A Framework for Mapping and Comparing …...1 A framework for mapping and comparing behavioral...
Transcript of A Framework for Mapping and Comparing …...1 A framework for mapping and comparing behavioral...
Center for Behavior, Institutions and the Environment
CBIE Working Paper Series
#CBIE-2015-010
A Framework for Mapping and Comparing Behavioral Theories in Modelsof Social-Ecological Systems
Maja SchluterStockholm Resilience Centre
Andres BaezaNational Socio-Environmental Synthesis Center (SESYNC)
Gunnar DresslerUFZ, Helmholtz Centre for Environmental Research e UFZ
Karin FrankUFZ, Helmholtz Centre for Environmental Research e UFZ
Jurgen GroeneveldUFZ, Helmholtz Centre for Environmental Research e UFZ
Wander JagerUniversity College Groningen
Marco A. JanssenSchool of Sustainability
Ryan R.J. McAllisterCSIRO
Birgit MullerUFZ, Helmholtz Centre for Environmental Research e UFZ
Kirill OrachStockholm Resilience Centre
Nina SchwarzUFZ, Helmholtz Centre for Environmental Research e UFZ
Nanda WijermansStockholm Resilience Centre
December 16, 2015
The Center for Behavior, Institutions and the Environment is a research center located within the Biosocial ComplexityInititive at ASU. CBIE can be found on the internet at: http://csid.asu.edu. CBIE can be reached via email [email protected].
c©2015 M. Schluter. All rights reserved.
2
A Framework for Mapping and Comparing Behavioral Theories in Models ofSocial-Ecological Systems
Maja Schlutera, Andres Baezab, Gunnar Dresslerc Karin Frankd Jurgen Groenevelde Wander Jagere
Marco A. Janssene Ryan R.J. McAllistere Birgit Mullere Kirill Orache Nina Schwarze NandaWijermanse
aStockholm Resilience Centre;bNational Socio-Environmental Synthesis Center (SESYNC);cUFZ, Helmholtz Centre for Environmental Research e UFZ;dUniversity College Groningen;eSchool of Sustainability;fCSIRO;
Corresponding author:Maja SchluterStockholm Resilience Centre, Stockholm University, Krftriket 2b, 10691 Stockholm, [email protected]
Abstract:Formal models are commonly used in natural resource management (NRM) to studyhuman-environment interactions and inform policy making. In the majority of applications, humanbehavior is represented by the rational actor model despite growing empirical evidence of itsshortcomings in NRM contexts. While the importance of taking into account the complexity of humanbehavior is increasingly recognized it remains a major challenge to integrate it into formal models.The challenges are multiple: i) there exists many theories scattered across the social sciences, ii) mosttheories cover only a certain aspect of decision-making, iii) they vary in their degree of formalization,iv) causal mechanisms are often not specified. With this paper we provide a framework to facilitate abroader inclusion of theories on human decision making in formal NRM models. It serves as a tool andcommon language to describe, compare and communicate alternative theories. In doing so, we notonly enhance understanding of commonalities and differences between theories, but also support a firststep in tackling the challenges mentioned above. This approach may enable modelers to find andformalize relevant theories, and be more explicit and inclusive about theories of human decisionmaking in the analysis of social-ecological systems.
1
A framework for mapping and comparing behavioral
theories in models of social-ecological systems Maja Schlütera, Andres Baezab, Gunnar Dresslerc, Karin Frankc, Jürgen Groeneveldc, Wander Jagerd,
Marco A. Janssene, Ryan R.J. McAllisterf, Birgit Müllerc, Kirill Oracha, Nina Schwarzc, Nanda
Wijermansa
a Stockholm Resilience Centre, Stockholm University, Kräftriket 2b, 10691 Stockholm, Sweden,
[email protected], [email protected], [email protected]
b National Socio-Environmental Synthesis Center (SESYNC), 1 Park Place, Suite 300, Annapolis, MD
21401, USA, [email protected]
c UFZ, Helmholtz Centre for Environmental Research e UFZ, Department of Ecological Modelling,
Permoser Str. 15, 04138 Leipzig, Germany, [email protected], [email protected],
[email protected], [email protected], [email protected]
d University College Groningen, Hoendiepskade 23-24, 9718 BG Groningen, The
Netherlands,[email protected]
e School of Sustainability, Arizona State University, PO Box 875502, AZ 85287-5502Tempe, USA,
f CSIRO PO Box 2583 Brisbane Q 4001 Australia, [email protected]
Corresponding author
Maja Schlüter
Stockholm Resilience Centre
Stockholm University
Kräftriket 2b
10691 Stockholm, Sweden
2
Abstract 1
Formal models are commonly used in natural resource management (NRM) to study human-2
environment interactions and inform policy making. In the majority of applications, human behavior 3
is represented by the rational actor model despite growing empirical evidence of its shortcomings in 4
NRM contexts. While the importance of taking into account the complexity of human behavior is 5
increasingly recognized it remains a major challenge to integrate it into formal models. The 6
challenges are multiple: i) there exists many theories scattered across the social sciences, ii) most 7
theories cover only a certain aspect of decision-making, iii) they vary in their degree of formalization, 8
iv) causal mechanisms are often not specified. With this paper we provide a framework to facilitate a 9
broader inclusion of theories on human decision making in formal NRM models. It serves as a tool 10
and common language to describe, compare and communicate alternative theories. In doing so, we 11
not only enhance understanding of commonalities and differences between theories, but also 12
support a first step in tackling the challenges mentioned above. This approach may enable modelers 13
to find and formalize relevant theories, and be more explicit and inclusive about theories of human 14
decision making in the analysis of social-ecological systems. 15
16
1. Introduction 17
Formal models have been used extensively to study the interactions between humans and their 18
environment, as well as inform policy making (e.g., Meadows et al, 1972; Clark, 1976; Nordhaus, 19
1994). In natural resource management (NRM), modelling has advanced our understanding of the 20
dynamics of natural resources, their response to management interventions and environmental 21
change, as well as their vulnerabilities and regenerative capacities. This has informed policy 22
decisions on harvest quotas, CO2 emission budgets, agri-environmental schemes as well as the 23
identification of biodiversity hotspots and corridors for inclusion in protection programs 24
(Karagannakios, 1996; Meinshausen et al, 2009; Myers et al., 2000; Simberloff and Cox, 1987). The 25
focus on understanding natural resource dynamics and their optimal management has, however, 26
neglected humans and their behavior as a key uncertainty for sustainable management (Fulton et al. 27
2011). Given that natural resource use systems are social-ecological systems in which humans shape 28
and depend on their biophysical environment (Berkes and Folke 1998), their adaptive responses to 29
policy and environmental change cannot be neglected (e.g., Palmer and Smith 2014). Therefore, 30
modelling approaches need to explicitly combine ecological dynamics and human behavior to 31
address the interactions between the different domains. However, integrating human behavior into 32
formal models on natural resource use and management is still a major challenge (Janssen and 33
Jager, 2000; Fulton et al. 2011, Milner-Gulland 2012, Schlüter et al. 2012). 34
Common approaches for integrating human behavior into formal models of social-ecological systems 35
couple economic theory with resource dynamics (e.g., Clark, 1976; Nordhaus, 1994), capture human 36
aggregated responses in feedback loops (e.g., Meadows et al., 1972), or use ad hoc assumptions 37
(Smajgl and Barreteau, 2014). While the former is prescriptive in that it aims to determine the 38
optimal resource management strategy or the optimal policy option given a set of constraints, the 39
latter two aim to describe actual system dynamics by explicitly incorporating human behavior. The 40
first approach is often based on very simple assumptions about human decision-making, namely the 41
concept of the selfish rational actor, also referred to as homo economicus. The frequent use of the 42
3
rational actor in modelling human behavior and decision-making in NRM is not surprising since it is 43
the standard model in economic theory and is straightforward enough to include in mathematical 44
formulations. However, the key assumptions of the rational actor—that she has perfect knowledge 45
and stable preferences, is selfish and makes calculations to identify an optimal decision that 46
maximizes utility—are in contrast with empirical observations of how people actually make decisions 47
concerning natural resource use (Siebenhuner, 2000; Van den Bergh et al., 2000). Also the 48
assumption that these “deviations from optimal behavior” can be considered as “noise,” and hence 49
would cancel out in large populations does not hold because much of these deviations are 50
systematic. For example, in real life, people have their cultural habits, learn from other people, and 51
often obtain utility from interacting with and helping other people (e.g., see Gintis 2000; Fehr and 52
Gintis 2007, Fehr and Schmidt 1999). Such behavioral drivers and processes are expected to have 53
consequences for the dynamics and performance of social-ecological systems at large. 54
The importance of including the relevant complexity of human behavior in the study of human-55
environmental interactions has been alluded to in recent publications (e.g., Janssen and Jager, 2000; 56
Worldbank, 2015). The World Bank’s report on “Mind, Society and Behavior” (Worldbank, 2015) 57
explicitly acknowledges the importance of capturing the most advanced understanding of how 58
humans think and how context shapes thinking for the design and implementation of policies. 59
Others argue that the current focus on a small set of theories of human decision-making in policy 60
assessment (such as climate policy) limits the relevance of those exercises (Victor, 2015). Since 61
formal models are used to inform policy making, the lack of inclusion of social science expertise can 62
considerably limit both the usefulness of formal models and the effectiveness of policies. 63
There is an abundance of theories in the social sciences that describe and test how people behave in 64
various contexts. For example, in social and cognitive psychology, research has focused on processes 65
of decision-making (e.g., Todd, Gigerenzer & the ABC Research Group 2012), social influence (e.g 66
Cialdini and Goldstein, 2004), information processing (e.g., Anderson 1990), time discounting (e.g., 67
Hardisty et al, 2012), and reinforcement learning (Skinner, 1953), just to mention a few. Theories 68
have been developed on the gains and losses of group decision-making and situational and 69
procedural contexts that affect outcomes (for an overview see e.g., Kerr and Tindale, 2004). In 70
behavioral economics, the focus is directed at heuristics and biases, prospect theory and the framing 71
of decisions (see e.g., Kahneman, 2003). However, this impressive body of knowledge has barely 72
found its way into the field of natural resource management in general and social-ecological systems 73
modelling of resource management contexts in particular. 74
Modelers who aim to introduce alternative theories on human behavior and decision-making in their 75
models of natural resource management, however, face several challenges (see section 2 for a more 76
detailed discussion): (i) the vast amount of theories on human decision-making, some of which are 77
even competing, makes orientation in the field very difficult. Moreover, the available knowledge is 78
fragmented across different disciplines and disciplinary languages. Theories can have different foci, 79
such as emphasizing the importance of selected social or environmental aspects. (ii) As a 80
consequence, some theories on human decision-making address very detailed aspects of decision-81
making, while others are very broad and comprehensive. Modelers need to recognize this diversity 82
in scope and aims and may even have to combine several theories in order to model the process of 83
human decision-making in a comprehensive way. (iii) Depending on their methodological 84
background (experimental, conceptual, empirical), theories on human decision-making vary also 85
with respect to the degree of formalization. This implies that modelers will have to specify elements 86
4
and/or processes of these theories to varying extents. (iv) Modeling social-ecological systems means 87
simulating these systems over time, requiring the specification and representation of causalities in 88
the models. Many theories on human decision-making tend to focus on correlations and thus lack an 89
understanding of causal mechanisms that can be translated into a dynamic model. Modelers, thus, 90
have to make assumptions about causalities when using such theories in their models. Overall, these 91
issues make the selection of relevant theories for natural resource management situations and their 92
formalization in social-ecological models and comparison very challenging. 93
This manuscript is a modest step toward providing a framework to address these challenges and 94
facilitate a broader inclusion of knowledge on human decision-making in formal models of SES. The 95
aim is to support the identification, operationalization and integration of alternative behavioral 96
theories into formal models of natural resource management. More specifically we aim to 97
encourage modelers to think more systematically about the implementation of human decision-98
making in their models and make use of the diversity of human decision-making theories from the 99
social sciences where possible. The purpose of this framework is therefore threefold: 100
● to provide a tool and common language for mapping, describing, organizing, comparing and 101
communicating theories of human decision-making and by doing so 102
● to enhance understanding of commonalities and differences such that modelers can make 103
informed choices of which theory is relevant for a given context and research question, and 104
● to support the operationalization of behavioral theories in formal models by providing 105
guidance on relevant factors and processes of decision-making and facilitating a more 106
systematic implementation procedure 107
To provide a framework that meets these purposes is an ambitious goal. In order to make concrete 108
progress, we narrowed down the type of decisions we were focusing on. The result is a focus on 109
resource users (representing individuals, households or villages) making decisions on when, where, 110
how and how much to appropriate from a resource— these are decisions on what crop to plant, 111
where to fish and how many trees to cut. We do not include, for now, higher- level collective choice 112
decisions, such as changing institutional rules, but we do include decisions on compliance to rules 113
and social norms. 114
The remainder of the paper is organized as follows: In section 2, we discuss the challenges modelers 115
face when formally modelling human behavior. In section 3, we introduce the framework and apply 116
it to a number of concrete theories from various social sciences in section 4. In section 5, we discuss 117
the framework and conclude by considering how we may use this framework to implement different 118
theories in simulation models, compare them and inform policy analysis. 119
120
2. The challenge of formally modelling human decision-making 121
Modelers need to overcome a number of challenges if they want to include theories on human 122
decision-making in their dynamic models. As outlined above these relate to the need to navigate a 123
vast amount of theories scattered across many fields; to deal with different foci, levels of 124
comprehensiveness and formalization; and to assign causality where a theory may remain rather 125
unspecific. In the following we briefly discuss these challenges in more detail. 126
5
The first challenge is finding relevant theories and navigating the vast amount of available theories 127
and knowledge about human decision-making. Various social sciences (economics, psychology, 128
anthropology, sociology, political science, etc.) study human behavior in diverse and sometimes 129
specialized contexts, using various theoretical approaches and terminology. In fact, each discipline or 130
even sub-discipline experiences a range of alternative theories that aim to explain very specific 131
aspects of human behavior. For example, explaining why people cooperate in one-shot prisoner’s 132
dilemma experiments can depend on cultural factors, emotions, cognition, and neural or hormonal 133
processes (Henrich et al., 2001; Hopfensitz and Reuben, 2009; Moll and Tomasello, 2007; Rilling et 134
al., 2002; Eisenegger et al., 2011). This has lead to many specific theories that relate to partial 135
processes of human decision-making and behavior, next to a few generic (overlapping) theories. The 136
Theory of Planned Behavior (Ajzen, 1991), for instance, is a more generic theory describing how 137
behavioral beliefs and attitudes, normative beliefs and subjective norms, together with control 138
beliefs lead to an intention to perform a behavior and influence the actual execution of that 139
behavior. Hence this theory describes a full process, ranging from the formation of beliefs towards 140
the performance of behavior. An example of a specific theory would be the Elaboration Likelihood 141
Model (Petty and Cacioppo, 1984). This theory addresses the process of persuasion. It discriminates 142
between central processing, where involved and capable people process arguments in shaping their 143
beliefs, and a peripheral route, where less involved and less capable people evaluate cues—such as 144
the attractiveness of the source—in in shaping beliefs. This fragmentation of knowledge makes it 145
challenging to know the implications of different theories for different context interacting with each 146
other at a system level. 147
A second challenge is that, from a modeler’s perspective, decision-making theories are often not 148
complete. Formalizing a theory often uncovers logical gaps in the theory that must be filled in order 149
to make a simulation work. Assumptions thus have to be made by the modeler that were not given 150
by the theory (Sawyer, 2004). For example, when modelling human decision-making according to the 151
Theory of Planned Behavior (Ajzen, 1991) one needs to specify the subjective norm, which refers to 152
an individual's perception about how significant others (e.g., parents, spouse, friends, teachers) 153
would judge the behavior under consideration. For the modeler the theory does not specify this in a 154
way that allows for formalization. It remains open what other agents are “significant others” and 155
what their level of significance is. The theory does not specify what principle should guide the 156
formalization of the significance of other agents. Obviously, more specific theories can serve as a 157
guide in this formalization, such as perspectives on interpersonal attraction. Variables can be 158
identified (such as similarity, expertise, physical attractiveness, familiarity) that have an effect on 159
who is considered to be a significant other, and, obviously, this differs among people, among 160
situations, and over time, confronting the modeler with a vast unlimited horizon. This requires that a 161
modeler identify the main drivers to be modelled, and make “crude assumptions” on the modelling 162
details and then carefully check the robustness of model results to these assumptions. These 163
challenges are not unique to formalizing of theories of human behavior. In fact most theories in the 164
life and social sciences experience some ambiguities when translated into formal equations. This is in 165
contrast to many theories in economics and physics that are described through mathematical 166
equations. 167
The third challenge is introducing causality. Simulating human interactions with the environment 168
over time requires the specification of causal relationships about how psychological, social and 169
environmental factors influence an agent’s decision-making. Many of the theories, however, do not 170
6
specify causal mechanisms since they are based on empirical correlations and focus on the 171
relationship between factors at one moment in time. Modelers thus need to make explicit 172
assumptions on processes and mechanisms, for example, how the provision of factual information 173
about the actual behavior of individuals within a community, such as the average energy 174
consumption in a neighborhood or the amount of alcoholic consumption in a student population, 175
affect decision-making (Cialdini and Goldstein, 2004). To translate this into causal relations for a 176
model, we need to specify functional forms on how information about the behavior of others 177
changes the likelihood of certain choices. This relationship can be explained in various ways, for 178
example, by the fact that people want to conform to others or that the option that is dominant in 179
the community becomes more salient. Ambiguities or lack of knowledge on how variables relate to 180
each other and determine the development of the system over time make it very important to be 181
explicit about assumptions made and their implications for model results. 182
In sum, the challenge of modelers is to identify and transform relevant theories on human decision-183
making into crisp causal relationships, while the best available knowledge is fragmented, context 184
dependent and descriptive. Given these challenges it is no surprise many models rely on rational 185
choice, which is based on a clear, unified theory that has and can be easily formulated in 186
mathematical equations. But rational choice theory does not represent our empirical knowledge on 187
human decision-making. By providing a framework to communicate the knowledge from various 188
relevant theories we want to support a first step in tackling these challenges and enable the 189
modelling community to find, filter and formalize relevant theories. This may enable them to be 190
more explicit and inclusive about the various theories of human decision-making in the analysis of 191
social-ecological systems. 192
193
3. A framework of human behavior in natural resource use contexts 194
We took an iterative approach in developing the framework that involved formulating its elements, 195
mapping theories and revising the formulation. We based its formulation on insights from a review 196
of reviews of human decision-making in social-ecological systems (Cooke et al. 2009; Meyfroidt, 197
2013; Scarlett et al. 2013; Van Vugt and Griskevicius, 2014) and in agent-based models (Carley and 198
Newell, 1994; Bousquet and Le Page, 2004; Hare and Deadman, 2004; Matthews et al. 2007; 199
Heckbert et al. 2010; An, 2012; Balke and Gilbert, 2014) as well as experience with our own 200
implementations of different decision-making models in agent-based models of natural resource 201
management. Definitions given below for the components of the framework draw upon general 202
definitions, e.g., given by the Merriam Webster dictionary, to avoid biasing the framework towards 203
one specific scientific discipline. 204
A framework that supports communicating and comparing different theories of human decision-205
making needs to be generic enough to capture the majority of theories and at the same time allow 206
for a meaningful distinction between them. With this aim in mind we decomposed the decision-207
making process within an individual into three major parts: 1) what comes in (perception), what goes 208
out (behavior) and what happens in between (i.e., rules’ and representations that lead to the 209
selection and execution of a behavior) (Figure 1). In Figure 1, the outer box represents the social and 210
biophysical environment and thereby the decision context of an individual. The individual herself 211
(inner box) is represented by the structural elements (state and perceived behavioral options) and 212
processes involved in decision-making. Decision-making involves both conscious and unconscious 213
7
processes that lie in the interface of the individual and the environment (perception and behavior) 214
and internal processes (evaluation and selection). We argue that different theories of behavior can 215
be described as alternative configurations (presence and/or specification) of the structural elements, 216
processes and context of an individual. 217
The seven elements that make up our framework are defined and further detailed in Table 1. They 218
are connected through flows of information or processes that make up the decision-making process 219
as indicated by the solid arrows in Figure 1. An agent perceives the state of its environment, 220
evaluates the information and possibly updates its state. The state and the perceived behavioral 221
options enter the selection process, where the behavioral option that fulfills given 222
goals/needs/satisfaction criteria is identified. The behavior is executed and affects the state of the 223
social and bio-physical environment. For simplicity we focus on the main interactions between the 224
elements, however, other ones are possible and likely. It is also important to note that not every 225
theory includes or specifies all of the steps and that they do not necessarily follow this sequence. 226
The dashed arrows represent the influence of one element on another. The state of an agent for 227
instance may influence the perceived behavioral options, by either constraining an originally broader 228
set of perceived behavioral options (e.g., due to limited available assets) or by enabling the agents to 229
choose from additional behavioral options (e.g., due to new knowledge). The set of perceived 230
behavioral options can also change over time as the result of learning, forgetting or changes in 231
attention. Furthermore, the state may impact on the selection process by activating a different 232
selection process (e.g. due to being dissatisfied). Finally, the perceived behavioral options may 233
influence the search process regarding the search routines that can be executed on a given set (e.g., 234
an optimization is not useful for a set with only one option). 235
8
236
Figure 1. The framework of individual decision-making, allowing for the comparison of selected 237
behavioral theories to model human behavior. Solid arrows and corresponding ellipses indicate 238
processes, boxes represent structural elements. Dashed arrows represent an influence of one 239
element on another, e.g. the state influencing the set of perceived behavioral options. 240
9
Table 1: Definition and specifications of the different elements of the framework. Definitions are adapted from the Merriam Webster dictionary.
Element Definition Specification/Examples
Context
Social & Biophysical Environment
The environment the individual and her behavior are embedded in
Social Env: actors and institutions1 that might affect individual decision-making through information exchange,
coordination, satisfying need for belonging, etc. Biophysical Env: the biophysical properties and dynamics of elements such as a resource, a population or an ecosystem relevant for a particular decision context. Actors’ behavior can affect the amount, quality and location of a resource or population or the integrity of an ecosystem such as its food-web or habitat quality.
Structural Elements
State The internal state of an individual Attributes of an individual that influence the behavior selection process and possibly the perceived behavioral options. There are four classes of attributes: needs/goals, knowledge, assets, values
State: Needs/Goals
Physiological, psychological or material requirements for the well-being of an individual
Needs are motivational goals/factors for behavior. Theories often include one or several of the following: utility, financial income, safety, reputation. Maslow developed a hierarchy of needs: physiological, safety, social, esteem and self-actualization needs (Maslow, 1943). Max-Neef classified subsistence, protection, affection, understanding, participation, leisure, creation, identity, freedom as basic human needs (Max-Neef, 1991).
State: Knowledge
The information and understanding an individual has about her social-ecological environment and her own behavior within this context
A) Declarative or factual: knowledge about the state and dynamics of the ecological system, the relation between actions and outcomes, memory of the outcomes of past actions or past system states; B) Procedural: knowledge about behaviors and the skills/abilities to perform; C) Relational: knowledge about the behavior and opinions of other relevant actors , e.g. knowledge about other actors the individual knows and what they typically do
State: Assets
Resources and other advantageous characteristics of an individual
A) Personal (e.g., skills), B) Social (e.g., social networks, trust); C) Financial assets. Note that knowledge or value aspects can also be considered assets in some theories.
State: Values
Something (as a principle or quality) intrinsically valuable or desirable, i.e. not directly linked to the well-being of an individual or her motivational goals
Values reflect deep, slowly changing beliefs, e.g. a conservation value reflecting how an agent values the conservation of nature per se without any direct monetary benefits or the value of future benefits (discount rate). Note that depending on the context and on the theory, considering others could be either a need or a value. Similarly, preferences can be related to either values or needs.
Perceived behavioral options
The set of options the individual perceives and thus can choose from
Continuous, e.g. the amount of time an agent allocates between labor and leisure; or discrete, e.g. a set of behaviors the agent can choose from.
Processes
Perception The process by which an individual senses the surrounding social and
Complete: the individual receives all possible information from the environment that is relevant to a decision; or Incomplete: i.e., the individual does not receive all information and as such is bounded in her knowledge
1 in the sense of formal and informal rules/norms
10
biophysical environment
Evaluation The process by which an individual determines the significance, worth, or condition of the perceived state of the social and bio-physical environment
For example evaluation of the outcome of a past decision with respect to the returns provided, of fairness or equity of distribution of benefits within a group/society, or of the compliance with a social norm in the population.
Selection The process by which an individual chooses her behavior from the set of perceived behavioral options taking its state into account
Consists of either one or a multiple of different types of selection processes, e.g. ● Random: randomly select one of the possible options ● Optimization: evaluating all options and choosing the one with highest expected outcome ● Satisficing: evaluating options until a behavior is found that is expected to satisfy ● Imitation: selecting the most popular behavior of your friends, or larger community. ● Social Comparison: evaluating options observed among others similar to you and selecting the one with the
highest possible expected outcome. ● Habitual (2 selection processes): 1. Automatic: (If satisfied) Repeating the behavior performed earlier otherwise
switch to 2. Deliberate: (If not satisfied) any other selection process to select, explore and evaluate another behavior.
Behavior the action that an individual executes as a result of the decision process
Behavior impacts the socio-environmental system and, in addition to perception, is the second interface between an individual and its environment. Selected behaviors may fail to be executed if the behavior is physically impossible.
11
4. Positioning different behavioral theories in the framework 241
We will now apply the framework to a selection of well-known theories of decision-making from 242
different disciplines. By doing so we demonstrate its potential use and highlight how theories differ 243
from each other, but also indicate challenges of mapping different theories. We selected the 244
example theories with the aim to span a broad range of theories with respect to their application in 245
different fields (economics, psychology) and their comprehensiveness in covering different drivers of 246
human behavior (individual, social, environmental). Rational actor and bounded rationality are 247
theories widely used in natural resource management. Prospect theory is a rather specialized 248
psychological theory about individual perceptions of gains and losses used in economics. 249
Reinforcement learning is a behavioral theory of how very basic processes of reward shape future 250
behavior. The theory of planned behavior is a comprehensive theory widely used in environmental 251
psychology to explain pro-environmental behavior. And, finally, the theory of descriptive norms 252
focusses on the role of the social context. 253
For the mapping we went through the concepts and relationships of each theory and the elements 254
of the framework in an iterative way to relate a concept from the theory with a framework element. 255
We for example identified self-interest as a value of the rational actor (state), utility maximization as 256
its needs (state) and optimization as the selection process (Figure 2). When a theory specifies a 257
certain element the specific details of the theory with respect to this element are included in the 258
figure. Elements that are not mentioned in the theory still remain in the mapping figure; however, 259
they do not include any details (e.g. evaluation in the rational actor model, Figure 2). 260
12
4.1. Rational actor 261
262
Figure 2: The rational actor mapped to the framework. 263
The rational actor theory (homo economicus) takes root in neoclassical economics. Scholars cite 264
works of economic philosophers Thomas Hobbes and Adam Smith as the earliest foundations of the 265
rational behavior theory (Monroe, 2001). For the rational actor - who is sometimes also referred to 266
as the “economic man” - rationality first and foremost means maximization of personal utility 267
(Simon, 1978). Selfish maximization implies that a rational actor always chooses the behavioral 268
option that results in highest personal utility gain (Frank, 1987). She is not only self-interested and 269
maximizing, but also goal-oriented; she has agency, stable and organized preferences, advanced and 270
even perfect knowledge of the environment and unlimited cognitive abilities for calculating 271
outcomes of all possible behavioral options (Monroe, 2001). 272
The principle of rationality as self-interested utility maximization and the rational actor model itself 273
have been frequently applied in economics and political science, as well as in psychology, 274
international relations and other social sciences. As stated in the introduction to the paper, the 275
rational actor has been and still is used widely for modelling human decision-making in natural 276
resource use - especially when dealing with the so called “tragedy of the commons” or guiding 277
environmental policy (Jager et al., 2000; van den Bergh et al., 2000). 278
When applying our framework to the rational actor model, the state reflects self-interested needs of 279
the homo economicus, a certain utility function, clear and stable values (or preferences), an 280
individual skillset and a complete knowledge about the social-ecological system. The agent is aware 281
13
of all behavioral options available to her and thus the “perceived” behavioral options include all 282
possible options. The options are nevertheless still restricted by the agent’s skillset. We identify 283
optimization as the selection process of the homo economicus model. When optimizing, a selfish 284
rational agent selects the behavioral option that maximizes her own utility, given the possible costs 285
and constraints. Since the rational agent is “all-knowing”, with unrestricted cognitive capacity, she is 286
always able to calculate and select the optimal option. Her actions then affect the social-ecological 287
system. If the rational actor considers a long-term time horizon (and therefore has perfect 288
knowledge about outcomes that occur), the feedback from the system in this time step does not 289
carry any information that is new for this actor. However, in some cases, a rational actor can be 290
modelled with complete information only about outcomes of current actions and therefore would 291
need to perceive additional knowledge from the system in order to decide on behavior during the 292
next time step. The perception of the rational actor is not restricted by cognitive capacity, and she is 293
fully aware of her impacts on the system and payoffs resulting from her actions. 294
4.2. Bounded rationality 295
296
Figure 3: Bounded rationality mapped to the framework. 297
The theory of bounded rationality was proposed by Simon (1957) who argued that the model of 298
economic man does little in terms of describing how actors behave rationally under real-world 299
constraints such as uncertainty and limitations in the capacity of human mind (van den Bergh, 2000; 300
Monroe, 2001). Simon’s model retains some key assumption of rationality (goal-oriented, conscious 301
self-interest), however suggests introducing constraints on information-processing ability of a 302
14
rational actor (Simon, 1972; Monroe, 2001). Specifically, the actor is “bounded” by her own 303
cognitive capacity as well as the environment and has incomplete or uncertain information about 304
the world. Finally, the goal of the boundedly rational actor is not limited to utility maximization. 305
According to Simon (1957) the actor may choose a course of action that is “good enough” instead of 306
continuing to search for the best one. In this case the actor is still being rational, but engages in 307
satisficing, rather than maximizing behavior. Apart from maximizing and satisficing, a boundedly 308
rational actor may choose behavior according to heuristics - simple ‘rules of thumb’ that can be 309
adapted to the environment and do not involve calculation of possible outcomes (Gigerenzer and 310
Selten, 2001). 311
Models of boundedly rational behavior have found use in a broad variety of disciplines, starting from 312
economics - to address economic decision-making of individuals and organizational behavior; for 313
studying political behavior, voting decisions and linking individual behavior to macro-politics in 314
political science. The frequent use of boundedly rational actors in agent-based modelling of social-315
ecological systems is not explicitly mentioned since the theory does not specify the many different 316
ways to implement boundedly rational actors. 317
Bounded rationality can be represented within the framework as a modification of the rational actor 318
model. As mentioned above, apart from utility maximization, a boundedly rational actor’s goal may 319
be getting just enough to be satisfied. Most importantly, it has incomplete or uncertain knowledge 320
about possible behavioral options and their outcomes as well as system processes (e.g., rate of 321
resource renewal). When considering possible behavioral options, the boundedly rational actor may 322
look for the optimal one (albeit using imperfect information and considering search costs), select the 323
one that potentially fits a certain satisfaction criteria or apply known heuristics. Information the 324
actor perceives from the system may be incomplete (e.g., due to cognitive, physical or financial 325
constraints). When evaluating new information, a boundedly rational agent may adjust preferences, 326
gain new knowledge or update and learn heuristics. 327
15
4.3. Theory of planned behavior 328
329
Figure 4: Theory of planned behavior mapped to the framework. 330
The theory of planned behavior (Ajzen 1991) assumes that behavior is mediated by intentions and 331
perceived behavioral control. Intentions are based on three beliefs: behavioral beliefs (attitudes), 332
normative beliefs (subjective norm), and control beliefs (perceived behavioral control). Attitudes are 333
aggregated beliefs about the strength of the effect of the behavior (e.g., how important resting a 334
pasture is for nature conservation) and their normative value (e.g., whether conserving nature is bad 335
or good). Subjective norms are aggregates of the beliefs of approval/disapproval of the behavior by 336
important individuals or groups and the motivation to comply with important others; perceived 337
behavioral controls are aggregates of the beliefs about a control factor (e.g., money) and the 338
perceived power of the control factor (e.g., is money important). 339
The theory of planned behavior has been used extensively in different contexts, such in 340
environmental psychology, among others, to explain pro-environmental behavior or lack thereof. It 341
is mostly applied in empirical studies to measure the three beliefs and to relate them to intentions 342
as well as to behavior. However, recently there are some modelling studies that use the theory to 343
study the adoption of new practices or technologies (Schwarz & Ernst, 2008; Kiesing et al. 2012). 344
When mapping the theory on our framework, attitudes, subjective norms and perceived behavioral 345
control are placed in the state. Attitudes relate to the knowledge (belief of the effect and its 346
strength) and values or needs (evaluation, e.g., good or bad). The subjective norm relates to the 347
knowledge of the agent (beliefs about the approval or disapproval of the respective behavior by 348
16
others and their importance for the agent). The control factors correspond to the assets, while the 349
perceived power of a control factor is part of the knowledge. Attitudes, subjective norms and 350
perceived behavioral control act as a filter that determines the intentions for the different perceived 351
behavioral options. The selection process is a function (which needs to be specified by the modeler) 352
that links behavioral options to performed behavior, mediated by the perceived behavioral control. 353
Options that have a higher intention and that are perceived as being in the control of the actor are 354
more likely to get executed by the agent. 355
4.4. Habitual/reinforcement learning 356
357
Figure 5: Habitual Decision-making/Reinforcement learning (psychology) mapped to the framework. 358
Habit - "is a behavior we do often, almost without thinking" (Graybiel, 2008, p. 359). For example, 359
always eating potatoes, meat and vegetables, not considering a pasta or stir fry. Habits are "learned, 360
repetitive, sequential, context-triggered behaviors" (Graybiel, 2008, p. 363). They perform the best 361
in relatively stable situations, and save on cognitive effort (e.g., Jager, 2003). Reinforcement learning 362
is an approach to describe/represent habitual behavior. It addresses a relevant principle in 363
behavioral learning that originates in the classical (Pavlov, 1927) and operant (Skinner, 1953) 364
conditioning theories. These behaviorist approaches indicate that if an organism receives a reward 365
after performing new behavior, the chances increase that this behavior will be repeated. Theories on 366
habits/reinforcement learning are - in contrast to many other theories - explicitly incorporating 367
feedback, i.e., the positive outcomes of performed behavior result in reinforcement of this behavior 368
in a next time-step. This cyclicity is an essential component of habitual/reinforcement learning. 369
17
When mapping the key concepts of reinforcement learning onto the framework the importance of 370
cyclical processes of action and rewards is reflected in the focus on the selection (of action/behavior) 371
and evaluation (reward) processes. The theory includes two selection processes whose choice is 372
influenced by the state of the agent, namely the satisfaction of its needs. If a person's state is 373
satisfactory, the selection of behavior will be automatic, and relies on a script that defines the 374
behavior to perform under this condition (Abelson, 1981). The script is part of the agent’s 375
knowledge. A script can address a single behavior-outcome link (e.g., irrigating when the soil is dry), 376
but also more complicated structures of behavior-outcome relations are possible (e.g., a farmer may 377
have a repertoire of habitual responses to different conditions of weather, season and state of the 378
crop). The performance of a behavior is evaluated concerning the satisfaction of different needs 379
(evaluation). When agents repeatedly go through the cycle from perception and evaluation to 380
behavior, choices that satisfy are reinforced. The more often new behavior is being performed, and 381
the more consistent the person experiences reinforcement, the stronger the script gets, and the 382
more stable/strong the new habitual behavior becomes. 383
When no reinforcement follows after performing the habit, after some time the need satisfaction 384
will drop below a certain critical level (e.g., linked to the person’s ambition level), causing the person 385
to switch towards deliberating about alternative behaviors. For example, one or a few times fishing 386
with a very disappointing catch is not likely to change the behavior of a fisher. However, after a few 387
instances the fisher will start thinking about possible causes and alternative behaviors. This 388
deliberation will also include the longer-term outcomes and goals, which are not being considered 389
while in automatic mode. Note that the habitual/reinforcement learning does not specify the 390
deliberative processes when a habit is being reconsidered. Here different theories can be applied 391
depending on the context of deliberation, such as the other theories addressed in this paper (i.e.,: 392
the rational actor, prospect theory, theory of planned behavior or descriptive norms). 393
394
18
4.5. Descriptive norm 395
396
Figure 6: Descriptive norm mapped to the framework. 397
Within different social science disciplines social norms are studied as a key element affecting 398
decision-making (Kallgren et al. 2000; Berkowitz, 1972; Fishbein & Ajzen, 1975; Kerr, 1995; Staub, 399
1972; Triandis, 1977; Bandura 1977; Borsari and Carey, 2003; Cialdini et al. 1990). A common 400
distinction is descriptive versus injunctive norms (Cialdini et al., 1990). Descriptive norms refer to the 401
influence of perceiving what other people actually do, while injunctive norms refer to what one’s 402
perception is about socially acceptable behavior. Observing the behavior of what other people do 403
may, under certain circumstances, have an impact on a person’s behavior. This observation may take 404
place in an almost subconscious manner, where the observed behavior becomes more salient for 405
selection, or this observation may be more deliberately processed, as where other people’s behavior 406
serves as a cue in deciding the proper action to take in a particular situation. 407
Experimental research has demonstrated that descriptive norms influence environmentally related 408
behavior, such as littering (Cialdini, 2003), reuse of towels in hotels (Goldstein et al. 2008), voting 409
behavior (Gerber & Rogers, 2009) and energy use (Schultz et al. 2007). Except for the work of Feola 410
and Binder (2010) the authors do not know of the explicit use of descriptive norms in models of 411
social-ecological systems. 412
When placing the descriptive norm within the framework we can directly relate it to perception and 413
behavior options and observe how it affects behavior through the decision-making process by 414
making the dominant behavior of others more salient. Perception is the process of observing the 415
19
actions of others, either by direct information or by receiving information on other people’s 416
behavior. People learn the descriptive norm based on repeated observations (Kashima et al, 2013). 417
During the evaluation, the observations of the behaviors of others are evaluated and the dominant 418
behavior of others is determined. This makes it more likely that this dominant behavior is selected. 419
Values within the state component are important in understanding how a person may react to a 420
descriptive norm. Here a distinction can be made between conformism, non-conformism and anti-421
conformism (see e.g. Levine & Hogg, 2009). A non-conformist is not much affected by the observed 422
behavior of other people in contrast to the conformist and anti-conformist. Here the conformist is 423
likely to comply to the group behavior, and the anti-conformist wants to deviate from the group 424
behavior. The perceived dominant behavior of others influences the saliency (activation) of the 425
perceived behavioral options. Also, following the majority behavior can result in a higher social 426
need satisfaction in the state for conformists, and a lower social need satisfaction for anti-427
conformists. The dominant behavior becomes more salient for those who are more socially 428
susceptible, and is therefore more likely to be selected by conformists, and less likely by anti-429
conformists. 430
431
4.6. Prospect theory 432
433
Figure 7: Prospect theory mapped to the framework 434
Prospect theory introduces important aspects from cognitive psychology to the rational actor model, 435
particularly with respect to how people make decisions between alternative options that involve 436
20
probabilistic events. The central assumption in prospect theory is that people bias a rational 437
decision because the context (social or physical setting of a decision situation) shapes their aversion 438
to risk (Kahneman and Tversky, 2000). Prospect theory generally assumes a degree of risk aversion, 439
whereby actors bias decisions toward avoiding loss over chancing a gain (Hastie and Dawes, 2001). 440
When the stakes are small however, actors tend to “gamble” and seek more risk (Lefebvre et al., 441
2010). 442
The theory has been used to understand human decisions in multiple arenas, such as international 443
relations (Goldgeier and Tetlock, 2001; Levy, 1992), financial risk management (Fiegenbaum, 1990), 444
and insurance markets (Sydnor, 2010). Recently the theory has been applied to problems related to 445
NRM in variable environments (Rajagopalan et al., 2009). 446
When mapping Prospect theory onto the framework, much emphasis is put on the evaluation 447
process as this is where the probability of events is evaluated as well as the values of an agent such 448
as its risk attitude. In the context of prospect theory people evaluate possible future outcomes 449
differently based on the probability of the events. Each of these events occurs with a certain 450
probability that may, or may not, be perceived by the decision-maker. Commonly, people 451
underestimate chances of rare events eventuating, for example extreme storms that have a 452
probability of occurring once every 100 years. In contrast, people can overestimate the implication 453
of such rare events. An example is the perceived risks from Ebola virus vs. influenza in the USA. 454
While an Ebola case occurs with much less probability people tend to weigh the risks more heavily 455
and this may influence the expected value of the possible actions. 456
Furthermore people’s values play a major role in defining the perceived behavioral options. Agents’ 457
attitude toward risky behavior will weigh gains and losses differently, but just how will depend both 458
on the person, and how large the potential loses or gains are. What defines a loss vs. a gain is a 459
threshold, or more precisely a reference point that is a reflection of people's expectations or beliefs 460
about past outcomes. It is important to note that the function that defines the risk attitude can 461
represent different types of behavior, such as risk aversion, relentless or indifference to the 462
perceived risk (Wakker, 2010). 463
While a range of factors influence how an individual weighs loses and gains, in a modelling context, 464
once the total value function for each behavioral option can be defined, the selection of behavior is 465
thereafter based on selecting the optimal return, and hence prospect theory can be operationalized 466
as a specialized version of the rational actor model. 467
4.7. Summary of mapping of all theories on framework 468
Table 1 summarizes the main attributes of the different theories as identified from the mapping 469
process. 470
471
472
21
Table 1: Summary of the mapping of the theories 473
Theory Perception Evaluation Selection State Perceived behavioral options
Rational actor all information needed for decision available
not specified optimization self-interest/utility/ preferences
all are known
Bounded rationality
constrained by cognitive capability or bio-physical reality
can include learning
optimization satisficing heuristics
preferences limited knowledge about available options
Prospect Theory
probabilities of events
weighting of different possible outcomes
optimization risk attitude reference point
filtered by risk perception and attitude
Theory of Planned Behavior
social network approval/ disapproval of behavior by others
intention with highest value becomes most important
beliefs, strength of beliefs, others’ importance, resources, opportunities
intentions
Descriptive Norm
behavior of others
dominant behavior
dominant behavior becomes more salient
social need social susceptibility
behaviors and dominant behaviors of others
Habitual/ Reinforcement Learning
not specified need satisfaction
automatic or deliberate
needs scripts
scripts
474
475
5. Discussion 476
Managing natural resources is managing people. Despite the widespread recognition of the 477
importance of complex human behavior for sustainable resource management, many formal models 478
are still based on the over-simplifying assumptions of the rational actor. This is particularly critical 479
when models are used for real world policy support where humans may behave very differently than 480
assumed in these models. Empirical and experimental research in psychology, behavioral economics, 481
sociology, or anthropology on the other hand has developed many alternative theories of human 482
behavior. We argue that this gap between our representations of human behavior in formal models 483
and the rich evidence of the nuances of human decision-making in the real world is to some extent 484
due to the plethora of theories developed scattered within and among the different disciplines 485
(some of which are generic while others are very specific), the different terminologies used in 486
different fields, and the lack of understanding of causal mechanisms that explain how factors 487
interact over time in determining decisions and behaviors. When implementing a behavioral theory 488
in a formal model a modeler thus faces the challenge to identify a relevant theory, potentially 489
dealing with incompleteness in the representation of the decision-making process and the lack of 490
specification of causal relationships. 491
In this paper, we proposed a framework for operationalizing theories of human decision-making in 492
natural resource management contexts to tackle some of these challenges. The framework supports 493
22
mapping, describing, organizing and comparing different behavioral theories as a first step towards 494
choosing and implementing them in formal models of social-ecological systems. Each theory uses a 495
vocabulary that is specific to the discipline it belongs to. Our framework provides a generic language 496
that can be used to describe multiple theories using the seven elements of the framework – social 497
and biophysical context, perception, evaluation, state, perceived behavioral options, selection 498
process, and behavior. This facilitates mapping and comparison of different theories with respect to 499
their structure and dynamics and their implications for understanding natural resource 500
management. Furthermore it can be used to communicate how a theory was formalized to 501
implement it into an agent-based model. We do not claim that the framework can capture all 502
theories of human decision-making in a natural resource use context, but we hope that, as a first 503
step, it will be useful to capture many relevant theories and in the long run supports a broader 504
inclusion of these theories in social-ecological models. 505
5.1. Mapping theories to clarify their focus and underlying assumptions 506
Our experience with the mapping of the six selected theories showed that it was possible - with a bit 507
of flexibility in the interpretation of the elements - to map each theory onto the elements. The 508
comprehensive process of specifying the elements of a behavioral theory facilitates communication 509
about the assumptions of the theory and can help clarify the meaning of the diversity of concepts 510
used in different theories. It also makes explicit the focus of a theory, e.g. which of the elements are 511
considered important determinants of behavior (Table 2 – “Focus”). The mapping highlights which 512
ones are represented and which ones not, how they are represented and how they may interact 513
over time to determine the behavior of the actor. As such it can facilitate the identification of 514
commonalities and differences between theories. For instance, a theory may focus on the role of 515
actor characteristics (state), the options an actor can choose from (perceived behavioral options) or 516
the way she makes her choices (selection process). For instance, our mapping of the rational actor 517
clearly shows that, in this theory, actors maximize their needs (e.g. expected utility) by exploring all 518
available options. In contrast, in the descriptive norm theory, decisions are almost exclusively 519
influenced by the perception of the behavior of others in the social environment. In some theories 520
the needs and goals of the actor are driving a decision (e.g., bounded rationality, prospect theory) 521
while others don’t specify any needs and goals (e.g. theory of planned behavior, descriptive norm). 522
The theory of planned behavior is very specific on elements of the state of an agent such as its 523
attitudes, subjective norm and perceived behavioral control, even specifying the components that 524
determine its value. Yet it remains very unspecific on the actual selection process that selects among 525
different intentions. 526
The mapping also reveals where a theory lacks detail or relevant information on particular aspects of 527
the decision -making processes needed to implement it in a formal model (Table 2 – “degree of 528
completeness”). In the rational actor model, for instance, changes of behavior over time are not 529
made explicit, i.e., the informational feedback from the social and biophysical environment is not 530
specified. In the theory of planned behavior all components of attitudes, subjective norms and 531
perceived behavioral control could possibly change over time if aspects of that are perceived and 532
evaluated dynamically. However, the theory itself does not state so. In the boundedly rational actor 533
the selection process is not specified, rather several processes may be used. The 534
habitual/reinforcement learning theory does not describe how an agent decides deliberately when it 535
explores a new behavior. Gaps in a theory may however open up opportunities to link it with 536
23
another theory that specifies the missing process. For instance one could combine the habitual 537
model with the descriptive norms model to represent deliberate decision-making such that an agent 538
follows the norm when it explores and carries out its habitual behavior when it is in an automated 539
mode. The framework can thus facilitate a process of integrating different theoretical perspectives. 540
5.2. Limitations of the framework 541
The mapping of the theories, however, also revealed limitations of the framework or difficulties in its 542
application. It does not capture well the processes that go beyond a single time step such as 543
learning. This becomes particularly problematic when the decision of an agent depends on the 544
decision- making process in the previous time step as is the case in the habitual model. Here the 545
decision of an agent to explore new behavior can only happen when it did not receive positive 546
reinforcement after repeatedly performing the habit. Hence processes related to learning and 547
habitual behavior can only be understood in terms of repeatedly running through the framework 548
cycle. The dynamics of behavior and behavioral change are a result of iterations of the decision- 549
making processes over time (i.e., running the model). For example, learning emerges as a change in 550
selected variables of the state such as knowledge or values or as a change in the behavioral options 551
resulting from new information or the observation of the behaviors of others (social learning). 552
For some theories it was difficult to map specific terms of a theory onto the framework vocabulary. 553
In general, flexibility is required to relate a theory- specific concept to the concepts of the 554
framework. The term “preferences” for instance could be mapped as needs and goals or values 555
depending on the theory or application context. The concept “intentions” from the theory of 556
planned behavior does not have a direct equivalent in the framework but can be mapped onto the 557
behavioral options. It also does not specify whether the intention with the highest values is always 558
executed. When mapping and later implementing a theory in a formal model the modeler has to 559
make many assumptions on causal relationships, particularly when the theory is based on 560
correlations as in the case of the theory of planned behavior. Finally, as the framework focusses on 561
individual decision-making it cannot deal with processes or theories that include group or collective 562
decision-making. 563
564
24
5.3. Operationalizing behavioral theories in SES models 565
A comparison of the six theories analyzed here revealed their strengths and weaknesses with 566
respect to the elements of decision-making that are well represented and those that are missing. 567
Table 2 summarizes an assessment of the different theories with respect to additional criteria that 568
may be relevant for the choice of one or several theories to be implemented in a social-ecological 569
model such as the degree of formalization, degree of completeness, the representation of dynamics. 570
One can see, for instance, that only one theory explicitly includes details on the change of decision-571
making over time (dynamics) and that most theories except for the rational and the boundedly 572
rational actor have so far not been used much in formal models. 573
When choosing a theory for implementation in a formal model it is important that the selected 574
theory represents those aspects of human decision-making in detail that are considered crucial for 575
the particular context or research focus of the model. If for instance for a given context cooperation 576
between users of a natural resource is critical, then theories focusing on interactions with others will 577
automatically be relevant for the modeler. A choice may also be influenced by the degree of 578
formalization of a theory since as a theory that is already highly formalized will be easier to 579
implement into a model and less assumptions on missing elements will have to be made. Finally the 580
degree of completeness of a theory can also be important since as a theory that already covers many 581
aspects means that modelers do have to make less assumptions than for theories that focus on 582
single elements and leave out others. The framework can help identify which aspects a theory 583
focusses on (Table 2) and which details it includes. 584
When operationalizing a theory for use in a social-ecological model, one has to make choices on 585
causal mechanisms. In the case of the theory of planned behavior, for instance, one has to decide 586
how the behavioral, normative and control beliefs are aggregated to determine intentions and how 587
an intention leads to behavior. These choices are necessarily more or less well-founded assumptions 588
that will vary between different model implementations. The framework can thus be very valuable in 589
encouraging modelers to make these assumptions explicit and their effects comparable between 590
different models. It also stimulates a critical reflection on whether the degree to which one has to 591
make assumptions on causality limits the usefulness of a particular theory for SES modelling. Such a 592
process can thus help identify theories that are most relevant. 593
Furthermore the structure of the framework makes modelers aware of the different elements of 594
decision-making that need to be tackled in any kind of implementation in a formal model. By going 595
through the framework in a systematic manner a modeler will have to think about the relevance of 596
each element, even if a theory does not specify it. For instance, a theory may not be explicit about 597
what information an agent can perceive from the biophysical and social environment. For example, 598
the implementation of the rational actor model shows that one needs to specify a feedback from the 599
biophysical environment if the agent makes decisions over time. The ideal model of the theory 600
cannot be implemented (next to that the agent will always be a boundedly rational actor). The 601
mapping will alert the modeler to the fact that she needs to investigate whether this is an important 602
aspect for the particular decision and decision context to be modelled. This process can be 603
complementary to the use of the ODD+D protocol (Müller et al. 2013) in guiding the development of 604
the decision model. The framework however goes beyond ODD+D in that it structures the decision- 605
making process in more detail into specific elements and specifies how they interact. This enables a 606
more systematic and comparable development of a decision- making model. 607
25
Table 2: Assessment of the different theories with respect to their use in formal modelling of SES. 608
Theory Focus (element(s) of the framework that is/are most important)
Degree of formalization
Degree of completeness (are all elements described)
Representation of dynamics
Frequency of use in models of natural resource management
Rational actor Goal-oriented needs
optimization (selection)
High (equations) Medium No High
Bounded rationality
Imperfect perception and selection
Medium (no equations given, but clearly described)
High No High
Prospect Theory (E)valuation of
the probability of different events
High (equations) Medium No Medium
Theory of Planned Behavior
Elements of the state that
determine intentions
Medium High No Low
Descriptive Norms
Perception of
the behavior of others and evaluation of
dominant one
Low High No Low
Habitual/Reinforcement Learning
Two different types of selection processes
(automatic versus deliberative)
Low Medium Yes Low
609
5.4. Final remarks 610
There is an increasing recognition of the importance of including a broader knowledge base of our 611
understanding of human behavior in the study of social-ecological systems. The inclusion of this 612
knowledge in formal models faces major challenges due to the required specifications of 613
assumptions that are not addressed in the original theories. Our proposed framework is a modest 614
step to facilitate this translation process. The next step for this research endeavor will be the 615
systematic implementation of a set of behavioral theories in models of social-ecological systems. 616
Implementing different behavioral theories into a formal model allows for a sensitivity analysis of 617
human behavior within a natural resource management context. We see the value of this sensitivity 618
analysis initially as a conceptual understanding of how different behavioral theories affect the 619
dynamics of social-ecological systems. Moreover, including different behavioral theories on decision-620
making in formal models of social-ecological systems enables the ability to assess the consequences 621
of a mismatch in behavioral theories for designing policies. For example, one may optimize a tax 622
policy to meet some environmental management goals assuming rational choice of selfish actors. 623
Implementing alternative behavioral theories, we can test the consequences of this policy if actors 624
26
are not rational and selfish, but make decisions according to other behavioral theories. Such an 625
analysis enables assessing the robustness of the performance of policy options to different 626
assumptions of human behavior. 627
628
6. Acknowledgements 629
We acknowledge the financial support from the National Socio-Environmental Synthesis Center in 630
Annapolis, US (SESYNC), the Helmholtz Centre for Environmental Research (UFZ) in Leipzig, 631
Germany, and German Centre for Integrative Biodiversity Research (iDiv), Leipzig for meetings of 632
our working group. MS acknowledges funding by the European Research Council under the 633
European Union’s Seventh Framework Programme (FP/2007-2013)/ERC grant agreement no. 283950 634
SES-LINK and a core grant to the Stockholm Resilience Centre by Mistra. BM and GD were supported 635
by the German Federal Ministry of Education and Research (BMBF—01LN1315A) within the Junior 636
Research Group POLISES. We also acknowledge the valuable feedback of the audience on 637
presentations on this paper at UFZ and iDiv, and the students from the summer school class July 638
2015 on “How to model human decision-making in social-ecological agent-based models” in Kohren-639
Sahlis, Germany. 640
641
7. References 642
Abelson, R.P. (1981). Psychological status of the script concept. American Psychologist, 36, 715-729. 643
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision 644
Processes, 50, 179–211. 645
An, L. (2012). Modelling human decisions in coupled human and natural systems: Review of agent-646
based models., Ecological Modelling, 229, 25-36. 647
Anderson, J.R. (1990). Cognitive psychology and its implications (3rd ed.). A series of books in 648
psychology. New York, NY, US: W H Freeman/Times Books/ Henry Holt & Co.. xvi 519 pp. 649
Balke, T. & Gilbert , N. (2014). How do agents make decisions? A Survey. Journal of Artificial Societies 650
and Social Simulation, 17 (4), 13. 651
Bandura, A. (1977). Social learning theory. Englewood Cliffs, N.J.: Prentice-Hall. 652
Berkes, F., Folke, C. (Eds.) (1998). Linking Social and Ecological Systems: Management Practices and 653
Social Mechanisms for Building Resilience. Cambridge: Cambridge University Press. 654
Berkowitz, L. (1972). Social norms, feelings, and other factors affecting helping and altruism. In L. 655
Berkowitz (Ed.), Advances in experimental social psychology (Vol. 6, 63-108). San Diego, 656
CA: Academic Press. 657
Borsari, B., Carey, K.B. (2003). Descriptive and injunctive norms in college drinking: A meta-analytic 658
integration. Journal of Studies on Alcohol, 64(3), 331-341. 659
Bousquet, F., Le Page C. (2004). Multi-agent simulations and ecosystems management: a review. 660
Ecological Modelling, 176, 313-332. 661
27
Carley, K., Newell, A. (1994). The Nature of the social agent. Journal of Mathematical Sociology, 662
19(4), 221-262. 663
Cialdini, R.B., Kallgren, C.A. & Reno, R.R. (1990). A focus theory of normative conduct: A theoretical 664
refinement and re-evaluation of the role of norms in human behaviors. Journal of Personality and 665
Social Psychology, 58, 1015–1026. 666
Cialdini, R. B., Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annual Review of 667
Psychology, 55 (1), 591–621. 668
Cialdini, R.B. (2003). Crafting normative messages to protect the environment. Current Directions in 669
Psychological Science, 12(4), 105-109. 670
Clark, C.W. (1976) Mathematical Bioeconomics: The optimal management of renewable resources. 671
New York, NY: John Wiley & Sons. 672
Cooke, I.R., Queenborough, S.A., Mattison, E.H.A., Bailey, A.P., Sandars, D.L., Graves, A.R., Morris, J., 673
Atkinson, P.W., Trawick, P., Freckleton, R.P., Watkinson, A.R. & Sutherland, W.J. (2009). 674
Integrating socio-economics and ecology: A taxonomy of quantitative methods and a review of 675
their use in agro-ecology., Journal of Applied Ecology, 46, 269-277. 676
Eisenegger, C., Haushofer, J. & Fehr, E. (2011). The role of testosterone in social interaction. Trends 677
in Cognitive Science, 15, 263–271 678
Fehr, E. & Gintis, H. (2007). Human motivation and social cooperation: Experimental and analytical 679
foundations. Annual Review of Sociology, 33, 43-64. 680
Fehr, E. & Schmidt, K.M. (1999) A theory of fairness, competition, and cooperation. Quarterly Journal 681
of Economics, 114, 817-868. 682
Feola, G. & Binder, C. R. (2010). Towards an improved understanding of farmers' behavior: The 683
integrative agent-centered (IAC) framework. Ecological Economics, 69(12), 2323-2333. 684
Fiegenbaum, A. (1990). Prospect theory and the risk-return association: An empirical examination in 685
85 industries. Journal of Economic Behavior & Organization, 14(2), 187-203. 686
Fishbein, M. & Ajzen, I. (1975). Belief, Attitude, Intention, and Behavior: An Introduction to Theory 687
and Research. Reading, MA: Addison-Wesley. 688
Frank, R.H. (1987). If Homo-Economicus could choose his own utility function, would he want one 689
with a conscience. American Economic Review, 77, 593-604. 690
Fulton, E.A., Smith, A. D. M., Smith, D. C., & van Putten, I.E. (2011). Human Behavior: The Key Source 691
of Uncertainty in Fisheries Management. Fish and Fisheries, 12, 2–17. 692
Gerber, A.S., Rogers, T. (2009). Descriptive Social Norms and Motivation to Vote: Everybody’s Voting 693
and so Should You. The Journal of Politics, 71(1), 178–191. 694
Gigerenzer G. & Selten R. (Eds.) (2001) Bounded Rationality: The Adaptive Toolbox. Cambridge, 695
Massachusetts, USA: MIT Press. 696
Gintis, H. (2000). Strong reciprocity and human sociality. Journal of Theoretical Biology, 206(2), 169-697
179. 698
28
Goldgeier, J. M., & Tetlock, P. E. (2001). Psychology and international relations theory. Annual 699
Review of Political Science, 4(1), 67-92. 700
Goldstein, N.J., R.B. Cialdini, V. Griskevicius. (2008). A room with a viewpoint: Using social norms to 701
motivate environmental conservation in hotels. Journal of Consumer Research, 35(3), 472-482. 702
Graybiel, A.M. (2008). Habits, rituals, and the evaluative brain. Annual Review of Neuroscience, 31, 703
359–387. 704
Hardisty, D., Orlove,B., Krantz, D.H., Small, A.A., Milch, K.F., & Osgood, D.E. (2012). About time: An 705
integrative approach to effective environmental policy. Global Environmental Change, 22 (2012) 706
684–694. 707
Hare, M. & Deadman, P., (2004). Further towards a taxonomy of agent-based simulation models in 708
environmental management. Mathematics and Computers in Simulation, 64, 25–40. 709
Hastie, R. & Dawes, R. M. (2001). Rational choice in an uncertain world : The psychology of judgment 710
and decision making. Thousand Oaks, CA: Sage. 711
Heckbert, S., Baynes, T. & Reeson, A. (2010). Agent-based modelling in ecological economics. Annals 712
of the New York Academy of Sciences, 1185, 39-63. 713
Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., & McElrath, R. (2001). Cooperation, 714
reciprocity and punishment in fifteen small-scale societies. American Economic Review, 91, 73- – 715
78. 716
Hopfensitz, A. & Reuben, E. (2009). The importance of emotions for the effectiveness of social 717
punishment. Economic Journal, 119, 1534-1559. 718
Jager, W. (2003). Breaking “’bad habits”: A dynamical perspective on habit formation and change. in: 719
Hendrick, L., Jager, W. & Steg, L. (Eds.) Human decision making and environmental perception: 720
Understanding and assisting human decision making in real-life settings. Liber Amicorum for 721
Charles Vlek. Groningen: University of Groningen. 722
Jager, W., Janssen, M.A., De Vries, H.J.M., De Greef, J., Vlek, C.A.J., (2000). Behavior in commons 723
dilemmas: homo economicus and homo psychologicus in an ecological–economic model. 724
Ecological Economics, 35 (3), 357–379. 725
Janssen, M.A. & Jager, W. (2000). The human actor in ecological-economic models. Ecological 726
Economics, 35(3), 307-310. 727
Kahneman, D. (2003). Maps of bounded rationality: Psychology for behavioral economics. The 728
American Economic Review, 93 (5), 1449-1475. 729
Kahneman, D. & Tversky, A. (Eds.) (2000). Choices, values, and frames. Cambridge, MA: Cambridge 730
University Press. 731
Kallgren, C.A., Reno, R.R., Cialdini, R.B. (2000). A Focus Theory of Normative Conduct: When Norms 732
Do and Do not Affect Behavior. Personality and Social Psychology Bulletin, 26(8), 1002–1012. 733
Karagiannakos, A. (1996). Total allowable Catch (TAC) and quota management system in the 734
European Union. Marine Policy, 20(3), 235-248. 735
29
Kashima, Y., Wilson, S., Lusher, D., Pearson, L. J. & Pearson, C. (2013). The acquisition of perceived 736
descriptive norms as social category learning in social networks. Social Networks, 35(4), 711–719. 737
Kerr, N.L. (1995). Norms in social dilemmas. In Schroeder, D. A. (Ed.) Social dilemmas: Perspectives 738
on individuals and groups. Westport, CT: Praeger. 739
Kerr, N.L. & Tindale, N.L. (2004). Group performance and decision making. Annual Review of 740
Psychology, 55, 623-655. 741
Kiesling, E., M. Günther, M., Stummer, C. & Wakolbinger, L.M. (2012). Agent-based simulation of 742
innovation diffusion: a review. Central European Journal of Operations Research, 20(2), 183-230. 743
Lefebvre, M., Vieider, F. M. & Villeval, M. C. (2010). Incentive effects on risk attitude in small 744
probability prospects. Economics Letters, 109(2), 115-120. 745
Levy, J. S. (1992). Prospect theory and international relations: Theoretical applications and analytical 746
problems. Political Psychology, 283-310. 747
Levine, J. M., & Hogg, M. A. (2009). Encyclopedia of groups processes and intergroup relations: 748
“Anticonformity” (19-20). Thousand Oaks, CA: Sage. 749
Maslow, A. H. (1943). A theory of human motivation. Psychological Review, 50, 370–396. 750
Matthews, R., Gilbert, N., Roach, A., Polhill, J., Gotts, N., (2007). Agent-based land-use models: a 751
review of applications. Landscape Ecology, 22, 1447–1459. 752
Max-Neef, M. ( 1991). Human scale development: conception, application and further reflections. 753
New York, NY: The Apex Press. 754
Meadows, D.H., Meadows, G., Randers, J., and Behrens III, W.B., (1972). The limits to growth. New 755
York: Universe Books. 756
Meinshausen, M., Meinshausen, N., Hare, W., Raper, S.C.B., Frieler, R., Knutti, R., Frame, D.J. & Allen, 757
M.R.(2009). Greenhouse-gas emission targets for limiting global warming to 2 oC., Nature, 458, 758
1158-1162. 759
Meyfroidt, P. (2013). Environmental cognitions, land change, and social-ecological feedbacks: an 760
overview. Journal of Land Use Science, 8(3), 341-367. 761
Milner-Gulland, E.J. (2012). Interactions between human behaviour and ecological systems. 762
Philosophical Transactions of the Royal Society of London, Series B, 367, 270–278. 763
Moll, H., & Tomasello, M. (2007). Cooperation and human cognition: The Vygotskian intelligence 764
hypothesis. Philosophical Transactions of the Royal Society of London, Series B, Biological 765
Sciences, 362, 639–648. 766
Monroe, K.R. (2001). Paradigm Shift: From Rational Choice to Perspective. International Political 767
Science Review, 22(2), 151–172. 768
Müller, B., Bohn, F., Dreßler, G., Groeneveld, J., Klassert, C., Martin, R., et al. (2013). Describing 769
Human Decisions in Agent-Based Models - ODD + D, an Extension of the ODD Protocol. 770
Environmental Modelling & Software, 48, 37–48. 771
Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A.B. & Kent, J. (2000). Biodiversity 772
hotspots for conservation priorities. Nature, 403, 853-858. 773
30
Nordhaus, W.D. (1994). Managing the global commons: The economics of climate change. 774
Cambridge, MA. and London: MIT Press. 775
Palmer, P.I., Smith, M.J., (2014). Earth systems: Model human adaptation to climate change. Nature, 776
512, 365–366. 777
Pavlov, I.P. (1927). Conditioned reflexes. New York: Oxford University Press. 778
Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on responses to argument quantity 779
and quality: Central and peripheral routes to persuasion. Journal of Personality and Social 780
Psychology, 46, 69-81. 781
Rajagopalan, B., Laciana, C. E., Weber, E. U., Katz, R. M., & Letson, D. (2009). Decadal climate 782
variability in the Argentine Pampas: regional impacts of plausible climate scenarios on agricultural 783
systems. Climate Research, 40, 199-210. 784
Rilling, J.K., Gutman, D.A., Zeh, T.R., Pagnoni, G., Berns, G.S. & Kilts, C.D. (2002). A neural basis for 785
social cooperation. Neuron, 35(2), 395-405. 786
Rodhe, H., Sörlin, S., Snyder, P.K., Costanza, R., Svedin, U., Falkenmark, M., Karlberg, L., Corell, R.W., 787
Fabry, V.J., Hansen, J., Walker, B. D., Liverman, D., Richardson, K., Crutzen, P. & Foley, J.A. (2009). 788
A safe operating space for humanity. Nature, 461, 472-475. 789
Sawyer, R. K. (2004). Social explanation and computational simulation. Philosophical Explorations, 790
7(3), 219–231. 791
Scarlett, L., Boyd, J. & Brittain, A. (2013). Catalysts for conservation: Exploring behavioral science 792
insights for natural resource investments., Resources for the Future Report. Washington DC: 793
Resources for the Future Press. 794
Schlüter, M., Mcallister, R.R.J., Arlinghaus, R., Bunnefeld, N., Eisenack, K., Hoelker, F., Milner-795
Gulland, E.J., Mueller, B. (2012). New horizons for managing the environment: a review of 796
coupled social-ecological systems modeling. Natural resource modeling, 25, 219–272. 797
Schultz, P.W., Nolan, J.M., Cialdini, R.B., Goldstein, N.J. & Griskevicius, V. (2007). The constructive, 798
destructive and reconstructive power of social norms. Psychological Science, 18(5), 439-434. 799
Siebenhuner, B. (2000). Homo sustinens: towards a new conception of humans for the science of 800
sustainability. Ecological Economics, 32 (1), 15–25. 801
Simberloff, D. & Cox, J. (1987). Consequences and Costs of conservation corridors. Conservation 802
Biology, 1(1), 63-71. 803
Simon, H. A. (1957). Models of Man, Social and Rational: Mathematical Essays on Rational 804
Human Behavior in a Social Setting. New York: John Wiley and Sons. 805
Simon, H. A. (1972). Theories of bounded rationality. In: McGuire, C.B., Radner, R. & Arrow, K.J., 806
(Eds.) Decision and organization: A volume in honor of Jacob Marschak. North-Holland Pub. Co. 807
Simon, H.A. (1978). Rationality as Process and as Product of Thought. American Economic 808
Review, 68(2), 1–16. 809
Skinner, B.F. (1953). The behavior of organisms. New York: Appleton-Century-Crofts. 810
31
Smajgl, A., Barreteau, O. (Eds.), 2014. Empirical Agent-Based Modelling - Challenges and 811
Solutions. New York, NY: Springer. 812
Staub, E. (1972). Instigation to Goodness: The Role of Social Norms and Interpersonal Influence. 813
Journal of Social Issues, 28(3), 131–150. 814
Todd, P. M., Gigerenzer, G. & the ABC Research Group. (2012). Ecological Rationality: Intelligence in 815
the world. New York: Oxford University Press. 816
Sydnor, J. (2010). (Over) insuring modest risks. American Economic Journal: Applied Economics, 2(4), 817
177-199. 818
Triandis, H.C. (1977). Interpersonal behavior. Belmont, CA: Brooks/Cole Pub. Co. 819
Van den Bergh, J., Ferrer-i-Carbonell, A.,Munda, G., (2000). Alternative models of individual behavior 820
and implications for environmental policy. Ecological Economics 32 (1), 43–61. 821
Van Vugt, M. and Griskevicius, V. (2014). Naturally green: Harnessing Stone Age psychological biases 822
to foster environmental behavior. Social Issues and Policy Review, 8(1), 1-32. 823
Victor, D. (2015). Embed the social sciences in climate policy. Nature, 520, 27-29. 824
Wakker, P.P., (2010). Prospect theory: For risk and ambiguity. Cambridge, MA: Cambridge University 825
Press. 826
World Bank (2015). World Development Report 2015: Mind, Society and behavior. Washington, DC: 827
World Bank. 828
A framework for mapping and comparing behavioral theories
in models of social-ecological systems Maja Schlütera, Andres Baezab, Gunnar Dresslerc, Karin Frankc, Jürgen Groeneveldc, Wander Jagerd,
Marco A. Janssene, Ryan R.J. McAllisterf, Birgit Müllerc, Kirill Oracha, Nina Schwarzc, Nanda Wijermansa
a Stockholm Resilience Centre, Stockholm University, Kräftriket 2b, 10691 Stockholm, Sweden,
[email protected], [email protected], [email protected]
b National Socio-Environmental Synthesis Center (SESYNC), 1 Park Place, Suite 300, Annapolis, MD
21401, USA, [email protected]
c UFZ, Helmholtz Centre for Environmental Research e UFZ, Department of Ecological Modelling,
Permoser Str. 15, 04138 Leipzig, Germany, [email protected], [email protected],
[email protected], [email protected], [email protected]
d University College Groningen, Hoendiepskade 23-24, 9718 BG Groningen, The
Netherlands,[email protected]
e School of Sustainability, Arizona State University, PO Box 875502, AZ 85287-5502Tempe, USA,
f CSIRO PO Box 2583 Brisbane Q 4001 Australia, [email protected]
Corresponding author
Maja Schlüter
Stockholm Resilience Centre
Stockholm University
Kräftriket 2b
10691 Stockholm, Sweden