Keywords

19.1 Introduction

The idea that values are abstract drivers of behavior is not new. What is interesting about the use of values is that, at least according to Schwartz [1], there is a universal set of abstract values that can be attributed to people. Differences between people stem not from having different values, but from giving different priorities to the values. This makes it possible to use values as a starting point to compare behaviors. The downside of the value theory of Schwartz is that the values defined are very abstract and thus not directly related to behavior. Several steps are needed to translate abstract values into more concrete values and ultimately into behavioral choices. The way people concretize abstract values into concrete choices for action can also differ. Therefore, there is a need to describe this whole system in a precise and unambiguous way before it can be used for practical purposes. Some work of formally describing the relation between abstract values and actions, using values trees, has been done in [2]. In this paper we start from this logical framework and show how a quantitative framework can be designed that can be used to drive behavior of agents in social simulations.

A note should be made on the applicability of values as drivers of behavior. Not all behavior is primarily value driven. In normal life values usually play an explicit role only in larger (life changing) decisions, while smaller day-to-day behavior is governed by goals and norms. However, in many social simulations, we are exactly interested in situations where people do make life-changing decisions, such as moving houses, changing jobs, change for a more sustainable life style, etc. Thus it seems that the framework is relevant for many simulations.

In Sect. 19.2, we give some background on the value framework of Schwartz and the way we can connect these abstract values to concrete values and decisions on actions as described by Weide in [2]. In Sect. 19.3, we will discuss the framework that we propose to use to translate this theory into an implementable framework that can be used in agent-based social simulations. In Sect. 19.4, we illustrate how this framework can be used and leads to intuitive results in the domain of fishery management. Section 19.5 gives some conclusions and directions for future work.

19.2 Related Work

Schwartz et al. proposed ten basic values according to the universal needs of humans [1]. The Schwartz values are defined in the most abstract way that includes all the core values of every human all around the world.

As shown in Fig. 19.1, Schwartz value theory describes the dynamic compatibility and conflicting relation between all the value types by positioning them in a circle. Values close in the circle are more compatible than values on opposite sides. For example, pursuing Tradition and Hedonism are conflicting values, as Tradition is about restraining owns actions to conform traditions and Hedonism is about self-oriented need for pleasure. However, pursuing Tradition value is compatible with pursuing Conformity (to not violate social expectations in groups usually with close others) as both stress self-restraint and submission. In other words, the compatibility level of values in the Schwartz value circle decreases when the distance of them increases in the circle. The least distance belongs to two values next to each other, and the most distance belongs to two values that are on opposite position of each other.

Fig. 19.1
figure 1

Schwartz value circle, categorization and dynamicity of abstract personal values [1]

Full size image

Values and value systems such as from Schwartz have been used in many research efforts to explore the behavior of a complex system, studying human argumentation, managerial decisions, land-use behavior, and adaptation to climate change [3,4,5,6]. For example, the effects of individual values of a society on the general behavior of a complex system (including society, ecology, and economy) are studied in [7].

Bench-Capon et al.[8] show that promoting different values will lead to different arguments. Dechesne et al.[3] investigate how personal values (and other social phenomena) affect the behavior of people when introducing a smoking ban rule.

Mercuur [9] categorized usage of values in regard to doing an action into three main categories: pre-condition, post-condition, and deliberation navigator. In other words, values might be used as a measurement function to evaluate an action (post-condition), values can be used as a motivation to do an action (pre-condition) [10], and they can be used for justifying a decision of doing an action [2, 11].

Van der Weide [2] provides a formal model that can be used for modeling value-compliant decision-making. He shows how to form concrete values out of actions that can influence the abstract values. However, the relation between compatible and opposite values in the Schwartz value circle is not included in his model. Inspired by his formal model, we propose a framework for Schwartz value theory which not only considers translating Schwartz values into concrete values, but also the relation of values in the circle (Fig. 19.1), using values as pre-condition (filter) and justifying an action at the same time.

19.3 Framework

In this section, we propose a framework to make value-based decisions. Values can be used at different places in the deliberation cycle of the agents to select options (goals, plans, actions, etc.). If agents use goals and norms, then the values can be used to prioritize between those. Once goals are chosen and pursued, the values can be used to guide which plan is most in line with values. In this paper we focus on the motivational aspect of values which implies that they are at the basis of action selection. We start with the set of all salient actions (i.e., actions that can be taken at that moment because their pre-condition is true). If more actions are available the value tree and the current satisfaction of values are used to determine the highest priority of values. Then, the set of actions that are in line with the highest priority values will be chosen. From the resulting set of actions, one action is selected based on the current goal, the norms, and motives of the agent. In our current example, only social, economic, and ecological goals are used in this step. By performing an action, the agent updates his status. In what follows we explain the step in more detail.

In order to model the value circumplex of Schwartz, we define two sets. One of the input sets is a collection of Schwartz abstract values; V alues = {V 1, V 2, V 3, V 4, V 5, V 6, V 7, V 8, V 9, V 10}, where V 1 = Universalism, V 2 = Selfdirection, V 3 = Stimulation, V 4 = Hedonism, V 5 = Achievement, V 6 = Power, V 7 = Security, V 8 = Tradition, V 9 = Conformity, V 10 = Benevolence. We need the indexes to consider the position of each value in the Schwartz circle in the framework.

The second set is the amount that each V i ∈ V alues can get which is defined as Importance = [1, 100]. Any member V i ∈ V alues can get any value from Importance to identify its importance (which determines how often the value V i has to be satisfied).

Assume that we define a function τ : V alues → Importance in which τ(V i) gets the importance of value V i. For each V i ∈ V alues, if τ(V i) = 0, then value V i is silent and not playing a role in the system; if τ(V i) = 100, it is one of the important values if there are other values with the same importance.

To consider the relation of values, inducing the conflicting opposite values in the circle, we defined the following condition that shows how the importance of any members of V alues is related.

$$\displaystyle \begin{aligned} \mathrm{Condition\ 1:} \forall i,j \in {1..10} : 0 \leqslant |\tau(V_{\mathrm{i}}) - \tau(V_{\mathrm{j}})| \leqslant \frac{|i^\prime - j^\prime|}{10}*100, \end{aligned} $$

where:

$$\displaystyle \begin{aligned} i^\prime = &\begin{cases} i &\mbox{if } 1 \leqslant i \leqslant 5 \\ 10-i & \mbox{if } i >5 \end{cases} \qquad \qquad j^\prime = &\begin{cases} j &\mbox{if } 1 \leqslant i \leqslant 5 \\ 10-j & \mbox{if } j >5 \end{cases} \end{aligned} $$

in which 5 is the number of abstract values in one half of the Schwartz circle. Regarding symmetric distances of abstract values in the Schwartz circle, we slightly transform the formula by changing some variables.

Researchers used different version of Schwartz value system with various number abstract values. For example, Schwartz used seven abstract values to study the meaning of work in different cultures [12]. Also, it is possible to define different distance for items. In the current formula for Condition 1, we assumed the same area of each sectors. As an instance, the distance between Universalism and Tradition is the same as the distance between any other successive values in the circle. However, it is possible to change the formula to adapt different distances for items. For instance, according to this condition, it is possible to have this setting: |τ(Universalism) −τ(Selfdirection)| = 15, |τ(Power) −τ(Achievement)| = 5, and the distance of the other successive values remains 10. It would be the modelers preference to make such a decision according to their research requirements.

$$\displaystyle \begin{aligned} {\mathrm{Condition 2:}} \begin{cases} \text{if } \tau(V_j) = 0 & \tau(V_i) > 50 \\ \mbox{if } \tau(V_j) \neq 0 & 0 \leqslant \tau(V_i) + \tau(V_j) \leqslant 100 \end{cases}, {\mathrm{where }}j = (5+i)\%10. \end{aligned} $$

According to Condition 2, when value V j is not included in the model (τ(V j) = 0), the opposite value of it in the Schwartz value circle should be high enough to have effects on the behavior of the system; otherwise, it can be ignored. All conditions can happen in extreme cases as well. For instance, it is possible to have all values with the same Importance. Also, having some values that do not play a role in the system (τ(V i) = 0) is possible.

In addition to the importance, values have level of satisfaction. In other words, people need to satisfy all of their values from time to time. But, the frequency of satisfaction differs due to their personal values. Function τ(V i) shows how often value V i should be satisfied. Therefore, there is a need to consider satisfaction level in the framework as well. To model changing in needs over time, we use the water tank model that determines the priority of satisfaction requirement for each value in Sect. 19.3.1.1.

19.3.1 Value-Based Selection

19.3.1.1 Value Satisfaction

In the Schwartz theory, all humans have ten introduced values, they consider their values in their life. Their life is consistent with their values. However, it is possible that in some conditions of the life, some values are not applicable. What makes a different personality is a different importance of values. As an example, consider a CEO of a multinational and an employee of an NGO. The NGO employee will do more activities that are in line with the Universalism value, and the CEO will do more activities that satisfy Power. But, it does not mean that the NGO employee does not do any activity toward Power. The difference is the frequency and types of actions of satisfying the values. But, all the values need some level of satisfaction from time to time.

To model these dynamics, we use the water tank model represented in [13]. We consider one tank for each V i ∈ V alues. Each tank has the following base parameters: fluid level λ i where 0 ≤ λ i ≤ 100 to indicate how much the value is satisfied and threshold τ(V i) where 0 ≤ τ(V i) ≤ 100 indicating when a value gets salient. The fluid drains every time step with a fixed amount of 10 to indicate that the value satisfaction is time dependent and increases when the agent does an action which is in line with the value. To be able to model the differing priorities of values of an agent, we use the threshold and calculate the priority. Agents try to fill up the tanks with the highest priority first. Priority of values is determined by using the following equation:

$$\displaystyle \begin{aligned}\rho = -((\lambda - \tau(V_i)) / \tau(V_i)) * 100\end{aligned}$$

We use negative sign as the priority of value satisfaction has reverse relation with its filled level. Filling up the tank can be done by performing actions that satisfy the abstract value connected to the tank. The increase amount is given by (100 − τ(V i)) ∗ σ. This formula makes more important values fill up slower, thus increasing the frequency of performing actions relating to those values. It is possible to assign different values to multiplier σ for different actions. For example, buying a house usually has a larger effect on your values than buying an ice cream and thus has a larger multiplier σ for more impact.

A sample of the water tank model for an agent is shown in Fig. 19.2. Each agent has ten tanks, each with the same capacity and the same draining level. Though, the value thresholds are different for various agents.

Fig. 19.2
figure 2

Example figure of the water tank model for an agent

Full size image

19.3.1.2 Value Tree

The water tanks are used for determining which abstract value has to be satisfied. As mentioned prior, the abstract values do not directly impact the behavior of people, but rather through a series of perspectives that link the abstract values to concretized values that are directly related to behavioral choices. To make values work, we need to define more concrete values. Concrete values are easier to implement, and it is easier to track their contribution to a decision.

Several steps might be taken to translate an abstract value to its concrete values. One possible solution of formally describing the relation between abstract values and concrete values is through defining value trees [14]. The root node of the tree is an abstract value from Schwartz values. Nodes that are closest to the leaves are more concrete.

To view a simple example, one could look at Sect. 19.4 Fig. 19.4a in which the abstract values are Power, Self − direction, Universalism, and Tradition. The abstract values are the roots of the trees. As the values get more concrete, the further we go down from the root. Leaves of the trees are the most concrete values that related actions are assigned to them. By looking at the parent nodes of an action, we can determine which values it can satisfy and vice versa. Different path from each action to the root is deliberation that an agent uses to justify his action. For example, being a fisher as an action can satisfy Tradition and/or Universalism.

People generally have different perspectives which can be modeled by giving only a subset of the total value tree to individuals. For example, to satisfy the Universalism value through caring for the environment, agent A might buy an electric car because the emission of an electric car in use is less than a petrol car. Agent B might think electric cars are actually worse for the environment than petrol cars because of the chemicals used to create the batteries. He will instead go with public transport instead of his own car. This illustrates that two agents might perform different actions to satisfy the same abstract value. It can also be the case that agents perform the same action to satisfy two different abstract values. For example, playing a sport for one person can satisfy the Achievement value (trying to win), while for the other, it satisfies the Tradition value (play a game with friends as a way to be together). In other words, it is possible to assign different subsets of value trees to agents.

Some actions (and therefore their related concrete values) can be linked to more than one abstract value. Considering definitions of types of values introduced in [15], we can assign actions to abstract values for our case of interest, which is studying the behavior of a fishery village. For example, people in a fishery village might go fishing because they like nature (Universalism), they like adventure (Stimulation), they want to make money to promote their social status (Power), or they want to comply with their family traditional profession (Conformity).

Actions that are linked to compatible values might be positively interrelated. For example, actions that satisfy Benevolence might have a positive effect on satisfying Universalism as well. In contrast, if an action promotes a value, it can hardly attain the value opposite of it in the Schwartz circle.

19.3.1.3 Value-Based Filtering

Using values agents makes initial selections among the available actions to perform. We find the highest priority value that needs to be satisfied using the following formula:

$$\displaystyle \begin{aligned}\operatorname*{\mbox{arg min}}_{V_i \in \mathrm{Values}} \rho (V_i) = \{V_i|V_i \in \mathrm{Values}, \forall V_j \in \mathrm{Values}: \rho(V_j) > \rho(V_i) \}\end{aligned}$$

This formula returns the most preferred value (highest priority) in the current situation that needs to be satisfied. Then the actions promoting the highest priority value that are available are selected first. To compare the priority of each two values in order to find the highest priority, we use the following formula:

$$\displaystyle \begin{aligned}\forall \textit{}V_i \in \mathrm{Values}, V_j \in \mathrm{Values}: \rho(V_i) = \rho(V_j) \ \mathrm{ if }\ \rho(V_j)-\rho(V_i) < \delta\end{aligned}$$

Meaning that ρ(V i) and ρ(V j) differ very little. Then all the actions that promote either V j or V i and are available get chosen.

The rules and conditions provided earlier are defined for abstract values in the Schwartz value system. All the concrete values in the value trees have the same importance as their root value. Therefore, all the rules and conditions of the abstract values (roots in value trees) are applicable to their related concrete values (leaves of the value trees).

It should be noted that it is possible to have some actions that are common between different value trees. For example, an agent can satisfy Power or Universalism by choosing to be a Captain as Captain is a shared action in these value trees (Fig. 19.3). However, the agent only satisfies one of the values by choosing action Captain which depends on which deliberation he did before picking up the possible actions. For example, if the agent wants to satisfy his Universalism by doing related actions and picks being Captain, he only satisfies his Universalism value (increasing the water level of Universalism water tank) and not the Power value.

Fig. 19.3
figure 3

Value tree of getting a job

Full size image

19.3.2 Making Decisions

After filtering the actions by values, we have a list of actions that are value consistent. Any of these actions that get chosen by the agent comply with his value system. Among all the value-consistent action, the agent needs to pick an action that can be done at the moment. Therefore, other filters can be applied. These filters can be motivations, social norms, goals, plans, etc. The number of filters and how those filters filter down the value-consistent action set is the modeler choice. But, the result of this function is making decision about which actions have to be performed at the moment.

19.4 Validating Value Framework

In this chapter, we validate and discuss the proposed value framework, how values play an important role in human decision-making, and how decisions of individual people in a society change the overall behavior of the society. We use an agent-based model of a fishery village and show two scenarios with different abstract value settings. As it comes from the field of exploring personal values, the whole study and therefore proposing a framework for it is a qualitative study. In [9] , validating a qualitative model is defined as the ability of the model to replicate the relations between variables. For instance, if the Universalism value gets promoted in a society (τ(Universalism) is high), the probability of hurting the environment decreases accordingly. As described in our previous study [7], one point that we want to include in our experiments is to consider the feedback loop between society, environment, and economy. Therefore, we develop all parts and feedback between them in our simulations. The attributes and mechanics of the simulation are denoted in Table 19.1.

Table 19.1 General simulation components
Full size table

19.4.1 Abstract Values Implementation

There are three main action sets that use the value framework; these are job selection, event organizing/attending, and donation/not donating. We developed value trees for those actions and for the values Power, Self − direction, Universalism and Tradition. The job selection value tree is shown in Fig. 19.3. Here we see that some jobs are capable of satisfying many values like a mayor (Tradition, Power and Self − direction), while other jobs have only one connected value, e.g., unemployed or factory worker (both Tradition). The value increase multiplier of job picking is σ = 1.

The event trees are denoted in Fig. 19.4b and show four possible actions. Organizing an event has a value increase multiplier of σ = 2 as only a small number of agents can organize an event (the maximum of events is 1 per 11 residents). Attending an event has a lower value increase multiplier; it is σ = 0.2.

Fig. 19.4
figure 4

(a) Value tree of donation. (b) Value tree of social events

Full size image

The donation trees are shown in Fig. 19.4a; there are only two possible actions here. The value increase multiplier is σ = 0.2, which is also low since donations actions can be done every tick (which is more frequent than job picking at every 4 ticks).

19.4.2 Results

We consider four values out of ten Schwartz values: Tradition, Universalism, Self − direction, and Power. These four values have been chosen because we can show that our framework works to show the dynamic relation between values. For example, Fig. 19.5 shows the dynamic behavior of the systems in two different settings. Figure 19.5a–c, g show the system output with setting (1) when there is a high priority of Power (τ(Power) = 90, τ(Self − direction) = 50, τ(Universalism) = 10, and τ(Tradition) = 50). Figure 19.5d–f, h show the behavior of the system when Universalism is promoted (τ(Power) = 10, τ(Self − direction) = 50, τ(Universalism) = 90, and τ(Tradition) = 50). Having high priority for Universalism means that agents need to do actions that satisfy Universalism more. As shown in Fig. 19.5e, agents satisfy their Universalism through donating public benefits, and there is almost always a maximum amount of fishers and captain, since this also satisfies Universalism. As agents have low priority for Power (they do not need to satisfy Power value very often), they organize commercial events and attend free events which are enough to keep them satisfied of Power value.

Fig. 19.5
figure 5

Results of simulation with two settings: (1) p = 90, s = 50, u = 10, t = 50 and (2) p = 10, s = 50, u = 90, t = 50. (a) Human age distribution, setting (1). (b) Value motivation of donation, setting (1). (c) Value motivation of social event, setting (1). (d) Human age distribution, setting (2). (e) Value motivation of donation, setting (2). (f) Value motivation of social event, setting (2). (g) Work distribution, setting (1). (h) Work distribution, setting (2)

Full size image

As shown in Fig. 19.5h, most of the agents make money as they have job. So, a lot of them earn enough money that makes them capable of donating. Therefore, they satisfy their Universalism value by donating in public benefits, working as a fisher, a captain, or elderly caretaker. As Power importance is low (as it is the opposite value in the Schwartz value circle), the other two values need to be satisfied with the same frequency; Self − direction and Tradition. These two values can be mostly satisfied with picking relating jobs. Working in the factory inside the village and working in the company outside the village satisfy Tradition and Self − direction, respectively. That is the reason we can see a fluctuation between workers in the company outside of the village and workers in the factory inside the village. There is a balance since both have free vacancies. The company outside of the village has no limit, and the factory can have a high number of employees since there is a high amount of fish coming in, because of the maximum amount of fishers.

One of the interesting simulation results is when a society is more into the Power value. As we can see in Fig. 19.5g, the number of employees for the jobs factory worker, fisher, and work outside the company fluctuates but follow a general trend. The amount of fishers is lower since people hardly ever need to satisfy universalism so they mainly become fisher because of tradition. The difference in average amount of factory workers and workers in the company outside is caused by a lower amount of vacancies as factory worker because there is less fish being caught. People satisfy their Power value by organizing commercial events and having well-paid jobs. In this case the maximum possible number of commercial events happens all the time. So, there are more chances to attend events for villagers to satisfy their Tradition value by attending the events. Besides, people tend to keep their paid job as they can make enough money to cover their living cost. The importance of Universalism (as the opposite value of Power in the Schwartz value circle) is low, and there is no need to put more effort than donating in public benefit to satisfy this value. Therefore, people who do not have chance of finding a job inside the village will look for a job outside. This justifies the higher number of people who work in the company outside. The simulation code is accessible via GitHub [16].

19.5 Discussion and Future Work

Different factors impact human behavior such as values, social norms, and environmental and economic factors. However, introduced models to study human behavior rarely considered social, environmental, and economic factors altogether. Many factors are involved to capture human behavior including personal, social, environmental, and economic factors. Values are strongly connected to behavioral choices of people among personal factors. One of the well-known theories in personal values is the theory introduced by Schwartz, and it has been used by many researchers. Schwartz came up with ten general values by studying people all around the globe. Though, using Schwartz values necessitates interpreting the abstract values to concrete values related to the case study. To the best of our knowledge, there is no standard way of using Schwartz values and transform them from general to concrete values. As of yet, researchers used them and translated them according to their taste. We introduced a framework of personal values that can be used as a guideline for those who consider values to study, model, implement, and reuse previous efforts regarding values. Using the introduced framework, it is possible to model heterogeneous agents in terms of their personality and deliberation and consider various statuses consciously. For example, two different people can do the same action for different reasons, or they can react differently in the same (social, environmental, economic) situation. In our framework, we make a value tree for each value in Schwartz value theory. The root of the tree is a general value, and value gets more and more concrete till the leaves of the tree are the most concrete values that are directly linked to implementable actions. A possible actions set is assigned to each concrete value. The result of doing one of the actions in the action set is satisfying the assigned value.

In the framework, there is a relation between Schwartz values that play an important role in decision-making. Such a relation is used to capture the circular relation of Schwartz values. The framework contains making decisions according to personal values. To make a decision at each time and determine which value is more important, we used the water tank model. We assigned a water tank to each value which drains in each time step and fills whenever the assigned value is satisfied. Using such a model, agents need to satisfy all the values during the simulation time. By changing the amount of draining and filling and therefore changing the satisfaction frequency of values, we can capture different personalities. We illustrate the use of the framework by using it to build a normative architecture for developing a socio-ecological complex system. The normative architecture is a modular one that proposes developing flexible socio-ecological complex models. This architecture includes social, environmental, and economic factors, as well as decision-making process of agents. Therefore, it is possible to make a model both for micro- and macro-analysis depending on the decision of the modeler. Another aspect of this is that manipulating different factors is possible. A model may include any of the social, ecological, and economic factors. As an example of social factors, a model might contain personal values, social norms, motives, social practices, etc.