained.
What makes these algorithms special is that they cannot be confirmed or disconfirmed, e.g., one would have to die to confirm/disconfirm the belief in life after death.
This theory, originally laid out in Stark and Bainbridge (1987), may or may not be correct. The problem is that the simulation does not seem to tell us anything we did not know before the simulation.
***** (1)
Agent-based social simulations are useful tools for social theory development when they result in the emergence of results that could not have been foreseen by the researchers before spending considerable time and effort to design and run simulations and to analyze the results.
The problem with all the simulations presented here is that they do not tell us anything we did not know before.
Agents in simulations of Chapter 8 were specifically programmed to favor the theories about the presence of phantom groups and the fact that they exhibit this bias does not tell us anything new about Bainbridge’s theory of religion.
****** (2)
Similarly, relabeling red and green agents Catholics and Protestants and rerunning Schelling’s simulations does not tell us anything that we haven’t known for 40 years.
There is however, another deeper problem with using traditional agent-based social simulation systems to simulate religious phenomena.
Traditional agent-based social simulation systems are designed based on the keep-it-as-simple-as-possible principle.
*****
The idea is that if complex social patterns can emerge from a simulation employing agents with simple decision making and agent-interaction rules and extremely limited memory (e.g., 1 or 2 bits) then it is easy to compute the causal links between the micro-level cognitive processes and macro-level social patterns.
***** (3)
The problem is that what makes religious beliefs interesting and religious is the very fact that they are richly connected with other religious and non-religious beliefs.
Such richly connected beliefs cannot emerge from a society of agent whose memory capacity is limited to one-bit (Doran, 1998; Epstein, 2001).
***** (4)
A reformulation of traditional agent-based social simulation approaches is needed to allow us to model complex cultural phenomena such as217 the formation and propagation of religious beliefs (Sun,218 2006; Upal, 2007; Upal & Sun, 2006, 2007).
In order to have complex shared beliefs emerge at the societal level, individual agents need to be able to represent such beliefs and be able to acquire and modify them
*****
To design predictive computational models we need to design agents that can models cognitive processes of information comprehension, information integration/belief revision, and communication.
My students and I have designed one such multiagent society called CCI (Communicating, Comprehending, and Integrating agents) and embedded it into a multiagent version of Rusell and Norvig (1995) Wumpus World Domain (MWW).
As shown in Fig. 1, MWW is an NNboard game with a number of wumpuses and treasures that are randomly placed in various cells.
Wumpuses emit stench and treasures glitter. Stench and glitter can be sensed in the horizontal and vertical neighbors of the cell containing a wumpus or a treasure.
Once the world is created, its configuration remains unchanged i.e., the wumpuses and treasures remain where they are throughout the duration of the game.
MWW is inhabited by a number of agents randomly placed in various cells at the start of the simulation.
*****
*****
Even so our experiments with a version of the society where we disabled communication revealed that patterns of false beliefs emergent in such a society have a particular structure to them; agents are more likely to have false beliefs about wumpuses than about treasures (Upal, 2007).
The reason appears to be that while hypotheses about the presence and absence of wumpus are harder to confirm and disconfirm for the agents than the hypotheses about the presence and absence of treasures.
This is because agents seek the cells where they believe treasures lie but avoid cells where they believe wumpuses live.
*****
This is exactly what Stark and Bainbridge (1987) argued, hypothesis which are harder to confirm and disconfirm are more likely to be believed by believers.
Furthermore, our subsequent experiments have shown that when agents are allowed to communicate with other agents this pattern continues to hold.
///////
fuck me, i actually appreciate someone talking about Hunt the Wumpus
which was one of the best maze games of the 1970s
And why on earth this is being babbled about in a philosophico-religio-sociological-bunko squad of Religion and Artificial Intelligence
is amusing and maddenningly crazy at the same time
///////
//////
you got it
i got zero faith in reductionist sociologists
not much hope in people or religion
and less than zero faith in computer programmers and non-artificial intelligence
maybe David H. Ahl was right, computers are all about fun, and it can be mildly educational.
I think simulations and games are excellent learning tools
but it's strange how i see the sociologists as peculiar as the religious oddballs
July 7, 2025
One of the most beautiful writings about religion I have ever read.
Most books of this kind really have an axe to grind. Either they are for or against religion. This is the first time I have read one that really feels neutral and fair.
Either Bainbridge is a good storyteller or he just really appeals to my autism. I have had experience with books I find so natural and so easy to read that my friends say are impenetrable while what they say is easy to read I have zero hope of finishing the whole book without torturing myself. This might be one of them? I'm not sure.
True to the anthropological studies he cited in the last chapter, I really did relate to these non existent automatons in his machine learning project. I really feel sorry for them. Or more accurately I really feel sorry for the people they are representing.
Maybe I myself am or have been or were, not sure, religious? Or literary but the story that forms is really heartbreaking and bittersweet. Especially the part about people who cannot get real rewards subsisting on rewards that do not exist instead. Having practiced Buddhism before, there's just something very dependent-origination about the way Bainbridge approaches religion. A lot of the book felt like looking at a virus under the electron microscope kind of nature documentary narrated by Attenborough. At such a scale there is no judgement. These are smart people. These are stupid people. No such thing. Just the strange feeling of looking at a rat struggling to navigate a maze built by scientists and imagining that you are also a rat, albeit a way bigger one, in our own cages and mazes. That feeling. I don't know the name of that emotion.
There's something very meditative about the abstract, formal, mathematical way Bainbridge approaches religion. Like there's many kinds of sympathy. People who bully you, yeah that makes you feel that they acknowledge your existence. People who love you, yeah that's really positive affirmations of my existence. And then there's this book. It's the cold but fair sympathy of "I understand". Zero judgements. Like instead of reassuring your friend that even the most beautiful of girls do pee themselves accidentally, you tell them exactly how the urinary system works and conclude that it could have been a worse injury. Dry pants but bladder requires catherization because it held shut for so long and under so much pressure that it cannot open again without mechanical intervention.
Usually books I finish reading are books that I feel offer some clues about just what the hell it means to be alive. This one definitely has that. At least for me. It's like this whole book was written with Legos.
This book is definitely not for everybody. But then again only the ones that say "New York Times Bestseller" are.
People who want a more traditionally human kind of treatment of religion should instead check out "Violence and the Sacred" by Rene Girard. But I still haven't finished that one because it's too human for me. If you feel more like British humor like haha you bloody idiot then Richard Dawkins got you covered. For a more Christian perspective like Dies Irae, la crimosa then check out Pelican books' Introduction to the Bible.
Most books of this kind really have an axe to grind. Either they are for or against religion. This is the first time I have read one that really feels neutral and fair.
Either Bainbridge is a good storyteller or he just really appeals to my autism. I have had experience with books I find so natural and so easy to read that my friends say are impenetrable while what they say is easy to read I have zero hope of finishing the whole book without torturing myself. This might be one of them? I'm not sure.
True to the anthropological studies he cited in the last chapter, I really did relate to these non existent automatons in his machine learning project. I really feel sorry for them. Or more accurately I really feel sorry for the people they are representing.
Maybe I myself am or have been or were, not sure, religious? Or literary but the story that forms is really heartbreaking and bittersweet. Especially the part about people who cannot get real rewards subsisting on rewards that do not exist instead. Having practiced Buddhism before, there's just something very dependent-origination about the way Bainbridge approaches religion. A lot of the book felt like looking at a virus under the electron microscope kind of nature documentary narrated by Attenborough. At such a scale there is no judgement. These are smart people. These are stupid people. No such thing. Just the strange feeling of looking at a rat struggling to navigate a maze built by scientists and imagining that you are also a rat, albeit a way bigger one, in our own cages and mazes. That feeling. I don't know the name of that emotion.
There's something very meditative about the abstract, formal, mathematical way Bainbridge approaches religion. Like there's many kinds of sympathy. People who bully you, yeah that makes you feel that they acknowledge your existence. People who love you, yeah that's really positive affirmations of my existence. And then there's this book. It's the cold but fair sympathy of "I understand". Zero judgements. Like instead of reassuring your friend that even the most beautiful of girls do pee themselves accidentally, you tell them exactly how the urinary system works and conclude that it could have been a worse injury. Dry pants but bladder requires catherization because it held shut for so long and under so much pressure that it cannot open again without mechanical intervention.
Usually books I finish reading are books that I feel offer some clues about just what the hell it means to be alive. This one definitely has that. At least for me. It's like this whole book was written with Legos.
This book is definitely not for everybody. But then again only the ones that say "New York Times Bestseller" are.
People who want a more traditionally human kind of treatment of religion should instead check out "Violence and the Sacred" by Rene Girard. But I still haven't finished that one because it's too human for me. If you feel more like British humor like haha you bloody idiot then Richard Dawkins got you covered. For a more Christian perspective like Dies Irae, la crimosa then check out Pelican books' Introduction to the Bible.
December 31, 2013
"God from the machine" (deus ex machina), has been long waiting by many singularitians where 'artificial intelligence'; has passes human intelligent and leave us behind.. It's not a religion but kind of concept where in the future there will be superintelligent machine could help us from many problems such as climate change, space exploration, etc.
Displaying 1 - 3 of 3 reviews