• TABLE OF CONTENTS
HIDE
 Introduction
 The cognitive revolution
 Agricultural decision making before...
 Agricultural decision making after...
 Decision tree modeling
 A real-life example: The Malawi...
 Elimination criteria
 Stage-two criteria
 The risk subroutine
 Results
 Conclusion
 Reference
 Figure 1a: The decision between...
 Figure 1b: Motivations to use chemical...
 Figure 1c: Resource constraints...
 Figure 1d: Resource constraints...






Title: Indigenous knowledge systems, the cognitive revolution, and agricultural decision making
CITATION PAGE IMAGE ZOOMABLE PAGE TEXT
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00071957/00001
 Material Information
Title: Indigenous knowledge systems, the cognitive revolution, and agricultural decision making
Physical Description: 23 leaves : ; 28 cm.
Language: English
Creator: Gladwin, Christina H
Publication Date: 2002
 Subjects
Subject: Traditional farming -- Decision making -- Data processing   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Spatial Coverage: Mexico
Mozambique
Malawi
 Notes
Bibliography: Includes bibliographical references (p. 15-19).
Statement of Responsibility: Christina H. Gladwin.
General Note: Typescript.
General Note: Per author's email, work presented at ICRAF seminar in 2002(?).
Funding: Electronic resources created as part of a prototype UF Institutional Repository and Faculty Papers project by the University of Florida.
 Record Information
Bibliographic ID: UF00071957
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 74668832

Table of Contents
    Introduction
        Page 1
    The cognitive revolution
        Page 2
        Page 3
    Agricultural decision making before the cognitive revolution
        Page 4
    Agricultural decision making after the cognitive revolution
        Page 4
        Page 5
    Decision tree modeling
        Page 6
    A real-life example: The Malawi smallholder's decision between chemical and organic fertilizers
        Page 7
        Page 8
    Elimination criteria
        Page 9
    Stage-two criteria
        Page 9
    The risk subroutine
        Page 10
    Results
        Page 11
        Page 12
    Conclusion
        Page 13
        Page 14
    Reference
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
    Figure 1a: The decision between organic and chemical fertilizers on local maize - stage 1 constraints
        Page 20
    Figure 1b: Motivations to use chemical or organic or both
        Page 21
    Figure 1c: Resource constraints to use of organic fertilizer
        Page 22
    Figure 1d: Resource constraints to use of chemical fertilizer
        Page 23
Full Text

/6, (ao
c-, OVA,'

INDIGENOUS KNOWLEDGE SYSTEMS, THE COGNITIVE REVOLUTION,
AND AGRICULTURAL DECISION MAKING

by

Christina H. Gladwin,
University of Florida

The school of thought now dubbed IKS, indigenous knowledge systems, aims to elicit the

expert systems of indigenous peoples -- peasants -- who are sometimes not thought of as experts.

These knowledge systems are brought back to agricultural research centers and ministries of

agriculture and used to educate agricultural scientists and policy makers so that they can design

better technologies and policies to improve peasants' standards of living. The idea that local

knowledge is a valuable resource to be understood and used to alleviate poverty is beginning to

diffuse throughout the Third World in practical applications to development problems, and in the

academic literature, as is evidenced by these papers and others (Brokensha 1989). Starting with

the collection of the same name, IKS scholars have described the expert systems of peasants to do

any number of things: to farm, to fertilize, to manage their soils and natural resources in a

sustainable way, to adopt or not adopt new "improved" technologies, to take risks, to make a

living (Brokensha, Warren, and Werner 1980). Farming systems programs in international

agricultural research centers have seen the value of knowing "how the farmers think" about their

crops, pests, forests, and water resources (CIMMYT 1984, Matlon, Cantrell, King, and Benoit-

Cattin 1984), before agricultural scientists design improved technologies for them. Increasingly,

it is accepted wisdom for an agronomist to get feedback from peasants about new improved seeds

and biotechnologies before their official release from the experiment station.

It is also generally recognized that the goals of the IKS school go way back to the

anthropologist Malinowski (1922) who aimed to "grasp the native's point of view, his relation to

life, to realize his vision of his world." To see the insiders' world through the insiders' eyes has

long been the aim of ethnographers whose work describes a culture from the "native's" or

insider's point of view and not from the researcher's or outsider's point of view (Spradley 1979).

To minimize the researcher's own ethnocentricity, i.e., the viewing of another culture through the









2

lens of one's own cultural values and assumptions, the ethnographer seeks to learn from people,

to be taught by them like a child is taught. The aim is to discover the cultural meaning of the

insiders' relationships, native terms, rules, and way of life (Spradley 1979: 3). This is a far

different goal from that of collecting data about people and testing a model based on the

outsider's view. Models of indigenous knowledge systems should thus contain "emic" categories,

i.e., units of meaning drawn from the culture bearers themselves, which can be contrasted with

"etic" categories which may have meaning for researchers but need not have meaning for the

people of the specific culture under study (Pike 1954; Harris 1979: 32-45).

What is not yet accepted wisdom is the importance of cognitive science to this school of

thought and agricultural development policies, and the influence of "the cognitive revolution"

which since the 1950s and 1960s has shaken the foundations of six social sciences: artificial

intelligence, anthropology, linguistics, neuroscience, psychology, and philosophy (Gardner 1985).

Perhaps agricultural scientists who are used to empirical laboratory experiments are

uncomfortable in realizing that farmers' acceptance of their new biotechnologies partly depends

on the mystery of how the human mind -- the ultimate black box -- works. Yet as more and

more agricultural scientists give up the naive notion that their technologies will automatically

diffuse, more and more research on farmer decision making, especially of the type "Why don't

they adopt," is done. One lesson gleaned from the adoption literature, by now too vast to

adequately cite, is that the success of efforts to predict farmers' decision making depends on the

underlying assumptions that are made about cognition, i.e., the thought process itself. These

assumptions changed drastically with the cognitive revolution. Before the cognitive revolution,

agricultural decision models were mathematically sophisticated, with risk-aversion often being

measured by the sign of the second differential; after the cognitive revolution, simpler decision

rules and trees replaced optimization of a continuous, twice-differentiable function.


The Cognitive Revolution

Scientific assumptions about cognition were radically changed in 1956 with the publication

of a seemingly-innocuous article entitled, "The Magic Number Seven, Plus or Minus Two" by the









3

psychologist George Miller. Miller's new idea was that the human computer is of limited

capacity, unlike a real computer; its short-term memory is limited to roughly 5 to 7 items at a

time. Indeed, people seem to categorize or discretize variables, rather than deal with continuous

quantitative variables as a real computer does. People use logic rather than perform complicated

mathematical operations as a real computer does. The limited "rehearsal buffer" of humans

affects the way people organize their language and thought processes, such that they are

hierarchicallyy" organized (Miller, Galanter, and Pribram 1960).

This revolutionary idea was taken up by the linguist Noam Chomsky, whose "trees" or

transformational grammars spread to most known languages around the world. In the field of

artificial intelligence, it stimulated scientists to invent computer programs called "expert systems"

which mirrored the way humans think, rather than expect humans to think like sophisticated

computers. In psychology, the revolution stimulated new theories of problem solving (Newell and

Simon 1972) and plans or scripts (Schank and Abelson 1977), and caused the abandonment of

decision making models of "expected utility" in favor of theories of "elimination-by-aspects" and

"preference trees" (Tversky 1972; Kahneman and Tversky 1972, 1982; Tversky and Kahneman

1981).

In anthropology, it resulted in the spread of cognitive models of taxonomies, schematas,

and decision processes which replaced the use of psychological traits and raw-shock tests to

explain cultural differences. Given the aims of ethnographers to model the insiders' point of

view and knowledge, it was not surprising that the cognitive revolution found the field of

anthropology a fertile ground to grow in. The products of cognitive anthropology can now be

seen in frame analysis (Frake 1964), taxonomic analysis (Berlin and Kay 1969), componential

analysis (Romney and D'Andrade (1964), schematas or folk models (D'Andrade 1981, 1984, H.

Gladwin 1974, Quinn 1989), plans or scripts (Werner and Schoepfle 1987), and decision tree

models or tables. Before these models are described, a look at what decision models were like

before the cognitive revolution is in order.









4

Agricultural Decision Making Before the Cognitive Revolution


In the field of agricultural decision making, long dominated by (agricultural) economists,

decision models were quantitative, linear-additive, and often normative (e.g., linear programming

models, expected value and expected utility models, stochastic dominance) but typically not

empirically grounded. They were not usually tested against a set of choice data to see how well

they predicted the choices of individuals in a group (see Anderson 1979, Anderson, Dillon, and

Hardaker 1977). Instead, they were either used as behavioral assumptions in a model of aggregate

supply or demand, or as normative models to tell people how they should make decisions (Raiffa

1968), or tested to see if they "fit" the observed behavior of one "representative" individual

(Benito 1976).

They were not empirically tested against choice data because the test was usually so

complicated that it was not worth the effort. For example, in a test of the expected utility

model, the researcher had to measure each individual's "utility function," which could vary in

shape across individuals depending on how risk-averse they were; and then independently

measure each individual's "subjective" probability distribution which differed from the objective

probability distribution. This proved to be such a job that it was rarely done; but textbooks

described how it could be done (Anderson, Dillon, and Hardaker 1977). When it was done, errors

arose due to inconsistencies (Officer and Halter 1968), or when midway through the experiment,

riskless gambles were replaced by risky gambles, suggesting that even the same individual might

have more than one utility function (Tversky 1967).


Agricultural Decision Making After the Cognitive Revolution


Cognitive psychologists and anthropologists in the 1970s rejected the expected utility

theory of choice, and searched for more cognitively-realistic models of the choice process. They

claimed that people in real-life choice contexts don't make holistic assignments of utility or

satisfaction to each alternative in the choice set, and separately formulate subjective probabilities

(Quinn 1978), and then pick the alternative with the most "expected utility" (Kahneman and









4

Agricultural Decision Making Before the Cognitive Revolution


In the field of agricultural decision making, long dominated by (agricultural) economists,

decision models were quantitative, linear-additive, and often normative (e.g., linear programming

models, expected value and expected utility models, stochastic dominance) but typically not

empirically grounded. They were not usually tested against a set of choice data to see how well

they predicted the choices of individuals in a group (see Anderson 1979, Anderson, Dillon, and

Hardaker 1977). Instead, they were either used as behavioral assumptions in a model of aggregate

supply or demand, or as normative models to tell people how they should make decisions (Raiffa

1968), or tested to see if they "fit" the observed behavior of one "representative" individual

(Benito 1976).

They were not empirically tested against choice data because the test was usually so

complicated that it was not worth the effort. For example, in a test of the expected utility

model, the researcher had to measure each individual's "utility function," which could vary in

shape across individuals depending on how risk-averse they were; and then independently

measure each individual's "subjective" probability distribution which differed from the objective

probability distribution. This proved to be such a job that it was rarely done; but textbooks

described how it could be done (Anderson, Dillon, and Hardaker 1977). When it was done, errors

arose due to inconsistencies (Officer and Halter 1968), or when midway through the experiment,

riskless gambles were replaced by risky gambles, suggesting that even the same individual might

have more than one utility function (Tversky 1967).


Agricultural Decision Making After the Cognitive Revolution


Cognitive psychologists and anthropologists in the 1970s rejected the expected utility

theory of choice, and searched for more cognitively-realistic models of the choice process. They

claimed that people in real-life choice contexts don't make holistic assignments of utility or

satisfaction to each alternative in the choice set, and separately formulate subjective probabilities

(Quinn 1978), and then pick the alternative with the most "expected utility" (Kahneman and









5

Tversky 1972, 1982). In line with Miller's results on the limitations of human computational

capacities, they felt that decisions are made in a decomposed fashion using relative comparisons,

because it is cognitively easier to compare alternatives on a piece-meal basis, i.e., one dimension

at a time (Schoemaker 1982). Indeed, people do not rank order alternatives holistically when they

make a decision. They just choose one out of several alternatives without ranking them (Quinn

1971), in which case the decision model is what Arrow (1951) calls a "choice function not built up

from orderings," i.e., simply a set of rules. In some choice contexts, these rules may result in an

incomplete order (Gladwin 1975) and intransitive preference structure (Tversky 1969).

Cognitivists even objected to the linear-additive decision models called probit analysis and

logit analysis, which have the advantage over expected utility and linear programming models of

being testable with data on choices made by many rather than one individual. Unfortunately,

they also are not cognitively-realistic models of the choice process. No-one assigns weights to

several variables and then adds them up to determine which of several outcomes is better; people

compare alternatives one dimension at a time. But probit and logit analyses do have the

advantage of providing a statistical test; they thus can be used alongside rule-based decision

models to show whether there is a significant correlation between a particular independent

variable (or decision criterion in the rule) and the decision outcome chosen. In this way, they can

provide an indirect test of a rule-based model (Mukhopadhyay 1984). Unfortunately, they can

provide this test for only a few of the variables or criteria in an individual's decision process.

Cognitivists also rejected the quantitative nature of decision variables in a probit or logit

analysis or linear programming model. Following Miller (1956), they claimed that decision

makers use discrete decision criteria in real-life choices, even when faced with a variable

amenable to quantification such as cost. The decision criteria used in decision tree models are

thus not continuous quantitative variables; they are discrete constraints that must be passed or

satisfied (e.g., Is cost of car < $4000?) or orders and semi-orders of alternatives on aspects (e.g.,

Is cost of car1 < cost of car2?). An alternative is assumed to be a set of aspects or constraints

(Lancaster 1966, Tversky 1972, Gladwin 1980), but criteria are discrete. An algebraic form of

choice model results. The decision process is also assumed to be deterministic rather than









6

probabilistic: an alternative either passes the criteria or constraints with a probability of one or it

does not. There are thus no probabilities -- other than 1 or 0 -- facing the individual on each

branch, as in Raiffa (1968). A decision tree is thus a sequence of discrete decision criteria, all of

which have to be passed along a path to a particular outcome or choice.


Decision Tree Modeling


Ethnographic decision tree modeling starts from the assumption that the decision makers

themselves are the experts on how they make the decisions they make. It uses ethnographic

fieldwork techniques to elicit from the decision makers themselves their decision criteria, which

are then combined in the form of a decision tree, table, flowchart, or set of if-then rules or

"expert systems" which can be programmed on the computer (Gladwin 1989). There are thus two

distinctive features about the method: its reliance on ethnographic fieldwork techniques to elicit

the decision criteria, and its insistence on a formal, testable, computer-based model of the

decision process which is hierarchical or treelike in nature.

Ethnographic decision tree modeling is not a black box technique to test the researcher's

interpretation of the insiders' culture, like some quantitative methods (e.g., factor analysis,

multidimensional scaling, cluster analysis). Because of its dependence on eliciting procedures, the

model is culturally tuned by some specific group of individuals, and then tested against choice

data from other individuals in the group. The form of a decision tree model is amazingly simple,

with the choice alternatives in a set at the top of the tree, denoted by { } and the decision

criteria at the nodes or diamonds of the tree denoted by < >, and the decision outcomes

denoted by [ ] at the ends of the paths of the tree. The decision maker starts at the top of the

tree and, independently of other decision makers, is asked the set of questions in the criteria at

the nodes of the tree, and based on his or her responses is "sent down" the tree on a path to a

particular outcome.

Cognitive anthropologists in the 1970s and 1980s applied these ideas to real-life choice

contexts in a number of different cultures. These included economic decisions made by

Ghanaian fish sellers in deciding between markets (H. Gladwin 1971, C. Gladwin 1975, Quinn









7

1978), farmers' adoption decisions in Puebla, Mexico (C. Gladwin 1976, 1977, 1979a, 1979b),

Californian families' decisions regarding the sexual division of labor within the family for daily

routine tasks (Mukhopadhyay 1984), farmers' cropping decisions (Barlett 1977, C. Gladwin 1983),

peasants' choice of treatment for illness in Pichatero, Mexico (Young 1980, 1981), U.S. car

buyers' choice of cars (H. Gladwin and Murtaugh 1984, Murtaugh and H. Gladwin 1980),

economic development decisions of the Navajo tribe (Schoepfle, Burton, and Morgan 1984), and

U.S. farmers' decisions to cut back production and sell land during a farm crisis (Zabawa 1984,

C. Gladwin and Zabawa 1984, 1986, 1987). In each case where the methodology of decision trees

has been used, the predictability has been as high as 85 to 95 percent of the historical choice data

used to test the model. These success rates are remarkable only because the pre-cognitive

decision models of accepted wisdom (e.g., expected utility) could not even be tested to see how

well they could predict a set of choice data.


A Real-Life Example: The Malawi Smallholder's Decision
Between Chemical and Organic Fertilizers

How do small-scale farmers in the Third World decide whether or not to use chemical

fertilizers? Why do or don't they use organic fertilizers (manure, compost, green manure) as

substitutes for chemicals? Which constraints to chemical fertilizer use are more important:

farmers' lack of capital or credit, or their indigenous beliefs in organic fertilizers

(manure/compost) as the right way to fertilize their crops, or their fear of dependency on

chemicals?

This decision is a crucial one for African governments facing a "food crisis" in their cities

and trying to increase the food surplus produced by small farmers in the countryside, because

food production is linked to quantity of fertilizer used on most food crops. It is even more

important in a country like Malawi in southern Africa that is land-locked, faces high transport

costs to the sea (due to the cutting of the Beira railroad line in Mozambique) and imports all

chemical fertilizer or its feedstocks. It is a key decision for those policy makers interested in the

potential of sustainable agriculture or low-input agriculture (Brush 1989). The decision model in









8

figures la, lb, Ic, and Id explains why some farmers use both chemical and organic fertilizer,

while others use only chemical, while some use only organic (defined as animal manure, compost,

or certain kinds of green manure). It also tests whether the main reason for nonuse of either kind

of fertilizer is simply farmers' lack of cash or credit, as opposed to an indigenous trust in local

organic fertilizers or a more invidious fear of dependency on chemicals.

In April and May, 1987, the model was built and tested with choice data on fertilizer use

of smallholders in Malawi during personal interviews; the test sample was comprised of 40

farmers in three agricultural districts: Lilongwe, Kasungu, and Salima. Although farmers in this

sample were on average bigger, more experienced farmers than is the norm (with average

landholdings of 3.02 ha.), in other respects the sample is fairly representative of Malawi

smallholders: 26 farmers were credit club members, and 14 were not; 22 farmers got credit for

fertilizer in 1986/87, while 18 did not. Seventeen farmers were women household heads, three

were couples interviewed together, and 20 were male household heads; the sample has a high

proportion of female household heads because Dixon (1982) estimates that women in Malawi

perform about 50 percent of the agricultural labor. Of the 40 farmers interviewed, 33 were

household heads. Questions about the sexual division of labor and income within the family

revealed that groundnuts is a women's cash crop while tobacco, cotton, and hybrid maize are

men's cash crops, and local maize and beans are grown for the whole family's consumption.

Because the outcome chosen by the farmer is different for different crops, this model is

specific to the local variety of maize, which constitutes 90 percent of maize production and is the

staple food crop. Also for simplicity of modeling, it is here assumed that every farmer

incorporates some crop residues on maize during the "banking up" of soil around the secondary

roots after the second weeding and fertilizing. This use of crop residues is no doubt beneficial to

the soil, but is not the same terrific stuff as animal manure or compost. Hence organic fertilizer

here means manure and/or compost and/or green manure, but not crop residues.

The model posits that farmers must first pass a set of simple "elimination-by-aspects"

constraints (Tversky 1972) in figure la. They then must have a need or motivation to use either

or both kinds of fertilizer (figure lb). They then pass to a set of resource constraints specific to









9

each kind of fertilizer, and will use that kind if they satisfy or pass each constraint (figures Ic

and ld). Farmers will use both kinds of fertilizer if they think the crop needs both, and they

pass both sets of resource constraints.

Elimination Criteria


Farmers must first pass a simple set of constraints in figure la which eliminate use of both

organic and chemical fertilizers if: their type of soil doesn't need or respond to either kind of

fertilizer (criterion 1), or their type of local maize seed doesn't need either kind of fertilizer

(criterion 2), or they let most of their land lie fallow for two or more years so that after the

fallow period it doesn't need either chemical or organic fertilizers (criterion 3), or they lack the

cash or credit for either chemical or organic this year (criterion 4). If a farmer is eliminated at

this first stage of the decision process, he or she doesn't have to decide between organic and

chemical fertilizers because both are eliminated and the decision is simple.


Stage-Two Criteria


Farmers who pass "stage 1" criteria do have a complicated decision, however, and continue

on to stage-two criteria in figure lb. If they think that their local maize variety needs both kinds

of fertilizer to produce good yields (criterion 5) and using both kinds is more profitable (criterion

7), they are sent on to both sets of resource constraints in figures Ic and Id. If they think local

maize needs only organic fertilizer (criterion 6) and it is more profitable than chemical, they are

sent only to figure Ic. Similarly, if they think local maize needs only chemical fertilizer

(criterion 6) and it is more profitable than organic, they are sent only to figure Id.

In figure Ic, farmers will apply organic fertilizer to local maize if: they have enough

animals to make enough manure or compost to use on their local maize (or a crop (e.g., tobacco)









9

each kind of fertilizer, and will use that kind if they satisfy or pass each constraint (figures Ic

and ld). Farmers will use both kinds of fertilizer if they think the crop needs both, and they

pass both sets of resource constraints.

Elimination Criteria


Farmers must first pass a simple set of constraints in figure la which eliminate use of both

organic and chemical fertilizers if: their type of soil doesn't need or respond to either kind of

fertilizer (criterion 1), or their type of local maize seed doesn't need either kind of fertilizer

(criterion 2), or they let most of their land lie fallow for two or more years so that after the

fallow period it doesn't need either chemical or organic fertilizers (criterion 3), or they lack the

cash or credit for either chemical or organic this year (criterion 4). If a farmer is eliminated at

this first stage of the decision process, he or she doesn't have to decide between organic and

chemical fertilizers because both are eliminated and the decision is simple.


Stage-Two Criteria


Farmers who pass "stage 1" criteria do have a complicated decision, however, and continue

on to stage-two criteria in figure lb. If they think that their local maize variety needs both kinds

of fertilizer to produce good yields (criterion 5) and using both kinds is more profitable (criterion

7), they are sent on to both sets of resource constraints in figures Ic and Id. If they think local

maize needs only organic fertilizer (criterion 6) and it is more profitable than chemical, they are

sent only to figure Ic. Similarly, if they think local maize needs only chemical fertilizer

(criterion 6) and it is more profitable than organic, they are sent only to figure Id.

In figure Ic, farmers will apply organic fertilizer to local maize if: they have enough

animals to make enough manure or compost to use on their local maize (or a crop (e.g., tobacco)









10

rotated with local maize)1 every two or three years (criterion 8), or they can buy the

manure/compost they need (criterion 9) and they have or can borrow/rent an oxcart and oxen to

carry the manure/compost to their fields (criterion 10), and they have the time or (full- or part-

time) labor to carry it to their fields (criterion 12), which are not too far away to reach by oxcart

(criterion 11). If all these constraints are passed, the model predicts that the farmer applies

manure and/or compost to local maize (or a crop rotated with local maize). If a farmer fails one

constraint, the model predicts no manure/compost is applied.

In figure Id, farmers will apply chemical fertilizer to local maize if: there was chemical

fertilizer available at the time needed, either to buy or get on credit (criterion 13), and the

farmer had either the cash or credit at the time needed to get the fertilizer (criterion 14), and the

farmer could take the risks associated with chemical fertilizers (criteria 15-17).


The Risk Subroutine


The main risk of chemical fertilizer is what I call the "dependency" of the land on chemical

fertilizer. Farmers interviewed in Malawi claim that their land (and in some cases their

traditional seeds or germplasm) get dependent on chemical fertilizer such that, if they stop

applying it for one year, their yields decrease drastically. Whether or not this is due to a change

in the land itself or merely in farmers' expectations of the yields they should get from the land is

irrelevant for the purposes of assessing whether farmers will take the risk of chemical fertilizer.

What is relevant here is whether they consider chemical fertilizer to be risky (criterion 15), and

whether they will take the risk (criteria 16-17). If farmers are worried about the dependency of

the land on chemicals, the model predicts that they will take the risk if: either they have a

farming practice or way to reduce the risk substantially (e.g., also apply manure) (criterion 16) or

they feel they must take the risk anyway because they cannot now return to using only




1 Farmers in Kasungu ADD normally rotate tobacco (the man's cash crop) with local maize every
other year. Farmers that I interviewed apply quite a bit of manure on tobacco: 20 to 50 oxcarts per
acre. Given that the effects of manure and/or compost last for two or more years because they are
slow-release fertilizers, this manure is also considered to be applied to local maize for the purpose
of testing this model.









11

manure/compost without a drastic reduction in their yields (criterion 17). With the latter line of

reasoning, they are weighing the risks of using fertilizer with the risks of dropping it altogether,

which entails a possible drastic reduction in maize yields and consumption and maybe even

hunger. As was the case with manure/compost, the model predicts that chemical fertilizer will be

applied if all the constraints on chemical fertilizer are passed, and not otherwise.

The reader should note that in all stages of this decision model there is a capital or credit

constraint. Results of testing this model should thus allow us to compare the limiting effect of

the cash/credit constraints with the limiting effect of farmers' beliefs such as "Local maize

doesn't need chemical fertilizer." If farmers do have such beliefs, they will drop out of the

decision process in the model at criterion 2 and not even reach the cash/credit constraints in

criteria 4, 9, and 14. This will allow us to determine just how important lack of cash/credit is, as

a factor limiting farmers' use of chemical fertilizer.


Results


Results of the test of this model in Malawi show that of the 40 farmers interviewed, only

eight farmers did not use either chemical fertilizer or manure, due to lack of cash or credit.

These farmers believed quite strongly that local maize needs at least one kind of fertilizer, but

had the money/credit for neither. Thirty-two farmers went on to the tough decision between

organic and chemical fertilizers; 26 farmers felt that both kinds of fertilizer were necessary,

while five farmers felt only chemical was necessary on local maize, and one farmer felt only

organic was necessary. (However, the model erred in his case, because he in fact uses both

chemical and organic.) Thus the model predicts that 31 farmers will use chemical fertilizer if

they can also pass all the constraints on chemical (availability at the right time, cash/credit, and

the farmer's willingness to take the risks of fertilizer). Twenty-six farmers will use organic if

they can also pass the constraints on organic (enough animals to make it or the cash to buy it, an

oxcart available, enough time to apply it, and maize gardens close enough to their home so they

can deliver it).









12

Results of putting the cases down the set of chemical fertilizer constraints show that 28 of

the 30 farmers pass them and apply chemical fertilizer. Only two farmers do not have the

cash/credit and only one farmer cannot take the risk of chemical fertilizer in this sample. This

does not mean there is not a high risk associated with chemical fertilizer. In fact, all farmers

interviewed felt that the land became dependent on chemical fertilizer after initial use, so that

once you started it, you had to keep applying more and more of it. However, three of 29 farmers

did not think this was risky; 19 more farmers had a farming practice to reduce the risk

(complementary manure application, appropriate crop rotations, etc.); while 6 farmers had no

farming practice to reduce the risk but felt they had to take it because otherwise, their yields

would decrease drastically or completely.

Results of putting the 26 cases down the set of organic constraints show that only 15 cases

end up applying manure/compost to local maize; the remaining 11 farmers could not pass one

constraint or another. Only 18 out of the 26 cases either had enough animals to make the

manure/compost or could buy manure in their village. Of those 18 farmers, 2 farmers could not

get hold of oxcarts to transport the manure, and one farmer (a big commercial farmer) did not

have the time to transport and apply manure to local maize. One farmer is an error of the model

because he did pass the organic constraints and should have applied organic but "didn't want to

waste time with manure when chemical fertilizer was available nearby."

Thus results show that, in this sample of Malawi smallholders, only 15 out of 40 farmers

apply manure while 28 of 40 farmers apply chemical fertilizer. Why? The majority of farmers

claim that both kinds of fertilizer, organic and chemical, are necessary; but more of this sample

of farmers can pass the chemical constraints than can pass the organic constraints, which are

stiffer. I conclude that manure/compost is a highly desirable kind of fertilizer, but farmers won't

depend on it as their sole source of plant nutrients when chemical fertilizer is available at

reasonable prices. What Malawi smallholders will do, given the resources, is apply both chemical

and organic fertilizers to local maize as complementary inputs. In their indigenous knowledge

system, chemical fertilizers give plants their nutrients and organic fertilizers build up the

structure of the soil and more importantly, reduce the riskiness of the land's becoming too









13

dependent on chemical fertilizer. For these farmers, application of organic fertilizer is a hedge

against too strong a dependency on chemical fertilizer.


Conclusion


Of what use are decision trees and other cognitive models for applied agricultural

scientists? The goal of decision studies is to model how people make real-world decisions and to

identify the specific decision criteria used by most of the individuals in a group. An applied

social scientist might use these results to identify intervention points in the decision process,

whether they be important constraints blocking a desired action which could be alleviated by

policy makers, or frequently-used reasons for a desired action which could be encouraged by

policy makers with new policies. For example, results from testing the Malawi farmer's decision

to use chemical vs. organic fertilizers show policy makers that lack of capital/credit is the main

constraint to farmers' use of chemical fertilizer, not an indigenous belief in organic fertilizers or

low-input agriculture. Policy recommendations made to the Ministry of Agriculture thus focused

on expanding credit use by smallholders so that more, not less, chemical fertilizer could be used

by smallholders in maize production. Similar policy recommendations can be drawn from any

study of real-life choice people make. For example, previous results helped agronomists and

planners of a rural development project to drop an agronomic recommendation that was impeding

spread of a "package" of new technologies (Gladwin 1976, 1979a, 1979b).

Decision tree models can also provide valuable feedback to social planners about why an

applied project, aimed at helping some target clientele group do something, is failing.

Experience with such projects in Third World settings show that the most frequent reason they

fail is that they end up getting a very good answer to the wrong question. In Mexico, for

example, planners of the Puebla Project designed fertilizer demonstrations so that farmers would

fertilize at planting and increase their yields; most farmers did not adopt because with "the

traditional way," they applied fertilizer before the rains came anyway, and so the new technology

would not increase their yields (Gladwin 1979a).









14

In each case in which the development project failed, social scientists were involved in the

project at the design stage as they should be (Matlon, Cantrell, King, and Benoit-Cattin 1984),

gathering lots of good quantitative data. In the case of the Plan Puebla, for example, good socio-

economic "baseline data" was collected, as well as data about centimeters of rainfall, soil types,

and communication networks (CIMMYT 1971). But still farmers did not adopt. What was

missing? At the start of the project, social scientists did not find out why the farmers

traditionally do what they do. They did not identify farmers' cognitive strategies and the

decision criteria behind "the traditional way" before they tried to improve on it (Gladwin 1979a).

They did not know the farmers' problems before they designed the solution, and so they ended

up getting a good answer to the wrong question; no-one adopted, and the project failed.

An understanding of the knowledge systems of indigenous peoples and a modeling of the

decision processes underpinning their farming systems can stop these failures of applied projects.

If project designers understand the decision criteria and logic used by the target population, they

can find the answers to the right questions. Clearly, the applicability of decision tree

methodology and other cognitive models to development problems and issues is limited only by

our imagination.












REFERENCES


Anderson, Jock
1979"Perspective on Models of Uncertain Decisions." Risk. Uncertainty, and Agricultural
Development, J. Roumasset, J. Boussard, and I. Singh, eds. New York: Agricultural
Development Council, pp. 39-62.


Anderson, J.R., Dillon, J.L., and J.B. Hardaker
1977 Agricultural Decision Analysis. Ames, Iowa: Iowa State Univ. Press.


Arrow, Kenneth J.
1951 "The Nature of Preference and Choice." Social Choice and Individual Values. New
York: Wiley & Sons, Inc.


Barlett, Peggy
1977 "The Structure of Decision Making in Paso." American Ethnologist 4(2): 285-307.


Benito, Carlos A.
1976 "Peasants' Response to Modernization Projects in Minifundia Economies." American
Journal of Agricultural Economics 58(2): 143-151.


Berlin, Brent, and Paul Kay
1969 Basic Color Terms. Berkeley: Univ. of California Press.


Brokensha, David
1989 "Local Management Systems and Sustainability." Food and Farm: Current Debates and
Policies, C. Gladwin and K. Truman, eds. Lantham, Md: University Press of
America.


Brokensha, David, Michael Warren, and Oswald Werner
1980 Indigenous Knowledge Systems. Lantham, Md: University Press of America.


Brush, Stephen
1989 "The Genetic Question in Agricultural Sustainability." Food and Farm: Current
Debates and Policies, C. Gladwin and K. Truman, eds. Lantham, Md: University
Press of America.


CIMMYT (Centro de Investigacion de Maiz y Trigo)
1971 The Puebla Project 1967-69. Mexico, D.F., Mexico: CIMMYT.


CIMMYT Economics Staff
1984 "The Farming Systems Perspective and Farmer Participation in the Development of
Appropriate Technology." Agricultural Development in the Third World, C. Eicher
and J. Staatz, eds. Baltimore: John Hopkins Univ. Press.











D'Andrade, Roy G.
1981 "The Cultural Part of Cognition." Cognitive Science 5: 179-195.

1984 "A Folk Model of the Mind." LaJolla, CA: University of California at San Diego,
Dept. of Anthropology mimeo.


Dixon, Ruth
1982 "Women in Agriculture: Counting the Labor Force in Developing Countries."
Population and Development Review 8(3): 539-566.

Frake, Charles
1964 "How to Ask for a Drink in Subanun." American Anthropologist 66(2): 127-132.


Gardner, Howard
1985 The Mind's New Science. New York: Basic Books.


Gladwin, Christina H.
1975 "A Model of the Supply of Smoked Fish from Cape Coast to Kumasi." Formal
Methods in Economic Anthropology, S. Plattner ed. Wash., D.C.: A Special
Publication of the American Anthropological Association, No. 4, 77-127.

1976 "A View of the Plan Puebla: An Application of Hierarchical Decision Models."
American Journal of Agricultural Economics 58(5): 881-87.

1977 A Model of Farmers' Decisions to Adopt the Recommendations of Plan Puebla. Ph.D.
Thesis, Stanford University.

1979a "Cognitive Strategies and Adoption Decisions: A Case Study of Nonadoption of an
Agronomic Recommendation." Economic Development and Cultural Change 28(1):
155-173.

1979b "Production Functions and Decision Models: Complementary models." American
Ethnologist 6(4): 653-674.

1980 "A Theory of Real-Life Choice: Applications to Agricultural Decisions." Agricultural
Decision Making: Anthropological Contributions to Rural Development, P. Barlett, ed.
New York: Academic Press, pp. 45-85.

1983 "Contributions of Decision-Tree Methodology to a Farming Systems Program."
Human Organization 42(2):146-157.

1989 Ethnographic Decision Tree Modeling. Beverly Hills, CA: Sage.


Gladwin, Christina, and Robert Zabawa
1984 "Microdynamics of Contraction Decisions: A Cognitive Approach to Structural
Change." American Journal of Agricultural Economics 66(5): 829-835.

1986 "After Structural Change: Are Part-Time or Full-Time Farmers Better Off?"
Agricultural Change: Consequences for Southern Farms and Rural Communities,
Joseph Molnar, ed. Boulder (CO): Westview Press, pp. 39-60.












1987 "Transformations of Full-Time Farms in the U.S.: Can They Survive?" Household
Economies and Their Transformations, M. Maclachlan, ed. Philadelphia (PA):
University Press of America.


Gladwin, Hugh
1971 Decision Making in the Cape Coast (Fante) Fishing and Fish Marketing System.
Ph.D. Thesis, Stanford University.

1974 "A Study of the Relationship Between Verbalization and Deeper Cognitive Skills in
Learning a Complex Task." Final Report, Project No. 2-0650, Grant No. OEC 0-72-
1879, U.S. Dept. of Health, Education, and Welfare.


Gladwin, Hugh, and Michael Murtaugh
1984 "Test of a Hierarchical Model of Auto Choice on Data from the National
Transportation Survey." Human Organization 43(3):217-226.


Harris, Marvin
1974 "Why a Perfect Knowledge of All the Rules One Must Know to Act Like a Native
Cannot Lead to Knowledge of How the Natives Act." Journal of Anthropological
Research 30(4):242-251.

1979 Cultural Materialism: The Struggle for a Science of Culture. New York: Random
House.


Kahneman, D, and A. Tversky
1972 "Subjective Probability: A Judgment of Representativeness." Cognitive Psychology 3:
430-54.

1982 "The Psychology of Preferences." Scientific American 246: 160-173.


Lancaster, Kevin
1966 "A New Approach to Consumer Theory." Journal of Political Economy 74: 132-157.


Malinowski, Bronislaw
1922 Argonants of the Western Pacific. London, England: Routledge and Kegan Paul, Ltd.


Matlon, Peter, Ronald Cantrell, David King, and Michel Benoit-Cattin
1984 Coming Full Circle: Farmers' Participation in the Development of Technology.
Ottawa, Canada: International Development Research Centre.


Miller, George
1956 "The Magic Number Seven, Plus or Minus Two: Some Limits on our Capacity for
Processing Information." Psychological Review 63: 81-97.


Miller, George, Eugene Galanter, and Karl Pribram
1960 Plans and the Structure of Behavior. Holt, Rinehart, and Winston, Inc.












Mukhopadhyay, Carol
1984 "Testing a Decision Process Model of the Sexual Division of Labor in the Family."
Human Organization 43(3):227-242.


Murtaugh, Michael, and Hugh Gladwin
1980 "A Hierarchical Decision-Process Model for Forecasting Automobile Type Choice."
Transportation Research 14A: 337-348.


Newell, Alan, and Herbert Simon
1972 Human Problem Solving. Englewood Cliffs, N.J.: Prentice-Hall, Inc.


Officer, R.R., and A.N. Halter
1968 "Utility Analysis in a Practical Setting." American Journal of Agricultural Economics
50: 257-277.


Pike, Kenneth
1954 Language in Relation to a Unified Theory of the Structure of Human Behavior. The
Hague: Mouton.


Quinn, Naomi
1971 "Simplifying Proceedures in Natural Decision-Making." Paper presented at the
Mathematical Social Science Board Seminar in Natural Decision Making Behavior, Palo
Alto, California, Nov. 22-25.

1978 "Do Mfantse Fish Sellers Estimate Probabilities in Their Heads?" American
Ethnologist 5(2): 206-226.

1989 A Model of American Marriage. Cambridge: Cambridge University Press.


Raiffa, Howard
1968 Decision Analysis. Reading, MA.: Addison-Wesley.


Romney, Kim, and Roy G. D'Andrade
1964 "Cognitive Aspects of English Kin Terms." American Anthropologist 66: 146-170.


Schank, Roger, and Robert Abelson
1977 Scripts. Plans. Goals and Understanding. New York: Wiley and Sons.


Schoemaker, Paul
1982 "The Expected Utility Model: Its Variants, Purposes, Evidence, and Limitations."
Journal of Economic Literature 20: 529-563.


Schoepfle, Mark, Michael Burton, and Frank Morgan
1984 "Navajos and Energy Development: Economic Decision Making Under Political
Uncertainty." Human Organization 43(3): 265-276.









19

Spradley, James
1979 The Ethnographic Interview. New York: Holt, Rinehart, and Winston.


Tversky, Amos
1967 "Additivity, Utility, and Subjective Probability." Decision Making, W. Edwards and
A. Tversky, eds. Penguin Books, pp. 208-239.

1969 "The Intransitivity of Preferences." Psychological Review 76: 31-48.

1972 "Elimination by Aspects: A Theory of Choice." Psychological Review 28: 1-39.


Tversky, A., and D. Kahneman
1981 "The Framing of Decisions and the Psychology of Choice." Science 211: 453-458.


Young, James C.
1980 "A Model of Illness Treatment Decisions in a Tarascan Town." American Ethnologist
7(1): 106-131.

1981 Medical Choice in a Mexican Village. New Brunswick, N.J.: Rutgers University
Press.


Werner, Oswald, and G. Mark Schoepfle
1987 Systematic Fieldwork, Vol 1 and 2. Newbury Park, CA: Sage Press.


Zabawa, Robert
1984 The Transformation of Farming in Gadsden County, North Florida. Ph.D.
dissertation, Northwestern University.








Figure la. The Decision Between Organic and Chemical Fertilizers on
Local Maize: Stage 1 Contraints.
A*0 c se
(Apply organic; chemical; both)


Does your soil need
or respond to either chemical
or organic fertilizers?


Eliminate
no Both
0 CoVvs


yes


Does your local maize
variety need either chemical or
organic fertilizers?
ye
yes


Eliminate
----)
no Both

0 Cases


Do you let most of your land
lie fallow for two or more years Eliminate
at a time so that afterwards you yes Both
don't need to apply chemical or
organic fertilizers? 0 Cases


no; I need them
afterwards


Have the cash or credit Eliminate
to apply either chemical or no Both
organic fertilizers this year?
1 3 Cases
yes

Go on to figure
Go on to figure lb


3R ea ses









Figure Ib. Motivations to Use Chemical or Organic or Both.


Given you've
passed figure ja
-3Q Is there a necessity
to use both chemical and
organic fertilizers on
local maize?


yes /
0. cases


Is it profitable
to use both?


yes /
Apply both organic and
chemical fertilizers if
you also pass constraints
in figures Ic and Id.

Cases





Organic

/
Apply only organic
fertilizer if you
also pass constraint
in figure Ic.


no
Sc-ases




no






Which type of
fertilizer does local maize
need more? Which is
more profitable?


Chemical


Apply only chemical
fertilizer if you
:s also pass all
constraints in
figure ed


1 case
(1 crror)


5" COses










Figure 1c. Resource Constraints to Use of Organic Fertilizer.
0% eases
"Do you have the animals
to make enough manure/compost
to use on local maize (or crop
rotated with local maize) every
two or three years?
yes no
\ caq \ c ase5
?Can you buy
the manure/compost
you need?
/ \
yes no
/I eases \\ e ses
No Organic
Can you get an
oxcart and oxen
to carry manure to -- no No Organic
your gardens?
I ; cOases
yes : /6 Cases

"/Is all your local maize No
planted too far away from -- yes---*Organic
your house to carry manure
to?
no : /6 ?ases

Do you yourself have the time No
or can you hire labourers to carry-no -- Organic
manure to your gardens?
I / ease
yes

1 I
Apply manure/compost
to local maize

15' CQseS
( ee ro r)










Figure Id. Resource Constraints to Use of Chemical Fertilizer.
3.1 01 Se &
irhis year, is there
chemical fertilizer available--no--- No
to buy or get on credit Chemical
at the time needed?

yes : 3 c 1 ses


IDo you have the cash
or credit to apply chemical---- no -*No
on local maize? Chemical


yes;: k cases

Do you think that the
dependency of the land on chemical
fertilizer is risky?


yes: &~ CQSeS



/6 u
Do you have a
farming practice that
you can use to reduce the
risk of dependency?


yes


Apply Chemical
Fertilizer

/? eases


no 3 cases


Apply Chemical
Fertilizer


no : 7- ca4S


'gRisk L
of dependency


Apply Chemical
Fertilizer

(P caseS


Risk
of dropping
fertilizer and
inviting hunger

o

Don't Apply
Chemical

I case
(1 error)




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs