ISHPSSB 2015 slides "Modeling complex cultural interactions in cooperation: The coevolution of sustainable rice farming and religious practices in Bali"

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Modeling complex culturalinteractions in cooperation:

The coevolution ofsustainable rice farming and

religious practices in Bali

Marshall AbramsDepartment of Philosophy

University of Alabama at BirminghamISHPSSB 2015, July 6, 2015

With additional material added later.http://members.logical.net/~marshall/AbramsISH2015bali.pdf

Marshall Abrams, UAB







2015

-07-

13Coevolution in Bali

Note: Pages with a large gray region at the top contain notes on thepreceding slide (displayed in the upper right corner). These pagespresent ideas that I conveyed verbally during my presentation, alongwith additional material. I’ve also added or rearranged material in someof the slides themselves in order to make this version of thepresentation function better as a standalone document.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Culture 1. Context

Modeling in Gene-Culture Coevolution/Dual-Inheritance Theoryassumes:

“Social learning” (vs. “individual learning”):• There are at least minor differences in beliefs, knowledge,

behavior, attitudes, inclinations, etc. between people within asociety.

• Characteristics in one person sometimes encourage similarones in another.

Like “memes”, but broader.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Example 1. Context

• Why did prehistoric Tasmanians have only very primitivetechnology (tools, clothing, etc.) compared tocontemporaries on the Australian mainland?

• Henrich’s (2004) mathematical models incorporatedabstract representations of learning from others.

• Showed that isolation and small population size could leadto loss of earlier technology.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Complex cultural interactions 1. Context

Gene-culture coevolution models are not necessarily mememodels, but they do usually treat cultural variants as simple, withfew interactions.

It’s implausible that cultural variants don’t influence other variants’adoption.

• Logical inference• Cognitive dissonance• Pedagogical methods• Statistical correlations between attitudes• Priming effects: implicit bias, analogy, etc.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Question 1. Context

• Can we extend modeling methods to investigate patterns ofinfluences of such complex cognitive influences?

• To gain insight about cultural change for specific, real-worldcases?

• To gain insight about general patterns of cultural change?

Answering this methodological question can’t be done from thearmchair.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Question 1. Context

• Can we extend modeling methods to investigate patterns ofinfluences of such complex cognitive influences?

• To gain insight about cultural change for specific, real-worldcases?

• To gain insight about general patterns of cultural change?

Answering this methodological question can’t be done from thearmchair.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Balinese rice farming 2. Lansing-Kremer model

Traditional southern Balinese rice farming requires:• Pest control requires nearby fields to be fallow together.• But there isn’t enough water for everyone to grow rices at

the same time.

Not just cooperation: complex cooperative coordination.



Traditional southern Balinese rice farming requires:• Pest control requires nearby fields to be fallow together.• But there isn’t enough water for everyone to grow rices at

the same time.

Not just cooperation: complex cooperative coordination.

2015

-07-


Lansing-Kremer model

Balinese rice farming

The solution is for each set of farms clustered together betweennatural pest barriers to adopt the same cropping pattern, whiledifferent clusters adopt different, complementary patterns.

Without fallow periods, numbers of rats and insects continue to grow.If neighboring farms don’t share fallow periods, pests just move fromfallow rice paddies to ones in which rice is growing. (In the 1970s, thegreen Revolution imposed continuous growing with pesticides and newrice varieties. This was a disaster–pesticides couldn’t control everypest–and it was eventually realized that it was best to go back totraditional methods.)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Balinese water temple groups 2. Lansing-Kremer model

(Lansing and Kremer, 1993), fig. 10Marshall Abrams, UAB

Balinese water temple groups 2. Lansing-Kremer model

(Lansing and Kremer, 1993), fig. 10

2015

-07-



Balinese water temple groups

This figure shows Lansing’s map from his study the two nearbywatersheds in Bali. Each icon represents a “subak”–a water temple towhich a group of farming villages belong. Icons with the same shapeshow subaks whose farmers used the same crop-scheduling plan. (Ialso drew shapes around groups of subaks that shared a plan.)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


How were water use and crop schedules coordinated in Bali?• Henk Schulte Nordholt: Central planning by representatives

of a king.• Stephen Lansing: Emergent from local decisions by

democratic “water temple” (subak) groups.

(Lansing 2007 [1991]; Lansing and Kremer 1993; Lansing 2006; Lansinget al. 2009; cf. Lansing et al. 1998; Lansing and Fox 2011; Lansing andde Vet 2012; Janssen 2007; Schulte Nordholt 2010, 2011)


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Lansing-Kremer model 2. Lansing-Kremer model

Lansing-Kremer (1993) model:• Crop plans involve different rice varieties, start months.• Tracks water flow, spread of pests.• Allows altering rainfall.• Allows altering pest growth and moveability.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Lansing-Kremer model 2. Lansing-Kremer model

Lansing-Kremer (1993) agent-based model for subaks (watertemple groups):

• Begin with random planting schedules (sequence of ricevarieties, fallow months, starting month).

• If any of your neighboring subaks have a better harvest thanyou do, adopt their planting schedule.

This is a model of cultural transmission model on a network withsuccess bias, with niche construction: Feedback from effects ofhuman behavior on the environment.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Lansing/Kremer/Janssen model 2. Lansing-Kremer model

After 25 years (Model is based on (Jannsen, 2012).)


Lansing/Kremer/Janssen model 2. Lansing-Kremer model

After 25 years (Model is based on (Jannsen, 2012).)

2015

-07-



Lansing/Kremer/Janssen model

This shows graphics from my modified version of Marco Janssen’sreimplementation of Lansing and Kremer’s model. Each multicoloredsquare is a subak, laid out according to the pattern in the real subaksthat Lansing studied. Colors represent crop-scheduling plans. Noticethat the clustering of colors is roughly like the clustering of shapes onLansing’s map, and that the average harvest across all subaks (upperright) goes up and stabilizes near its maximum. (Colored squaresrepresent sequences of rice varieties. Colored circles represent thestarting month of the sequence.)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


Lansing and Kremer’s model• provides a “how-possibly” explanation• Lansing provided further evidence in favor of his hypothesis

using other sources, e.g. historical texts, interviews.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Disruption and religion 3. Extending Lansing’s model

Based on interviews, surveys, interpretations of texts, Lansing’s2006 book Perfect Order suggested that

• Subaks don’t always act as rationally as the model assumes.There are departures due to internal and between-Subakdisputes about power, status, greed, etc.

but• “Religious” cultural patterns among rice farmers tend to

suppress these tendencies.



Based on interviews, surveys, interpretations of texts, Lansing’s2006 book Perfect Order suggested that

• Subaks don’t always act as rationally as the model assumes.There are departures due to internal and between-Subakdisputes about power, status, greed, etc.

but• “Religious” cultural patterns among rice farmers tend to

suppress these tendencies.

2015

-07-


Extending Lansing’s model

Disruption and religion

For convenience, I use “religion” and “religious” to refer to certaincultural patterns discussed in the rest of the presentation, but from the(emic) point of view of the Balinese rice farming culture(s), what we callreligious entities and properties are often just aspects of physical entitiesand properties—people, rats, etc. To call some dimensions of Balineserice farmers’ culture(s) “religion” provides a convenient shorthand andallows comparison with some other work such as Norenzayan’s (seebelow), but it encourages a misleading understanding of that culture,even from an outsider’s (etic) point of view. Nevertheless, investigatingcultural change requires distinguishing aspects of culture that are soconnected as to be unseparable from an emic perspective.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


Religious dimensions of Balinese rice farming:• System of identifications of natural, psychological, spiritual

entities.• Emphasis on order and beauty …• Danger of being corrupted by demons, etc. within and

without people.• Continuous spiritual/practical effort is needed to maintain

and restore order.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Religious variation 3. Extending Lansing’s model

Lansing (2006) argued that region religious patterns amongsubaks differ from:

• Religious patterns near top of mountain, where there is norice farming.

• Earlier royal/Brahmanic religious patterns.• Even earlier Javanese religious patterns.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

What explains cultural coherence? 3. Extending Lansing’s model

Why are religious dimensions of Balinese culturedifferent for rice farmers?

Why is it that aspects of Balinese rice farmers’ culturethat seem distant from practical needs, neverthelessseem well tailored to those needs?


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• Lansing didn’t suggest an answer to why historical culturalsources evolved toward a good fit with rice farming.

• D. S. Wilson Darwin’s Cathedral (2002) seemed to suggestthat religious patterns supporting the subak system evolvedby some kind of selection process, but didn’t suggest whatkind. Group selection seemed to be the only methodWilson entertained.

• It’s unclear how a group selection explanation of Balineserice farmers’ religious patterns would work; what are thegroups?


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• Aya Norenzayan and colleagues (Norenzayan, 2013;Norenzayan et al., in presss) argue that the prevalence of“Big Gods” religions (with all-powerful, omniscient,moralizing, punishing gods) results from group selection andsuccess-biased transmission:

• Groups with Big Gods are more successful, so big godsbelievers proliferate or are are copied because of theirsuccess.

• Balinese gods and spiritual beings partially share this pattern,but the cultural variants that support rice farming go wellbeyond the Big Gods pattern.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• My proposed explanation will use success-biased culturaltransmission without group selection per se, even thoughgroups place a crucial role via their environmental effects.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Extended model 3. Extending Lansing’s model

Starting from Janssen’s (2012) version of Lansing & Kremer’smodel

• Add “capriciousness”, modeled as probability of randomlychoosing a different cropping pattern than your bestneighbor’s.

• Add a second channel of cultural transmission of a religiousvariant, represented by a number between 0 and 1.

• Religious patterns near 1 tend to remove capriciousness.• Subaks copy religious patterns of subaks that have better

harvests.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Hypotheses 3. Extending Lansing’s model

Hypotheses:• When capriciousness is added to Lansing and Kremer’s

model, average harvest is decreased.

When, in addition, religious transmission is added, eventually:• Religious patterns with values near 1 become widespread.• Average harvest isn’t decreased.

Why? Subaks with “appropriate” religious patterns suppresscapriciousness, have greater harvests.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• Begin with random planting schedules, religions “patterns”between 0 and 1.

• After a 6000-month = 500-year “burn-in” period,• Run for 60,000 months = 5000 years.

100 simulation runs for each condition.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


Variations:

1. No capriciousness, no religious effects.

2. Capriciousness, no religious effects.

3. Capriciousness plus religious effects.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Harvest w/out relig. tran. 3. Extending Lansing’s model

Comparison of harvest with no religious effects,with/without noise in crop pattern transmission

harvest

0.0

0.2

0.4

0.6

1 2 3 4 5 6 7 8 9 100.0

0.2

0.4

0.6

nonoisenorelig

Avg over 5000 yrs, 100 runs. Blue: Lansing-Kremer; Pink: capriciousness.


Harvest w/out relig. tran. 3. Extending Lansing’s model

Comparison of harvest with no religious effects,with/without noise in crop pattern transmission

harvest

0.0

0.2

0.4

0.6

1 2 3 4 5 6 7 8 9 10

0.0

0.2

0.4

0.6

nonoisenorelig

Avg over 5000 yrs, 100 runs. Blue: Lansing-Kremer; Pink: capriciousness.

2015

-07-



Harvest w/out relig. tran.

This figure shows the 100-run distribution of averages over 5000 yearsand all subaks, of harvest values for two conditions. Blue: The pureLansing-Kremer model without added capriciousness; pink: With addedcapriciousness. As expected, harvests are better without capriciousness.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

What didn’t work 3. Extending Lansing’s model

• My original hypothesis was that allowing subaks to examineharvests of other subaks across the entire population wouldlead to the spread of “high” religious values.

• Informal investigation suggested that wouldn’t work.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

What didn’t work 3. Extending Lansing’s model

• My original hypothesis was that allowing subaks to examineharvests of other subaks across the entire population wouldlead to the spread of “high” religious values.

• Informal investigation suggested that wouldn’t work.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Why global transmission fails 3. Extending Lansing’s model

• When subaks compare harvests across the entirepopulation early in the simulation, there’s a random religiousvalues/harvest association.

• Since everyone copies the best harvests, variation inreligious value disappears.

• After that, narrowly-clustered religious values random-walk(roughly speaking) due to transmission noise.

• It turned out that the solution is to allow only intermittentcommunication between non-neighboring subaks.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


Capriciousness plus religious transmission condition models:

1. Always consider imitating religious values of yourneighboring subaks.

2. Also consider imitating religious values of random subakswhere mean number of subaks considered = .025, 1, 50,

using a Poisson distribution (truncated at 172, the numberof subaks).

3. Religious effect curves:

step at 0.5 step at 0.8 linear “sigmoidey”

100 runs for each of 12 models. 5000 years after burn-in.Marshall Abrams, UAB


Capriciousness plus religious transmission condition models:

1. Always consider imitating religious values of yourneighboring subaks.

2. Also consider imitating religious values of random subakswhere mean number of subaks considered = .025, 1, 50,

using a Poisson distribution (truncated at 172, the numberof subaks).

3. Religious effect curves:

step at 0.5 step at 0.8 linear “sigmoidey”

100 runs for each of 12 models. 5000 years after burn-in.

2015

-07-



Extended model

This slide describes the 12 different religious-variant transmissionmodels I ran, with 100 runs for each set of parameters. The religiouseffect functions plotted at the bottom of the page determine thedegree to which different religious values between 0.0 and 1.0 have theeffect of suppressing capriciousness. For example, in models using theleftmost function, religious values below 0.5 produce no suppression ofcapriciousness, while values above 0.5 fully suppress capriciousness.The second function is similar, but the cutoff is at 0.8. The third functionmakes the degree of suppression equal to the subak’s religious value,while the fourth function produces a gradual but increasing suppressiveeffect with rising religious value, plateauing near 0.7. (Each subak has0–4 neighbor subaks. “Neighbor subak”: i.e. near enough that pests cantravel between it and your subak.)

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

100−run means of mean relig−types over 5000 years

relig−type

0.00

0.05

0.10

0.15

0.02

0.04

0.06

0.08

05step0025global

0.02

0.04

0.06

0.08

08step0025global

0.02

0.04

0.06

0.08

linear0025global

0.02

0.04

0.06

0.08

0.00

0.05

0.10

0.15

sigmoidey0025global

0.00

0.05

0.10

0.15

05step1global

08step1global

linear1global

0.00

0.05

0.10

0.15

sigmoidey1global

0.00

0.05

0.10

0.15

05step50global

08step50global

linear50global

0.00

0.05

0.10

0.15

sigmoidey50global


100−run means of mean relig−types over 5000 years

relig−type

0.00

0.05

0.10

0.15

0.02

0.04

0.06

0.08

05step0025global

0.02

0.04

0.06

0.08

08step0025global

0.02

0.04

0.06

0.08

linear0025global

0.02

0.04

0.06

0.08

0.00

0.05

0.10

0.15

sigmoidey0025global

0.00

0.05

0.10

0.15

05step1global

08step1global

linear1global

0.00

0.05

0.10

0.15

sigmoidey1global

0.00

0.05

0.10

0.15

05step50global

08step50global

linear50global

0.00

0.05

0.10

0.15

sigmoidey50global

2015

-07-



These 12 plots show the 100-run distributions of the average, over 5000 years and all subaks, ofreligious values under each of 12 parameter combinations.

• 50global : Once a year, each subak adds to its neighbors 50 other subaks on average fromthe global population, and compares copies the religious value of the subak with the bestharvest.

• 1global : Same pattern, but selecting one subak, on average, from the global population.

• 0025global : Same pattern with 0.025 subaks from the global population, on average.

• 05step, 08step, linear , sigmoidey : The four religious effect curves shown in the previousslide. See notes on that slide.

Capricious-suppressing religious values tend to spread when communication betweennon-neighbors exists but is relatively rare. The effect is more pronounced with some religiouseffect curves, but more or less goes away when the number of non-neighbors that are consideredfor comparison is large (i.e. 50). Informal experiments suggest that using higher numbers of subaksfor comparison would not improve spread of high religious values.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

100−run avg harvest religious effects vs. no religious effects condition

harvest (avgharvestha)

0.0

0.2

0.4

0.6

1 2 3 4

05step0025global

1 2 3 4

08step0025global

1 2 3 4

linear0025global

1 2 3 4

0.0

0.2

0.4

0.6

sigmoidey0025global

0.0

0.2

0.4

0.6

05step1global

08step1global

linear1global

0.0

0.2

0.4

0.6

sigmoidey1global

0.0

0.2

0.4

0.6

05step50global

08step50global

linear50global

0.0

0.2

0.4

0.6

sigmoidey50global

no relig effectwith relig effect


100−run avg harvest religious effects vs. no religious effects condition


0.0

0.2

0.4

0.6

1 2 3 4

05step0025global

1 2 3 4

08step0025global

1 2 3 4

linear0025global

1 2 3 4

0.0

0.2

0.4

0.6

sigmoidey0025global

0.0

0.2

0.4

0.6

05step1global

08step1global

linear1global

0.0

0.2

0.4

0.6

sigmoidey1global

0.0

0.2

0.4

0.6

05step50global

08step50global

linear50global

0.0

0.2

0.4

0.6

sigmoidey50global

no relig effectwith relig effect

2015

-07-



These 12 plots allow comparison of the effect of the spread of highreligious values on harvest values, with the absence of this effect. The12 conditions are the same as in the previous slide. Pink bars show the100-run distribution of harvest values averaged over 5000 years and allsubaks; blue bars show the same distribution without any effect ofreligion on capriciousness. Notice that the presence ofcapriciousness-suppressing religion in the population always improvesaverage harvest, but the effect is more pronounced under theconditions in which high religious values tend to be more widespread.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Discussion 3. Extending Lansing’s model

• When harvest-success-biased religious transmission between subakgroups is occasional (mean per subak per year .025 or 1), religious valuesthat lead to local coordination for pest avoidance and globalcoordination for water use have a chance to become clustered in a fewsubak groups.

• These groups have more successful harvests, and so occasional copyingleads to the spread of their religious values.

• When global success-biased copying is more common (mean per subakper timestep = 50), the relationship between harvest success andreligious values is too noisy, and harvest-success-conducive religiousvalues are unlikely to spread.

• Success-conducive religious values improve harvest yields on averagerelative to no religious transmission.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Conclusion 4. Conclusion

• My extended Lansing-Kremer simulation shows that it canbe fruitful to model causal interactions between culturalpatterns in different domains.

• And that such effects can depend on the social structure ofcommunication

• and on the way in which cultural variants bias the effects ofother variants.

• This model illustrates these points both for a specificcultural context, and more generally.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


However:

• I ran the experiments only for high values for rainfall andpest growth and dispersal parameters that were available inLansing & Kremer’s model. In future experiments I may useother values, but I expect broadly similar patterns.

• The model treats cultural transmission as a phenomenon atthe level of subaks; a more realistic model would modelvillages and individual people as well.

• Lansing et al. (2014) showed statistically significant culturaldifferences between upstream and downstream subaks inpatterns relevant to water management and sharing. Myresults don’t address this variation.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• Boyer (2001) and others may be right that minimallycounterintuitive religious beings are more likely toremembered and transmitted, but that doesn’t explainculture-specific variations.

• Gervais et al. (2011) argue that general biases on culturaltransmission are needed to explain these variations, butdon’t focus on explaining interactions and apparentcoherence between cultural variants.

• Models of interactions between cultural variants have rarelybeen applied to details of real cultures.(Cavalli-Sforza and Feldman, 1981; Boyd and Richerson, 1985, 1987; McElreath et al., 2003;Mesoudi and O’Brien, 2008; Fogarty et al., 2011; Castro and Toro, 2014; Claidière et al.,2014)



• Boyer (2001) and others may be right that minimallycounterintuitive religious beings are more likely toremembered and transmitted, but that doesn’t explainculture-specific variations.

• Gervais et al. (2011) argue that general biases on culturaltransmission are needed to explain these variations, butdon’t focus on explaining interactions and apparentcoherence between cultural variants.

• Models of interactions between cultural variants have rarelybeen applied to details of real cultures.(Cavalli-Sforza and Feldman, 1981; Boyd and Richerson, 1985, 1987; McElreath et al., 2003;Mesoudi and O’Brien, 2008; Fogarty et al., 2011; Castro and Toro, 2014; Claidière et al.,2014)

2015

-07-


Conclusion

Conclusion

It’s valuable go beyond the abstract models that have been explored previously.

Even if some assumptions in a model such as mine are unrealistic, basing amodel on real-world details can expand the space of possibilities by leading usbeyond simple assumptions in abstract models. Thus in addition to the insightsabout particular cultures that a more applied model can provide, it can alsocontribute more general insights such as those provided by abstract models.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• Norenzayan and colleagues (Norenzayan, 2013;Norenzayan et al., in presss) argue that moralizing “BigGods” religions spread by success-biased transmission orgroup selection because they enhance cooperation.

• Norenzayan’s hypothesis may be broad enough to includesome features of Balinese rice-farming culture, but thecultural patterns that seem to support coordination amongBalinese rice farmers go well beyond Big Gods factors.

• The religion variables in my model can be viewed asrepresenting both Big Gods and other influences that leadto reduction of capriciousness in Balinese rice farmingcommunities.



• Norenzayan and colleagues (Norenzayan, 2013;Norenzayan et al., in presss) argue that moralizing “BigGods” religions spread by success-biased transmission orgroup selection because they enhance cooperation.

• Norenzayan’s hypothesis may be broad enough to includesome features of Balinese rice-farming culture, but thecultural patterns that seem to support coordination amongBalinese rice farmers go well beyond Big Gods factors.

• The religion variables in my model can be viewed asrepresenting both Big Gods and other influences that leadto reduction of capriciousness in Balinese rice farmingcommunities.20

15-0

7-13

Coevolution in BaliConclusion

Conclusion

It’s important to go beyond broad generalizations like Norenzayan’s to look athow detailed interactions between local cultural variants affect cultural change.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.


• This is part of a larger project to explore ways that modelscan incorporate insights from qualitative, humanisticresearch, in useful and tractable ways.

• Ongoing, related work:• Exploring a broadened parameter space using simpler, more

abstract models of the spread of religious values amongBalinese rice farmers.

• Modeling the role of analogies in the Balinese (cf. paper citedat top of on next slide).


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Previous work 4. Conclusion

• Qualitative work in the tradition of symbolic anthropology ascapturing effects of cognitive processing of analogies on culturaltransmission (Abrams, 2013). Incorporating a cognitive sciencemodel of analogy processing can allow us to model the effects ofanalogies (and some “symbolic” relations”) on cultural evolution.

• “A Moderate Role for Cognitive Models in Agent-BasedModeling of Cultural Change”, Complex Adaptive SystemsModeling 2013.

• Abstract mathematical models of psychological interactionsbetween cultural variants:

• “Coherence, Muller’s Ratchet, and the Maintenance ofCulture”, Philosophy of Science, in press, 2015.

• “Cultural variant interaction in teaching and transmission”(commentary), Behavioral and Brain Sciences, in press.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Thanks 4. Conclusion

Stephen Lansing, Marco Janssen, Emily Schultz, many others, includingaudience members at other conferences. Last but not least:

This work was supported in part by the research computing resourcesacquired and managed by UAB IT Research Computing. Any opinions,findings and conclusions or recommendations expressed in thismaterial are those of the author and do not necessarily reflect theviews of the University of Alabama at Birmingham.


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.








.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Abrams, Marshall (2013). A moderate role for cognitive models in agent-based modeling ofcultural change. Complex Adaptive Systems Modeling 1(16):1–33.

Abrams, Marshall (2015). Coherence, Muller’s ratchet, and the maintenance of culture. Philosophyof Science 82(5). In press.

Abrams, Marshall (in press). Cultural variant interaction in teaching and transmission. Behavioraland Brain Sciences .

Boyd, Robert and Richerson, Peter J. (1985). Culture and the Evolutionary Process. University ofChicago, Chicago, IL.

Boyd, Robert and Richerson, Peter J. (1987). The evolution of ethnic markers. CulturalAnthropology 2(1):65–79. Reprinted in (Boyd and Richerson, 2005).

Boyd, Robert and Richerson, Peter J. (2005). The Origin and Evolution of Cultures. OxfordUniversity Press, Oxford, UK.

Boyer, Pascal (2001). Religion Explained. Basic Books.

Castro, Laureano and Toro, Miguel A. (2014). Cumulative cultural evolution: The role of teaching.Journal of Theoretical Biology 347:74–83.

Cavalli-Sforza, Luca L. and Feldman, Marcus W. (1981). Cultural Transmission and Evolution: AQuantitative Approach. Princeton University Press, Princeton, NJ.

Claidière, Nicolas; Scott-Phillips, Thomas C.; and Sperber, Dan (2014). How Darwinian is culturalevolution? Philosophical Transactions of the Royal Society of London B: Biological Sciences369(1642):1–8.

Fogarty, L.; Strimling, P.; and Laland, K. N. (2011). The evolution of teaching. Evolution65(10):2760–2770.

Gervais, Will M.; Willard, Aiyana K.; Norenzayan, Ara; and Henrich, Joseph (2011). The culturaltransmission of faith: Why innate intuitions are necessary, but insufficient, to explain religiousbelief. Religion 41(3):389–410.Marshall Abrams, UAB

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Henrich, Joseph (2004). Demography and cultural evolution: How adaptive cultural processes canproduce maladaptive losses: The Tasmanian case. American Antiquity 69(2):197–214.

Jannsen, Marco A. (2012). Lansing-Kremer model of the balinese irrigation system (version 2).CoMSES Computational Model Library.URL http://www.openabm.org/model/2221/version/2

Janssen, Marco A. (2007). Coordination in irrigation systems: An analysis of the Lansing–Kremermodel of bali. Agricultural Systems 93:170–190.

Lansing, J. Stephen (2006). Perfect Order: Recognizing Complexity in Bali. Princeton University,Princeton, New Jersey.

Lansing, J. Stephen (2007 [1991]). Priests and Programmers: Technologies of Power in the EngineeredLandscape of Bali. Princeton University, Princeton, New Jersey. First edition 1991. Reprinted,with a new forward and preface, 2007.

Lansing, J. Stephen; Cheong, Siew Ann; Chew, Lock Yue; Cox, Murray P.; Ho, Moon-Ho Ringo; andArthawiguna, Wayan Alit (2014). Regime shifts in Balinese subaks. Current Anthropology55(2):232–239. NEED URL.

Lansing, J. Stephen; Cox, Murray P.; Downey, Sean S.; Jannsen, Marco A.; and Schoenfelder, John W.(2009). A robust budding model of balinese water temple networks. World Archaeology41(1):110–131.

Lansing, J. Stephen and de Vet, Thérèse A. (2012). The functional role of balinese water temples:A response to critics. Human Ecology 40(3):453–478.

Lansing, J. Stephen and Fox, Karyn M. (2011). Niche construction on Bali: The gods of thecountryside. Philosophical Transactions of the Royal Society B: Biological Sciences366(1566):927–934.URLhttp://rstb.royalsocietypublishing.org/content/366/1566/927.abstractMarshall Abrams, UAB

http://www.openabm.org/model/2221/version/2

http://rstb.royalsocietypublishing.org/content/366/1566/927.abstract

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Lansing, J. Stephen and Kremer, James N. (1993). Emergent properties of balinese water templenetworks: Coadaptation on a rugged fitness landscape. American Anthropologist 95(1):97–114.URL http://dx.doi.org/10.1525/aa.1993.95.1.02a00050

Lansing, J.Stephen; Kremer, James N.; and Smuts, Barbara B. (1998). System-dependent selection,ecological feedback and the emergence of functional structure in ecosystems. Journal ofTheoretical Biology 192(3):377 – 391.URLhttp://www.sciencedirect.com/science/article/pii/S0022519398906646

McElreath, Richard; Boyd, Robert; and Richerson, Peter J. (2003). Shared norms and the evolutionof ethnic markers. Current Anthropology 44(1):122–129. Reprinted in (Boyd and Richerson,2005).

Mesoudi, Alex and O’Brien, Michael J. (2008). The learning and transmission of hierarchicalcultural recipes. Biological Theory 3(1):63–72.

Norenzayan, Ara (2013). Big Gods: How Religion Transformed Cooperation and Conflict. PrincetonUniversity, Princeton, New Jersey.

Norenzayan, Ara; Shariff, Azim F.; Willard, Aiyana K.; Slingerland, Edward; Gervais, Will M.;McNamara, Rita A.; and Henrich, Joseph (in presss). The cultural evolution of prosocialreligions. The Behavioral and Brain Sciences .

Schulte Nordholt, Henk (2010). The Spell of Power: A History of Balinese Politics, 1650-1940. BrillAcademic Publishers.

Schulte Nordholt, Henk (2011). Dams and dynasty, and the colonial transformation of balineseirrigation management. Human Ecology 39(1):21–27.

Wilson, David Sloan (2002). Darwin’s Cathedral: Evolution, Religion and the Nature of Society.University of Chicago Press.Marshall Abrams, UAB

http://dx.doi.org/10.1525/aa.1993.95.1.02a00050

http://www.sciencedirect.com/science/article/pii/S0022519398906646

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Supplemental Slides


.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

100−run avg harvest religious effects vs. no noise condition


0.0

0.2

0.4

0.6

1 2 3 4

05step0025global

1 2 3 4

08step0025global

1 2 3 4

linear0025global

1 2 3 4

0.0

0.2

0.4

0.6

sigmoidey0025global

0.0

0.2

0.4

0.6

05step1global

08step1global

linear1global

0.0

0.2

0.4

0.6

sigmoidey1global

0.0

0.2

0.4

0.6

05step50global

08step50global

linear50global

0.0

0.2

0.4

0.6

sigmoidey50global

no noisenoise + relig effect


100−run avg harvest religious effects vs. no noise condition


0.0

0.2

0.4

0.6

1 2 3 4

05step0025global

1 2 3 4

08step0025global

1 2 3 4

linear0025global

1 2 3 4

0.0

0.2

0.4

0.6

sigmoidey0025global

0.0

0.2

0.4

0.6

05step1global

08step1global

linear1global

0.0

0.2

0.4

0.6

sigmoidey1global

0.0

0.2

0.4

0.6

05step50global

08step50global

linear50global

0.0

0.2

0.4

0.6

sigmoidey50global

no noisenoise + relig effect

2015

-07-


This supplemental slide superimposes the pure Lansing-Kremerno-capriciousness result (blue) on slides showing thecapriciousness-plus-religious transmission results from the 12-panelbarcharts shown earlier (pink). This slide shows that capriciousness plusreligious suppression of it produces harvests that are a bit worse, onaverage, than in Lansing and Kremer’s idealized system withoutcapriciousness.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Simulation parameters

NetLogo 5.1.0Source file: src/LKJplus/BaliPlus.nlogo (versions of 4/4–4/5/2015)burn-in-months = 6000months per run: 66000rainfall-scenario = highpestdispersal-rate = 1.5pestgrowth-rate = 2.4relig-tran-stddev = 0.02relig-influence = 1.5Used only the five crop plans that include only traditional rice varieties (1 and 2)(i.e. rice variety 3 wasn’t used in any model).For runs with capriciousness, ignore-neighbors-prob = 0.3Religious effect curves:

- step at 0.5- step at 0.8- linear (i.e. suppression effect = value of relig cultural variant)- “sigmoidey” with relig-effect-center = 2.25, relig-effect-endpt = 1.7 (see sourcecode for function definition)

Poisson means for addition of subaks from the global population to those neighboring subaks whoare candidates for transmission: subaks-mean-global = 0.025, 1, 50


ISHPSSB 2015 slides "Modeling complex cultural interactions in cooperation: The coevolution of sustainable rice farming and religious practices in Bali"

Documents