Service interaction design: A Hawk-Dove game based approach to managing customer expectations for oligopoly service providers

A dual-level (desired and adequate levels), dynamic model of customer expectations provides the optimal method for characterizing customer expectations (Parasuraman et al. 1991). The desired and adequate expectation levels represent the service levels that the customer hopes for, and those that are acceptable, respectively. The desired level combines what the customer believes the product “can be” with what they believe it “should be”. The adequate level is partially based on the customer predicted service level. The tolerance zone varies among customers, and even among situations for a single customer. Initial customer expectations are related to expectation disconfirmation, the degree of which influences customer satisfaction. Zeithaml et al. (1993) presented a conceptual framework for specifying service expectations and elucidated customer expectations using 11 antecedent factors (transitory service intensifiers, perceived service alternatives, customer self-perceived service role, situational factors, predicted service, enduring service intensifiers, personal needs, transitory service intensifiers, perceived service alternatives, customer self-perceived service role, situational factors and predicted service), which directly and indirectly affect desired and adequate expectations (as listed in Fig. 2) (Zeithaml et al. 1993). For example, when an automobile provider produces advertisements or contracts that explicitly promise customers it will provide perfect maintenance service, customer expectations increase accordingly. Perceived service alternatives are considered, for example: if customers can choose alternative service providers or services, then their levels of adequate service may increase highly. Accordingly, this work uses these determinants of expectation to define service components and thus manage customer expectations.

Understanding customer expectations requires service providers to successfully reach the customer franchise (Parasuraman et al. 1991; Kurtz and Clow 1993; Rust et al. 1999; Haeckel et al. 2003). The conceptualization of multiple expectations influences service provider allocation of resources (Walker and Baker 2000). Service providers can fulfill customer expectations and achieve customer satisfaction if they deliver services in a manner that considers the determinants of expectation. Understanding customer tolerance zone is crucial to ensuring customer satisfaction (Johnston 1995). Managing customer expectations is therefore incorporated into the proposed interaction design mechanism, affecting service delivery and ensuring customer satisfaction.

Competent service providers in an oligopoly market can manage customer expectations by raising the adequate expectation (from Adequate to Adequate’) and stabilizing the desired expectation to reduce the tolerance zone (as shown in Fig. 3) and thus increase barriers to competition. For example, since Apple uses branding and promotion to raise customer expectations of its product, customers expect the product to be superior to alternatives. Customer loyalty proves that Apple has successfully managed customer expectations by raising them, and has achieved a special position in the computer systems market.

In nature, individuals must forage and sometimes fight for food. The Hawk-Dove game represents one application of game theory to animal behavior (Kokko et al. 2006). The game describes the situation in which contestants can select either an aggressive (Hawk) or non-aggressive (Dove) strategy to compete for a collective resource. Hawks use every opportunity to steal or defend food, while Doves never steal or resist food thieves. These two birds thus comprise a polymorphic population. Mathematically, the Hawk-Dove game is a non-zero sum matrix game with the payoff matrix in Table 1, where V denotes the value of the contested resource; C is the cost of an escalated fight (Smith and Parker 1976), and the former is assumed to be less than the latter (C > V > 0).

The payoff matrix displays four possible fight situations. In the first case, when Hawk encounters Hawk, the winner obtains a resource of value V while the loser bears the cost of the fight, C, or both participants bear the cost of the fight equally, C/2. In the second case, when Hawk meets Dove, the Hawk monopolizes the entire resource and leaves the Dove with nothing. In the third case, when Dove meets Hawk, the Dove receives no resource. In the fourth and final case, two Doves meet and share the value of the resource with V/2.

The relationship between V and C is associated with different population equilibria. Equilibrium refers to the situation when game participants adopt the optimal combination of strategies. By approaching equilibrium, players can optimize resource value. In the aforementioned classic Hawk-Dove game, if C > V > 0, then the proportion of individuals in the population that adopts the Hawk strategy tends to equal V/C at equilibrium. (If C < V the environment favors Hawks, but this factor is not considered here.) Each individual considers value and cost in strategy selection. Strategy selection may eventually approach a stable evolutionary state (ESS). The concept of ESS involves the characteristics of specific Nash equilibriums, which are immune to the invasion of mutant strategies and were developed during the 1970s (Smith and Price 1973). ESS equilibrium has been shown strong enough to attract players who could perform better in other equilibriums, such as the dominated equilibrium (Berninghaus and Ehrhart 2003), particularly when players who are not fully informed of the payoff functions are unaware that they are selecting an ESS.

This study adopts the Hawk-Dove game to simulate the costs incurred, and the value gained, by oligopoly service providers and customers, to examine strategy selection. The evolutionary dynamics associated with consecutive games are considered to be based on repeated interactions between provider and customers in service delivery, and therefore the ESS equilibrium is used to guide strategy selection.

The context detection module is designed to identify the environment during service delivery. Service operations and resources must be carefully considered to optimize service delivery performance and efficacy. The proposed module must recognize customer feedback and behavior, which can be immediately and accurately inputted to the mechanism to modify the determinant decision module. The learning approach thus is critical to service delivery in a dynamic context.

The determinant decision module uses useful determinants to influence customer expectations based on inputs from the context detection module, customer preferences (such as age, brand interest, price etc.), and data from the encounter database, including previously collected customer data, information regarding delivered services, and so on. This module then automatically analyzes and converts the encounter data, customer preferences and the key performance indicators of the organizer (KPIs) into computable scores to yield the weight of each determinant, indicating the strength of its influence on customer expectations in the service context. This module selects candidate determinants of expectation that can be adopted to alter customer expectations regarding service delivery. Since this work applies the proposed approach to the exhibition service sector, one domain expert and one researcher with over a decade of experience of this service sector help define a reasonable range for each weighting and parameter.

The Hawk-Dove game module determines suitable determinants of expectation for exploitation in service encounters. After the determinant decision module obtains the weight of each candidate determinant of expectation, these determinants are inputted to the Hawk-Dove game module, which thus calculates the effective interaction design for use in service delivery. Figure 5 displays the procedure followed by the Hawk-Dove game module. Based on the input data from the determinant decision module and information on the ESS goal database and customer requirements, the ESS goal identification component can identify goals for customer expectations. Finally, these determinants are inputted to the solution selection module, which is outlined below.

The Hawk-Dove game models a competitive situation involving a shared resource, in which contestants can adopt either an aggressive (Hawk) or passive (Dove) competitive strategy. Mathematically, the Hawk-Dove game is a two-player non-zero sum matrix game (Grundman et al. 2009). This study applies the Hawk-Dove game to design interactions between an oligopolistic service provider and its customers. Either the oligopoly service provider or its customers must attempt to maximize profits while reducing the costs of service delivery. Not only the does the oligopoly service provider meet customer needs at relatively high cost, but customers must make more effort to obtain excellent service. From this perspective, each interaction can be regarded as a non-zero sum game.

After the Hawk-Dove game module calculates the determinants to be exploited in service delivery, the solution selection module identifies the most suitable service components for managing customer expectations in the value-bounded context. The identification is confirmed by predicting the effects of the chosen determinants using the expectation measurement module, which provides both the scores of the dual expectation measurements and appropriate expectation tactics.

This work assumes that each determinant of expectation is associated with corresponding service tactics (expectation tactics), which are used to select suitable service components, supporting a flexible method for delivering service components. The recommended product is an example of service tactics used to represent the determinant of expectation, Word-of-Mouth.

The effects of the selected determinants on customer mental state are evaluated to determine the feasibility of employing the expectation measurement module (as detailed in section 3.5). The expectation measurement module can measure customer expectations in real time and in dynamic environments that are affected by the selected determinants.

Based on the chosen expectation tactics, the solution selection module selects service components for implementation. When customers alter their behaviors and service encounter points, the proposed mechanism must interact with customers immediately and dynamically. In this case, the greedy algorithm selects the optimal service components for efficient solution identification. Selecting the optimal combination of service components is a knapsack problem in which the oligopoly service provider must select service components to maximize benefits given a limited budget. The knapsack problem is the following maximization problem.where q denotes the total number of service components, and each service component j has cost co _j and profit p _j . The oligopoly service provider determines the upper bound on the total cost, co _t , which is the sum of the costs of labor, technology and materials. This module sets the values of y _j at y _j  = 1 if the knapsack contains service component j and at y _j  = 0 otherwise. The greedy strategy selects the alternative with a maximum p _j /co _j that fits into the knapsack.

The service components identified by the greedy strategy may not be optimal for all implementations, but represent the best combination that the oligopoly service provider can identify given time constraints. Since all service components are allocated by customers and service providers, the oligopoly service provider can estimate costs and profits for each service component during the first stage of service component design. For example, when the oligopoly service provider adopts advertising, it can evaluate the cost and the effect of different advertisements (such as a commercial video or customer mail-outs). Following service component selection, this module considers the context factors (such as the distance between booths or how crowded each booth is, and uses this to arrange the encounters in order) to generate a service delivery template. After the service components are exploited in service delivery, the information obtained by the oligopoly service providers and customers (benefit and cost) is retrieved and fed back to the Hawk-Dove game module.

The expectation measurement module measures the likely performance of the selected determinants identified by the selection solution module by calculating the customer expectation scores. This process helps the solution selection model choose appropriate service components. The expectation measurement module is critical to determining customer mental state (Hsieh and Yuan 2010), and ensuring the integrity and effectiveness of the Hawk-Dove game-based interaction design mechanism. Owing to space limitations, the measurement model is not described in detail here.

The outputs of the expectations measurement model include the adequate expectation score, desired expectation score and a list of recommended expectation tactics. Once the oligopoly service provider determines actual customer expectations based on these outputs, it can propose suitable services to help customers achieve their business goals. Furthermore, the list of expectation tactics, derived from a real-time database, provides a reference. This list of appropriate expectation tactics can be mapped to specific service components to affect customer expectations via the solution selection module. After implementing the expectation tactics (service components), the solution selection module should store (the values of expectation variations and information of providers’ capabilities in a real-time database. The expectation measurement model can then use feedback control to reflect actual customer expectations.

The Hawk-Dove game-based interaction design mechanism is applied to enable oligopoly service providers to deliver services that ensure customer satisfaction by managing expectations. Initially, needs and preferences are specified to identify customer values, and enable service providers to design and deliver effective services during service experience delivery. Customer expectations determine customer satisfaction during service delivery in a dynamic context. This study applies an innovative method for enriching customer experiences that considers customer mental state. This work also utilizes animal behavior to mimic service interactions between customers and providers. After selecting the determinants of expectation, a greedy algorithm is adopted to choose appropriate low cost and high efficiency service components to enable customers to build a high performance eco-system. Therefore, the Hawk-Dove game-based interaction design mechanism can be considered a substantial and theoretical interface design artifact, comprising five core modules and using customer expectation management to answer the research questions. The following section tests the novel mechanism using experimental simulations.

The first set of experiments studies the number of iterations in each simulation of the use of determinants of expectation. When the number of iterations exceeds the threshold, the mechanism identifies the determinant of expectation as impractical. This set of experiments determines the optimal number of iterations for causing the simulation results to converge in a particular direction.

The Hawk-Dove game is designed to achieve an evolutionary balance between the various ecology approaches. In the exhibition scenario, equilibrium is defined as total customer profit for the scenario involving a stable oligopoly service provider. Customer reactions to the service context are important in designing effective customer encounter interactions. The total payoff for customers and the oligopoly service provider, and the strategies available for adoption by customers, thus are considered indicators of game stability. The experiments involve randomly setting game conditions (such as payoff) in various games. This work simulates ten games to examine the number of iterations within which most games reach equilibrium. At the beginning of each game, roughly half of all customers use the Dove strategy, while the remainder are assigned to the Hawk strategy. This experiment thus defines ten games that 1000 customers and one oligopoly service provider must adopt within 50 iterations. Table 7 indicates the detailed parameter settings.

In Fig. 9, the Y axis represents the percentage of customers adopting the Hawk and Dove strategies and the X axis represents the number of iterations. Customer strategies (in ten games – including game1, game2 and so on) may change significantly during the first ten iterations and then gradually evolve toward a single strategy. Consequently, a specific strategy is best suited to each game, each with its own conditions. The total payoffs to the customer and oligopoly service provider also change, as shown in Fig. 10. However, when the payoff from one strategy clearly exceeds that from the other, and when the behavior of the oligopoly service provider changes in response to learning, further iterations may be required to achieve ecological stability of both customers and oligopoly service provider agents. As shown in Fig. 10, most games cease to evolve after an average of ten iterations, and all converge within no more than 40 iterations.

To efficiently simulate the exhibition context and the behavior of customers and oligopoly service providers, this experiment attempts to optimize the number of iterations in the game simulation to minimize resource wastage in the remaining game runs. From the experimental results, 10 ~ 40 iterations can be used as the default number of iterations in subsequent Hawk-Dove game simulation experiments.

The Hawk-Dove game-based interaction is developed to manage customer expectations with reference to the objectives of oligopolistic service providers. Since customer expectations vary in dynamic environments, customer stereotypes and their preferences should be considered. The second experiment aims to analyze and manage customer expectations to refine the interaction mechanism by clarifying different customer stereotypes and their preferences. Accordingly, this experiment verifies that the Hawk-Dove game-based interaction design mechanism is an effective means of managing customer expectations and identifying customer stereotypes. Expectation management is designed to help oligopoly service providers increase their expectations of what is adequate and stabilize their expectations of what is desirable. Consequently, the zone of tolerance determined by the level of desired expectation is narrowed. This experiment exploits this mechanism to examine the effects of different customer stereotypes on customer expectations.

Five factors (including arrival, capability, effort, request, and subjective preference) influence customer behavior (Frei 2006). Table 8 lists various customer stereotypes. Stereotype 1 involves moderate effort; stereotype 2 makes a serious effort to bargain with the oligopoly service provider, and stereotype 3, which is the opposite of stereotype 1, involves little effort.

This experiment not only employs the above three stereotypes to demonstrate how the Hawk-Dove game-based interaction design mechanism can be used to manage customer expectations in service encounters, but also compares the measured adequate and desired expectations of each customer stereotype. For simplicity, this work assumes that all customers have the same initial scores associated with expectation states (where 3.5 means the adequate expectation while 7 indicates the desired expectation). The Hawk-Dove game-based interaction design mechanism organizes ten service encounters for each customer and uses the appropriate determinants in each encounter (as listed in Table 9). Each experiment involves three users of each stereotype to simulate and record the changing expectations associated with each encounter in ten runs. The results in Figs. 11–12 are obtained by averaging the variations in user expectations for a particular stereotype.

Figure 11 demonstrates that the adequate expectation of stereotype 1 increases with number of encounters. For stereotypes 2 and 3, the mechanism slightly increases adequate expectations. Regarding stabilization of desired expectations, Fig. 12 also shows that customer expectations are stable for each customer stereotype. Furthermore, with respect to the goals of increasing the level of adequate expectation and stabilizing the desired level, the experimental results indicate that the mechanism was successful, particularly for stereotype 1.

Stereotype 1 is frequently considered to represent the target customers of the oligopoly market, who are generally able to independently select the best service type since they have requisite domain knowledge (Kuusisto 2008; Rowley 1996). Stereotype 1 can also make more effort to become involved in service experience delivery (Kuusisto 2008; Rowley 1996). When stereotype 1 customers lack service choices, they may adopt medium or low expectations. The experimental results indicate that oligopoly service providers can provide stereotype 1 customers with the opportunity to reflect on their needs and become involved in service experience delivery using interactions designed specifically to manage expectations. The proposed mechanism can modify the determinants when customer expectations move in the opposite direction to that desired or exceed the expected range. Accordingly, the Hawk-Dove game-based interaction design mechanism not only provides customers with required services, but also reliably manages their expectations.

The third experiment tests whether the Hawk-Dove game-based interaction design mechanism can deliver services that meet customer needs. The Hawk-Dove game algorithm draws on a fight between a hawk and a dove and is used to simulate interactions between oligopoly service providers and customers. This experiment assesses the effectiveness of the Hawk-Dove game-based interaction design mechanism.

Table 10 shows the payoff matrix of the proposed experiment based on the concepts listed in Table 3. The recommendation determinant denotes the average number of customers who recommend a particular service component to others. A higher recommendation means more customers enjoy and benefit from the service. Regarding the costs of experiencing a service, customers must spend time and effort to use a service component. Accordingly, the customer payoff can be calculated as the determinant profit minus the determinant cost. Furthermore, the payoff of the oligopoly service provider is the benefit of the determinant minus the exercise cost. After interacting with the oligopoly service provider, customer agents can take various steps to increase profits from service encounters, as follows.

These experiments collect the customer satisfaction score for each service component and average these scores for service components arranged in a journey to indicate the degree of customer satisfaction with each journey. The experiments use the degrees of satisfaction of the journey arrangements to evaluate the performance of the service interaction mechanism in examining if the service components dynamically arranged in service encounters are appropriate for customers. In contrast with the effect of the Hawk-Dove game interaction design mechanism, the experiments also conduct an approach in which the service components are selected randomly during each service encounter of a journey. In these experiments, the customer reaction as measured by satisfaction score is derived based on analysis of the post-report of AutoTronics Taipei 2009 conducted by the Taiwan External Trade Development Council (TAITRA) which asked visitors to rate their impression for every service component, with the associated scores becoming the basis for assessing the effect of service components.

Figure 13 shows the results of customer satisfaction and every point represents the average degree of satisfaction for each journey of the three visitors. The curves indicate that the degrees of satisfaction for the journeys managed by the Hawk-Dove game based interaction mechanism (around 80 % of customer satisfaction) are higher than when the service components are randomly assigned (around 50 % of customer satisfaction). This indicates that the service components could effectively fulfill customer needs in delivering customer value and services when using the Hawk-Dove game interaction design mechanism.

This set of experiments is designed to determine how changing visitor expectations increase stakeholder benefits (corresponding to high performance). This set of experiments evaluates the overall performance of the Hawk-Dove game-based interaction design mechanism from a macro perspective. A high-performance ecosystem maximizes benefit to all stakeholders (oligopoly service providers and customers). Surplus value theory (Marx 1952; Carlsson and Davidsson 2002) is used to evaluate eco-system performance by measuring stakeholder perceptions of value. The equation of surplus value is as follows. where S denotes surplus value; P represents total value; C is total spending on investments and material; V denotes spending on labor, and (C + V) represents total cost. To model surplus value of an exhibition, the definition of surplus value is extended to include customer and oligopoly service provider value and costs.

This set of experiments uses pareto optimization (Hochman and Rodgers 1969; Sawaragi et al. 1985; Jin and Sendhoff 2008) to evaluate the performance of the Hawk-Dove game-based interaction design mechanism in the target ecosystem. A pareto-optimal outcome is one in which no-one is made better off without making someone else worse off. Customer services are supposed to be pareto-optimal solutions, obtained via the Hawk-Dove game-based interaction design mechanism, that consider various interaction design factors, including profit and cost. The measurement model thus measures the maximum surplus value of all stakeholders achieved using the Hawk-Dove game-based interaction design mechanism with the following objective functions.

Table 12 lists the definitions of the experimental parameters. P _p represents the ratio of variation in the quantity of the zone to the original quantity of the zone, and the total cost of implementing service components during the journey is (C + V) _p. The mechanism represents the customer satisfaction score for all service components experienced in the delivery of a service as P _c. Finally, the total time spent interacting with every service component in the service encounters is represented by (C + V) _c . The Hawk-Dove game-based interaction design mechanism arranges 25 customers. The customer satisfaction score for each encounter minus the effort made by the customer to interact with a service component is S_c. The variation in the size of the zone as a proportion of the original quantity of the zone minus the cost of implementing all service components is S_p. Twenty-five customers with randomly assigned S_c and S_p are considered to model all possible decisions made by all other possible mechanisms. The random rules are taken primarily from domain experts and practical reports. Comparing these two groups of customers in terms of overall surplus value clarifies the effectiveness of a Hawk-Dove game interaction design mechanism in creating a high-performance eco-system.

Based on the results of the ecosystem evaluation experiment (as shown in Fig. 14), each point represents a visitor. (A red square represents a customer served using the proposed mechanism and a blue rhombus indicates a customer with randomly assigned values of S_c and S_p). Except for a single square point, square points outperform rhombus points. If any points in the rhombus group exhibit quality performance of S_p, those same points must be associated with poor performance of S_c. In conclusion, the simulation results clearly indicate that the Hawk-Dove game interaction design mechanism can not only be used to build a well-performing ecosystem but can also benefit customers and providers by providing pareto-optimality.

Service interaction design is important in determining the performance of service experience delivery. This study uses a systematic and quantitative approach, the Hawk and Dove game, to model service interactions between oligopoly service providers and customers where customer expectations are dynamically but systematically managed. Restated, service providers consider customer expectation management during their design of the service interaction aspects of service experience delivery.

Traditionally, service providers have used the survey method (questionnaires) to determine customer expectations, but only after service completion. Understanding customer expectations in real time in service contexts is particularly difficult for service providers who deliver services in real time. This study presents a service interaction mechanism, which involves customer expectations based on elicited service-related requirements and needs. Service providers can thus provide required services based on this information. Customers can choose to receive available services and modify their requirements based on what they receive. Service providers thus can dynamically provide customers with satisfactory services by immediately analyzing customer feedback to clearly understand customer expectations in real-time service contexts. Consequently, both service providers and customers can achieve their goals by participating in interactive service activities.

As indicated above, customers can be grouped into numerous types with different levels of expectation. Service providers can deliver suitable services when they clearly understand customer expectations. For instance, if a service provider knows that a customer has low expectations of frontline services, they can reduce the number of service employees and devote resources to preparing other useful services for the customer. More generally, service providers can dynamically but effectively allocate existing resources to provide customers with more suitable services. Service providers do not need to provide customers with all services, necessarily increasing the cost of service provision. Hence, the proposed service interaction mechanism helps service providers who can flexibly modify their service-related resources provide services that meet customer expectations in a cost-effective manner.

One core feature of S-D logic is the co-creation of value by service providers and customers during service experience delivery. The proposed service interaction mechanism can be considered a platform on which service providers and customers can co-create value. When service providers provide services to customers via the proposed service interaction mechanism, customers can participate in the service activities to improve their own service experience. Customers interact with both service providers and other customers to share the experiences and knowledge gained during service experience delivery via the proposed service interaction mechanism. Both service providers and customers create value by participating in service activities. Therefore, customer involvement is essential to successful service experience delivery.

Since the proposed service interaction mechanism is designed to be trained and improved by the input of real-world data, the number of training sets and parameters requires careful consideration. In this work, experts on the exhibition industry helped specify suitable training sets and parameters for the experimental simulations. For example, the numbers of training iterations required to ensure the convergence of the HDG method and the customer expectation measurement model are 40 runs and ten runs, respectively. However, since the settings were only imprecisely specified, to imitate the behavior of real-world exhibition visitors, the parameters of the proposed service interaction mechanism must be further tuned by increasing the inputted visitor data. Although continuous learning and searching for suitable parameters in real-time service contexts is necessary, the constraints imposed on service providers by resource limitations and costs are the main factors in terminating training iterations.