Using agent-based models to design social marketing campaign
Luisa Perez-Mujica(a), Roderick Duncan(b),*, Terry Bossomaier(c)
(a) School of Environmental Sciences, Charles Sturt University
(b, c) Centre for Research in Complex Systems, Charles Sturt University
(b)School of Accounting and Finance, Charles Sturt University
(b) Institute of Land, Water and Society, Charles Sturt University
*Corresponding author. Tel. 61 2 6338 4982.Email address: .
ABSTRACT
Community-based social marketing (CBSM) involves members of the community as active participants in the marketing campaign for a social good. However behaviour of community members in CBSM is not well simulated using the standard tools available to marketers. We show how agent-based models (ABMs) can be used to simulate the behavior of community members at the individual level to determine how sensitive the outcome of a CBSM campaign is to assumptions around the effectiveness of marketing within the community. We develop an ABM for wetlands managers to use to simulate the outcome of a marketing plan for promotion of environmental tourism in a wetlands area. The wetlands managers must trade off the costs of marketing and the damage done by the tourism activities with the value of ecotourism for the wetland. We find evidence from the simulations that wetlands’ ecological health is sensitive to the design of the social marketing campaign.
Keywords: community-based social marketing, agent-based model, environmental tourism, wetlands
1.Introduction
The management of environmental tourism, or ecotourism, is a primary challenge for managers of areas of high environmental value (Lim & McAleer 2005). Ecotourism has benefits for the areas - both in terms of additional fee revenues for the area’s management but also in terms of raising the public profile and increasing political support for protection of the area. However the activities of the tourists will have a negative impact on the area due to land and water use, generation of waste, noise and other effects associated with tourism. These tourism activities degrade the same environmental assets the tourists are there to enjoy (Lynn & Brown 2003). Managers of the areas then have to balance use of the area with the needs of environmental and economic sustainability to achieve the goal of “ecologically sustainable tourism management”.
Community-based social marketing (CBSM) has been put forward as a marketing approach which could be of use to site managers in designing marketing campaigns to promote tourist visitors to their sites. CBSM (McKenzie-Mohr 2011; McKenzie-Mohr, Lee, Schultz & Kotler 2012) draws from research in social psychology to identify barriers to uptake of social marketing messages and communication strategies to overcome these barriers. It is hoped that the use of community-based mechanisms could overcome the traditional finding that information-based campaigns are ineffective in changing household behaviour (McKenzie-Mohr 2011). CBSM frequently makes use of communication channels within the community targeted (“social diffusion”), such as requesting those households already engaged in the activity to contact others about the campaign (McKenzie-Mohr 2011).
The design of a marketing campaign using CBSM however is hampered by the lack of tools to model the social diffusion of the campaign messages within the community. Previous CBSM research has used surveys and regressions to analyze household uptake of programs, however these methods are of limited use in predicting the impact of interactions between individuals in a community. Communication of the marketing message within the community is essentially a “black box”.
Agent-based models (ABM) are a simulation tool built around the behaviour and interaction of individual agents. Using an ABM, marketers can explicitly model the behaviour of individuals, the interactions between individuals in a community and the social diffusion of marketing messages. This method could be of use to marketers and managers designing social marketing campaigns with community-based communication. We present the results of an ABM for an ecotourism area: the Winton Wetlands.
The Winton Wetlands is an 8,750 ha transformed wetland site, located in the Goulburn-Broken Catchment in North-East Victoria, approximately 200 km north of Melbourne, Australia. In 1970, it was transformed into an artificial irrigation reservoir, Lake Mokoan, with the construction of a dam. Due to its inefficiency as an irrigation reservoir (increasing turbidity, algal blooms, and water losses), the State Government decommissioned the dam in 2004 (Goulburn Broken Catchment Management Authority, 2012).
After the decommissioning, the state Minister of Water established the Winton Wetlands Committee of Management. This community-based organization was charged with the preparation and implementation of two projects aiming to return the Winton Wetlands to its natural state as an important wetland system (approximately 2,900 ha of Red Gum woodland were destroyed with the construction of the dam) and develop the site as a sustainable nature-based touristic wetland.
The Winton Wetlands restoration project is an ideal case for this study because it is currently in the beginning stages of restoration and development of nature-based tourism. Furthermore, sustainable development and inclusion of stakeholders is deemed as important and the community has shown significant interest in taking part in the decision-making process.
2.Agent-based Modelling and community-based social marketing
2.1Agent-based modeling
Socio-environmental systems, such as those involving nature-based tourism, are dynamic, multi-scalar systems, which include interacting environmental, social and economic aspects (Musters, de Graaf, & ter Keurs, 1998). Socio-environmental systems are also complex, whose full description is impossible, where prediction of changes is difficult and unexpected changes are likely (Gibson, 2006). To overcome these difficulties, ABM is used to determine the behavior of autonomous agents, whose interaction with other agents and the environment gives rise to the behavior of the system as a whole (Bonabeau, 2002; Scholl, 2001).
Simulation has a long and distinguished history in Science, Technology, Engineering and Mathematical disciplines. Engineers build bridges and design aircraft, largely by computer. Economistsand social scientists are increasinglyoptimisticthat we can achieve similar goals using ABM for social systems of all kinds. There has certainly been substantial progress.
The Santa Fe Institute, founded by Nobel Laureates, Murray Gell-Mann (quantum electrodynamics) and Philip Anderson (disordered systems), played a foundational role in the development of complex systems as a discipline and fostered the early development of agent-based modeling.We can distinguish three different levels of agent modelling:canonical systems; heuristic ABMs (HABM), which are heterogeneous agent models using domain knowledge and heuristics for diving agent behaviour; and data-driven ABMS (DABMS), in which quantitative studies and big data parameterize agent behaviour in an attempt to get closer to the engineering goals.
2.1.1.Canonical systems (CABM)
Complex systems are interesting because diverse and particular behavior can emerge from simple agents and interaction rules. Thus, such behavior emerges in very simple, canonical systems. Of these, the best known are cellular automata (CA). Jon Conway’s Game of Life, found on many computer systems (Berlekamp, Conway, & Guy, 1982; Gardner, 1970) produces extraordinary patterns (Wuensche, 1998), characteristics of many co-called complex cellular automata (Wolfram, 1986). Since CAs have found applications in many areas from traffic (Nagelocd & Rasmussenaf, 1994) to soil erosion (Di Gregorio, Serra, & Villani, 1999) and urban planning (Clarke, 1997). Stuart Kauffman introduced Random Boolean Networks (RBN) (1993) for modelling gene regulatory networks, but they have now spread far and wide, even reaching into management and organizational science (Rivkin, 2001).
2.1.2.Swarm and its progeny
Chris Langton, who made a seminal contribution to the understanding of CA complexity in his famous Computation at the edge of chaos paper (1990), saw the future need for agent models, which embodied multiple agent types, diverse behavior and varied interactions. He convened the SWARM workshop at Santa Fe, from which the first serious ABM package arose, driven to completion by Nelson Minar. SWARM was written in Objective C, at the time primarily on the NEXT Platform. NEXT, a Steve jobs company, disappeared but its software morphed into MAC OS X and Objective C carries on with the Apple Community. But at the time it was overshadowed by C++, which still has a larger user base. Hence it was not until SWARM was rewritten in Java as RePast, that it became really popular. Since then, a range of ABM platforms have appeared. Leigh Tesfatsion’s website (Tesfatsion, 2013), an outstanding resource for ABM, has numerous examples. Netlogo (Wilensky, 1999), the platform used in this paper, is an excellent compromise between power and user tools, and intelligibility and usability for those who are not IT professionals. These packages support HABMs and can be infused with qualitative data from interviews and surveys.
2.1.3.Generative social science
Although engineering simulation seeks quantitative predictions, Joshua Epstein argues that prediction is not necessary an essential outcome of social science simulation (J. Epstein, 2008). Because of difficulties of parameterization and the insurmountable Popperian issues of the growth of human knowledge, a minimal desirable outcome is the generation of macro-level social observables from micro-level agent behavior and communication. Predicting emergent behavior in complex systems is inherently difficult. Thus finding the low level drivers of systems to produce observed outputs is usually non-trivial. Thus the sweet spot for ABMs at present are HABMs, which successfully describe trends and transitions rather than precise quantitative metrics. Examples abound, in Epstein’s book (2006) for anthropology, Paul Ormerod´s work in economics, with examples in books such as Positive Linking, or diverse social phenomena in Nigel Gilbert’s Agent Based Models(Gilbert, 2008).
2.1.
2.2Modelling social behavior for community-based social marketing
CBSM is an approach to marketing adopted from research in social psychology, which aims to identify barriers for the uptake of marketing messages, as well as communication strategies to overcome these barriers (McKenzie-Mohr 2011). For CBSM, individuals are embedded in a social, political, environmental and economic system. As their behavior depends on different factors, the identification of barriers require a systemic approach (Israel, Schulz, Parker, & Becker, 1998). The features of ABM allow Community-Based Social Marketers to simulate the behavior of a community in terms of their sensitivity to assumptions around the effectiveness of marketing within the community.
Studies of CBSM have worked with community partners using systematic, data driven marketing to design behavior change strategies (Flocks et al., 2001). These have ranged from the use of focus groups, observational studies, interviews and surveys. In some cases, statistical techniques such as regressions have been used to identify correlations between certain behaviours (McKenzie-Mohr 2011): for example, a study that compares non-composters and composters showed that individuals who compost, perceive that the environmental benefits of composting are so important that they far outweigh the inconvenient and unpleasant task of composting. However the approaches and techniques used so far for research in CBSM in environmental practices, such as environmental tourism, have marked limitations. These include the failure to include systemic approaches and the inefficiency of information-intensive marketing techniques. In this paper it is proposed that the use of ABM in CBSM campaigns could help make up for these limitations.
2.2.1.CBSM as a systemic approach
The bedrock of social marketing, according to Flocks and collaborators (2001), is an intensive formative research, which involves data collection and analysis activities before a marketing program begins, in order to understand and perceive a product. In addition, CBSM research has included systematic and intensive data collection techniques (surveys, interviews and focus groups), which also aim to give a systemic view of the product or service. Nevertheless, as with other “systemic” studies, these approaches resort to “cherry picking” in order to identify elements which are subjectively perceived to be of importance to the system (Bagheri & Hjorth, 2005).
In contrast, ABM is a mindset that consists of describing a system from the perspective of its constituents units, which in turn allows to look at the global behavior of the system (Bonabeau, 2002). ABM is interested in the emergent behaviour derived from behavioural rules of individual agents (Scholl, 2001). Emergence, in complexity theory, is understood as the property of complex systems where complex patterns arise from simple behavioural rules(Gilbert & Bankes, 2002). These rules include interactions with other individuals and the environment. Because of the interactions of agents, a system is more than the sum of its parts and the outcomes of the simulation can be counterintuitive (Bonabeau, 2002). Also, because we are dealing with individual agents, as opposed to aggregates, agent interactions are heterogenous and can generate network effects, through properties of memory and adaptation. This is of particular importance for this paper because CBSM makes use of social diffusion (communication among the members of a community) to overcome the barriers of the marketing campaign. Thus agents adapting and learning from the interactions with other agents and with the environment and the formulation of networks can assist marketers to simulate the effectiveness of marketing within the community.
2.2.2.From information-intensive to ABM marketing campaigns
Literature on CBSM articulates that information-intensive campaigns, such as media advertisements and booklets have been unsuccessful in influencing behaviors for the purpose of environmental enhancement (McKenzie-Mohr 2000). Millions of dollars have been spent in advertising energy or water efficiency. Nevertheless, the consumption of energy or water has not been significantly altered (McKenzie-Mohr, 2000). Aronson and Gonzales(1990) found that media advertising campaigns can be effective in creating public awareness but are limited when it comes to fostering behavioural changes.
In order to tackle this issue, ABM provides a highly visual, easy to understand and comparatively cheaper tool for marketers. Netlogo is an integrative modeling environment that allows users to visualize the emergent phenomena resulting for the behaviour of individual agents. It was designed for different audiences in mind and no programming background is needed to model related phenomena (Kornhauser, Rand, & Wilensky, 2007). Marketers can benefit from using Netlogo by presenting their clients, e.g. touristic service providers, with a simulation model of the marketing campaign and thus visually representing the possible implications of certain marketing decisions. Taking into account the benefits that ABM provide to CBSM, this paper provides a method for developing an ABM for the purpose of marketing environmental tourism, using the Winton Wetlands as a case study.
3.CASE STUDY: WINTON WETLANDS
The full modelling procedure consists of five stages. This paper addresses the building and testing of the ABM model in stage 3.
1. System boundaries: determine, through stakeholder involvement and revision of relevant documentation, the spatial and temporal boundaries of the system, which will help delimit the entire modeling process (Bagheri & Hjorth, 2005; Musters, de Graaf, & ter Keurs, 1998).
2. Main elements and interactions: identify the main social, environmental and economic elements of the Winton Wetlands system.
3. Model formulation: decision rules of the agents in Figure 2 are mapped using a flow chart and enteredinto Netlogo. Section 3shows a full description of the model.
4. Parameter estimation: parameter values are set for the model. This stage of the process can take place at the beginning of the modelling procedure and can be modified at any time.
5. Sustainability indicators: these are established for future use in monitoring of the system and to establish plausible scenarios of development.
To specify the ABM, the behaviour of the agents is translated into decision rules (Figure 1), which can then be translated into Netlogo code. The basic agents in the ABM are the “tourists”, who represent both the visitors to the site, as well as the general community who support the site through their taxes. The potential tourists start their journey in the city or hometown. Every turn, the tourists randomly decide if they wish to go to on a holiday or not (based on a probability of recreation). Once they decide to take a holiday, they choose to go to the wetland or to another destination (e.g. the beach) depending on their own previous experiences of the wetland (experience) as well as the experiences of other visitors to the wetland (word of mouth).
Adapted from the model formulated by Lacitignolaet al. (2007) in which they modeled interactions among the environment, nature-based tourism and tourists, the model proposed in this paper depends on several key assumptions about the behavior of the tourists, their interaction with marketers and their perception of and experience in the tourism site. These are detailed in Table 1.
Table 1: List of assumptions used to build the model
Number / Assumption1 / If tourists perceive that the ecosystem quality and infrastructure are poor, they will choose another destination
2 / If ecosystem quality and infrastructure are above a certain threshold, tourists will be more inclined to visit the site.
3 / Ecosystem quality and infrastructure are degraded by the arrival of tourists.
4 / Tourists’ perception of the site is dependent on their own previous experience at the site and word-of-mouth from other tourists and members of the community
5 / Pressures are exerted on ecological quality, including generation of waste and consumption of resources such as water and space (Buckley, 2004).
6 / Increasing tourist visits requires construction of new infrastructure as well as maintenance of existing infrastructure.
7 / Capital of the wetland increases through tourism revenue generated by the arrival of tourists
8 / Tourism revenue is a function of the number of visitors and level of community support.
9 / Tourism revenue and other grants, which the site management earns, are used for three main activities: the maintenance and construction of new infrastructure, the restoration of ecological quality of the wetland, and the implementation of the marketing campaign.
Assumptions 1, 2, 3, 5 and 6 relate to tourists’ perception of the wetland and their impact on the ecosystem quality and infrastructure. Assumption 4 also incorporates the role of word-of-mouth in tourists’ perception of the site. Assumptions 7, 8 and 9 describe the effect of tourism on change in tourism revenue and distribution of the revenue to different parts of the tourist site. The decision by the tourist site management to allocate the revenue is simulated by allowing the managers a fixed numbers of workers (agents), who can either be assigned to restore environmental capital (“ecologists”), site infrastructure (“builders”) or influence tourists through marketing campaigns (“marketers”).