See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/359134308

Developing a Service Quality Scale for Artificial Intelligence Service Agents

Article in European Journal of Marketing · March 2022

DOI: 10.1108/EJM-09-2020-0672

CITATIONS READS

27 2,986

3 authors:

Nurhafihz Noor Sally Rao Hill

James Cook University Singapore University of Adelaide
15 PUBLICATIONS 85 CITATIONS 93 PUBLICATIONS 2,199 CITATIONS

SEE PROFILE SEE PROFILE

Indrit Troshani
University of Adelaide
133 PUBLICATIONS 3,186 CITATIONS

SEE PROFILE

All content following this page was uploaded by Nurhafihz Noor on 30 April 2022.

The user has requested enhancement of the downloaded file.

 Developing a service quality scale for artificial intelligence service agents

Nurhafihz Noor
Adelaide Business School
The University of Adelaide
[email protected]

Sally Rao Hill

Adelaide Business School
The University of Adelaide
[email protected]

Indrit Troshani
Adelaide Business School
The University of Adelaide
[email protected]

Accepted for publication, suggested citation:

Noor, N., Rao Hill, S. and Troshani, I. (2022) ‘Developing a service quality scale for artificial intelligence service

agents’, European Journal of Marketing, doi: 10.1108/EJM-09-2020-0672.

1
 Developing a service quality scale for artificial intelligence service agents

Abstract

Purpose – Service providers and consumers alike are increasingly adopting artificial intelligence service agents

(AISA) for service. Yet, no service quality scale exists that can fully capture the key factors influencing AISA

service quality. This study aims to address this shortcoming by developing a scale for measuring AISA service

quality (AISAQUAL).

Design/methodology/approach – Based on extant service quality research and established scale development

techniques, the study constructs, refines and validates a multidimensional AISAQUAL scale through a series of

pilot and validation studies.

Findings – AISAQUAL contains 26 items across six dimensions: efficiency, security, availability, enjoyment,

contact and anthropomorphism. The new scale demonstrates good psychometric properties and can be used to

evaluate service quality across AISA, providing a means of examining the relationships between AISA service

quality and satisfaction, perceived value as well as loyalty.

Research limitations/implications – Future research should validate AISAQUAL with other AISA types as they

diffuse throughout the service sector. Moderating factors related to services, the customer and the AISA can be

investigated to uncover the boundary conditions under which AISAQUAL is likely to influence service outcomes.

Longitudinal studies can be carried out to assess how ongoing use of AISA can change service outcomes.

Practical implications – Service managers can use AISAQUAL to effectively monitor, diagnose and improve

services provided by AISA, whilst enhancing their understanding of how AISA can deliver better service quality

and customer loyalty outcomes.

Originality/value – Anthropomorphism is identified as a new service quality dimension. AISAQUAL facilitates

theory development by providing a reliable scale to improve the current understanding of consumers’ perspectives

concerning AISA services.

Keywords Artificial intelligence service agents; service quality; scale development; customer service

Paper type Research paper

2
Introduction

Recent years have seen rapid advances in artificial intelligence (AI) technology development

and the emergence of a plethora of applications powered by the technology (Huang and Rust,

2021, Rust, 2020, Davenport et al., 2020). AI can be defined as technology (e.g., machine

learning, big data, natural language processing and understanding) that enables software agents

to “act intelligently” (Poole and Mackworth, 2010, p. 3). AI software agents are systems or

machines that can complete tasks that typically require human intelligence (e.g., problem

solving) in rational ways, achieving the best possible (expected) outcome, given the

information available to them (Russell and Norvig, 2018, Poole and Mackworth, 2010).

There is a strong belief that AI is a key force in the expansion of the services industry

and will have far-reaching and broad impacts on business (Huang and Rust, 2021, Rust and

Huang, 2014, Noor et al., 2021b, Fountaine et al., 2021). Emerging evidence supports this

belief (Colback, 2020, Chui et al., 2020). AI offers business and service providers the potential

to greatly boost revenue (e.g., by improving support for business and marketing decisions) and

reduce operational costs (e.g., via automation) (Davenport et al., 2020, Neuhofer et al., 2020,

Prentice et al., 2020, Chui et al., 2020). Strong market value growth forecasts for AI agents

illustrate their potential (Makadia, 2020).

For example, the market value of chatbots and virtual assistants, two common types of

AI agents used by business to provide service to consumers (Zarouali et al., 2018, Hoy, 2018),

is expected to increase at the compound annual growth rate of 33% between 2020 and 2025

(AMR, 2020). This business case is motivating innovative service providers to use AI agents

to provide service to consumers over either part of or the entire customer service consumption

journey (Prentice et al., 2020, Oosthuizen et al., 2020, Robinson et al., 2020, De Keyser et al.,

2019, Paluch and Wirtz, 2020, Wirtz et al., 2018).

3
 In this study we are concerned with the broader issue of measuring how consumers

perceive the quality of service provided by AI service agents. Following Zeithaml (1988),

service quality can be defined as the overall excellence or superiority of the service

performance by a service agent as perceived by consumers. Traditionally, the role of service

agents has been performed by human service employees and/or technology-based agents, also

known as self-service technologies (SSTs), which enable consumers to (co)produce service

without the direct involvement of the employees (Meuter et al., 2000, Bitner et al., 2000,

Grewal and Levy, 2009, Lin and Hsieh, 2011). AI service agents, henceforth AISA, constitute

a new service agent type that is based on AI technology. Following Wirtz et al. (2018), we

define AISA as “system-based autonomous and adaptable interfaces that interact, communicate

and deliver service to an organization’s customers” (p. 909). Examples of AISA include

customer service chatbots, virtual assistants, social robots, and autonomous cars (Noor et al.,

2021a).

We argue that using AISA to provide service to consumers will change the nature of

service encounters. A service encounter is the moment of interaction between a consumer and

a firm (Bitner et al., 2000) where the firm is traditionally represented by service agents.

However, AISA are radically different to traditional, established service agents in that

computer-based applications act in ways that emulate humans (Park et al., 2021, Huang and

Rust, 2021). The differences can contribute to different consumer experiences and

consequentially dramatically alter consumers’ perceptions of service quality (Lu et al., 2020,

Paluch and Wirtz, 2020, Huang and Rust, 2018, Verhagen et al., 2014, Huang and Rust, 2021).

Service research in marketing has long established that “quality occurs during service

delivery” (Zeithaml, Berry, and Parasuraman, 1988, p. 35). Service quality is thus widely seen

as a key determinant of the service providers’ long-term performance and success (Zeithaml et

al., 2002, Fassnacht and Koese, 2006, Parasuraman et al., 1985, Zeithaml et al., 1996a,

4
Zeithaml et al., 1988). Indeed, research looking at the role of consumers’ perceptions of the

quality of service offered electronically (e.g., via websites) found that such perceptions can

affect a range of constructs, including customer satisfaction, loyalty, and attitudes towards

ongoing use and recommendations, which in turn affect retailers’ long term performance but

were also found to be a key determinant of long term performance and success (Ladhari, 2010,

Zeithaml et al., 2002, Fassnacht and Koese, 2006, Cronin Jr and Taylor, 1994). Additionally,

research from practice has shown that up to 42% of retail consumers show greater purchasing

interest after experiencing good customer service, whilst 52% stopped purchasing due to a

single disappointment in a customer service encounter (Zendesk, 2020).

There is limited understanding of how consumers’ service encounters with AISA affect

their perceptions of service quality and how the construct can be measured. This is problematic,

particularly given the increasing number of service providers that are using or considering

using AISA for service delivery (AMR, 2020). For example, Chui et al. (2020) report findings

from a 2019 McKinsey survey indicating that 58% of surveyed businesses were using AI in at

least one business function or unit, with product/service development and operations being the

most commonly used functions, including AI-based enhancements, feature and operations

optimization, and predictive service and interventions ranging between 19 and 24%. Similarly,

Makadia (2020) cites a recent Gartner report forecasting that over half of established businesses

are investing or considering investing in AI.

The business case for using AISA for service delivery is based on service providers’

belief that consumers will accept and use AISA. However, evidence in the literature is mixed,

challenging the foundation of this belief. For example, research discussing the use of AISA in

service has argued that consumers should experience a range of benefits from the service

provided by AISA, including greater efficiency and personalization (Huang and Rust, 2018,

Huang and Rust, 2021, Verhagen et al., 2014). These benefits are usually directly related to

5
service quality which implies that they should improve the customers’ perceptions of the

service quality delivered by AISA. However, recent research that has empirically assessed the

role of AISA in service shows mixed findings in terms of how consumers evaluate AI services.

For instance, looking at the role of AI in the hospitality and hotel context, Prentice et al. (2020)

find that, separately, both AI and employee service quality are significantly related to customer

satisfaction. However, when the two constructs were considered jointly (i.e., regressed in the

same equation), AI service quality had a negative effect on consumer satisfaction whilst

employee service quality explains a significant part of the overall assessment of the quality of

the service provided by both AI and service employees (Prentice et al., 2020). We note that

whilst Prentice et al. (2020) adopted 15 items from Makadia (2018) to represent a suite of AI-

based services specific to the hotel context, there is no indication (see e.g., Makadia, 2018)

how the items were generated and validated. More recently, research has also shown that how

consumers perceive AI service quality also depends on the type of service. For instance, a

recently published study by Park et al. (2021) shows that perceived AI service robots’

usefulness is significant for credence services (e.g., at a hospital) but not significant for

experience services (e.g., at a café). Overall, this evidence supports similar outcomes from

AISA services that were anticipated by Wirtz et al. (2018) whilst also extending similar

evidence by Elliot (2018) who also found that consumers prefer service employees to AISA in

service encounters.

Taken together, this research suggests that there might be a customer gap between

customers’ AISA service needs and expectations and their perceptions of the quality of the

services that are actually delivered by AISA (Prentice et al., 2020, Bitner et al., 2010). By

implication, there is a need to improve the current understanding of consumers’ perspective on

AISA service quality (Prentice et al., 2020, Lu et al., 2020, Paluch and Wirtz, 2020, Huang

and Rust, 2021).

6
 Noor et al. (2021b) follow a qualitative approach to identifying twelve service quality

dimensions representing AISA service quality. Ten of the proposed dimensions are adapted

from key, established human- and technology-based service quality dimensions, with two

dimensions reflecting AISA’s unique underlying characteristics. Whilst Noor et al. (2021b) is

a key study that makes important inroads into improving the understanding of AISA service

quality, the proposed dimensions lack empirical, psychometric development and validation

which limits their contribution to theory and practice.

We extend their work in this paper. Specifically, following on a strong tradition of

established, proven scale-development processes and techniques (e.g., Churchill Jr., 1979,

Netemeyer et al., 2003), we build on Noor et al. (2021b)’s proposed qualitative dimensions to

construct, refine and validate a multiple-item scale for measuring the AISA service quality of

chatbots and virtual assistants, two types of AI service agents that are commonly used by

service providers in the current market.

Our study culminates with an AISA service quality scale, henceforth AISAQUAL. For

the purposes of this study, we follow Parasuraman et al. (2005) to define AISAQUAL as the

extent to which AISA facilitate an overall perception of excellence or superiority as perceived

by consumers. AISAQUAL is comprised of six dimensions, including 26 item measures. We

thus contribute by providing a psychometrically validated, reliable service quality scale

specifically developed for AISA. This scale can inform and help better explain theoretical

findings and conclusions from emerging research in this area whilst also offering a new basis

and opportunities for the development of new theoretical insights into the growing, novel

domain of AI in service (Ranjan and Read, 2016). Specifically, the AISAQUAL scale can

improve current understanding of the factors that might help reduce or bridge the customer gap,

including service providers’ understanding of customer expectations of AISA services, service

design and standards, and service performance and delivery (e.g., Bitner et al., 2010, Prentice

7
et al., 2020). The AISAQUAL scale also offers a robust tool for service providers to better

understand the effective adoption of chatbots and virtual assistants as popular AISA in the

competitive service sector and emergent issues and implications. Overall, AISAQUAL and

related development work provides an important stepping stone towards addressing a critical

shortcoming in the service quality literature in the area of AI in service whilst also forming a

new platform for researching other types of AISA (Lu et al., 2020, Bock et al., 2020).

The paper is structured as follows. In the next section, we review related literature on

service quality measurement and the recent research concerning AISA service quality. We then

proceed to discuss the development of AISAQUAL, before concluding with the theoretical and

managerial implications as well as directions for future research.

Related literature
Technology in service

Rapid technological advancements have seen the growing role of technology in service

delivery (Bitner et al., 2010). Service providers have used technology aimed at improving

customer experiences. Broadly, technology has been used in two main ways (Bitner et al.,

2010). First, human service employees use technology to support interactions with consumers.

In a support role, technology contributes by improving both the effectiveness and efficiency of

the employees in service encounters with their customers. For example, service employees use

technology to facilitate customer communication (e.g., email) and processing of historical

transactional information to improve and personalize interactions with consumers (e.g.,

spreadsheets and database management systems). This form of technology use, where

technology is instrumental to the fulfilment by service employees of service roles, is

characterized by high interpersonal focus and contact between service employees and

8
customers, and has been described as the high (human) touch – low tech paradigm (Verhagen

et al., 2014, Salomann et al., 2007, Bitner et al., 2000).

Second, service providers have also used technology to partially or completely replace

human service employees. Technology-based service (i.e., via self-service technology) occurs

when consumers use the technology independently, without the involvement of service

employees, to provide service for themselves. Technology-based service helps consumers

improve the effectiveness and efficiency of their own service encounter. Examples include bill

payment portals, online shopping websites, self-service kiosks at airports and checkouts at

department stores, and ATMs in banking. With greater infusion and use of technology,

customers can access services conveniently anytime, anywhere, without the complications of

interpersonal exchanges (e.g., bias, errors of service employees), but also in non-personal ways

(Bitner et al., 2000). This has been described as the low (human) touch – high tech paradigm

(Verhagen et al., 2014, Krehl, 2020, Bitner et al., 2000).

Extant service quality and scale development research has predominantly focused on

the high (human) touch – low tech and low (human) touch – high tech paradigms (Verhagen et

al., 2014). A plethora of scales have been developed to measure service quality for different

types of traditional, established software agents including human employees and self-service

technology operating in different contexts (e.g., Ladhari, 2010, Zeithaml et al., 2002, Ladhari,

2009, Seth et al., 2005).

Use of AISA in service encounters is associated with the emergence of a new high

(humanlike) touch – high tech paradigm, where the AI technology provides humanlike

interaction, that is, it offers scope for customers to interact independently with AISA which, in

turn, emulate human interaction including undertaking rational approaches to problem solving,

and relational communication (Russell and Norvig, 2018, Paschen et al., 2020, Overgoor et al.,

2019, Verhagen et al., 2014, Huang and Rust, 2021).

9
 In the high (humanlike) touch – high tech paradigm, the AISA are radically different to

traditional, established service agents in many ways (Lu et al., 2020, Wirtz et al., 2018). First,

AI can efficiently process and use large volumes of structured and unstructured data from

heterogeneous sources, including traditional CRM or ERP databases, email correspondence,

social media, websites, location-based advertisements, and data types including text, audio,

image and video (Dwivedi et al., 2021, Mogaji and Erkan, 2019, Paschen et al., 2020). Access

and the power to process vast amounts of data enable AISA to solve service problems

effectively and efficiently, often in ways superior to that of service employees (Bock et al.,

2020, Black and van Esch, 2020, Overgoor et al., 2019, Beck and Libert, 2017).

Second, AISA can accurately, and efficiently profile consumers based on a broader

range of criteria than traditional, established service agents can, which enhances service

providers’ understanding of customer preferences, consequentially creating the scope for

effectively achieving personalization outcomes and managing relationships with consumers

(Shankar, 2018). Third, AISA can also overcome the limitations of traditional human service

agents, such as human judgement bias, fatigue, errors, and availability (i.e., by providing

convenient service access to consumers anytime, anywhere) (Mogaji et al., 2020, Wirtz et al.,

2018, Huang and Rust, 2018).

Fourth, AISA possess analytical and cognitive capabilities which make them capable

of engaging consumers in a humanlike fashion (Wirtz et al., 2018, Davenport et al., 2020,

Davenport and Ronanki, 2018). AISA can recognize human emotions, feelings and sentiments,

but also provide socio-relational and -emotional responses resembling human emotions, such

as empathy and compassion, when communicating with consumers, thereby addressing

usability issues that are often associated with established technology-based service agents such

as SSTs (Troshani et al., 2020, Schniter et al., 2020, Bitner et al., 2002, Bitner et al., 2000, Lu

et al., 2020, Wirtz et al., 2018). Indeed, extant marketing literature shows that a functionalist

10
approach to service delivery is necessary but not sufficient for achieving service quality

(Bharadwaj et al., 1993). Consumers place a premium on the socio-relational aspects that offer

emotional and social value such as trust, engagement, and rapport (Stock and Merkle, 2018,

Wirtz et al., 2018). Early experimental research looking at virtual customer service agents, a

form of AISA, found evidence that humanlike emotional responses exhibited by AISA such as

friendliness, expertise, and smiling determine social presence and personalization which in turn

influence consumers’ satisfaction in the service encounter (Verhagen et al., 2014).

Fifth, AISA have the ability to learn quickly from past interactions and historical

information (Huang and Rust, 2018, Huang and Rust, 2021) which makes them capable of

managing interactions in service encounters with greater scope and complexity relative to the

specific, predefined interaction scenarios that are typically managed by service technologies

such as SSTs (Lin and Hsieh, 2011, Orel and Kara, 2014, Considine and Cormican, 2016). For

example, in their comparative analyses of SST and AISA, Wirtz et al. (2018) argue that SSTs

follow rigid interaction scripts which are generally ineffective to recover from errors and

service failure (e.g., Le et al., 2020); by contrast, AISA have flexible interaction scripts and

provide alternative solutions to consumers to recover from errors and failures.

For the purposes of this study, we define the high (humanlike) touch – high tech

paradigm on the basis of the AI technology development and service use frameworks discussed

in Wirtz et al. (2018) and Huang and Rust (2021). Specifically, the high (humanlike) touch –

high tech paradigm includes types of AISA that can manage cognitive and analytical tasks that

are associated with low socio-emotional and relational complexity (Wirtz et al., 2018). Huang

and Rust (2021) refer to these types of AISA as mechanical and thinking AI which are

predominantly used for routine transaction-based services and utilitarian, data-based predictive

services, respectively. We note that Huang and Rust (2021) also identify feeling AI which can

manage feeling services, that is, cognitive and analytical service tasks associated with high

11
socio-emotional and relational needs (e.g., high-risk healthcare service). Feeling AI, however,

remains in the early stages of development with feeling services requiring the natural

intelligence of human service agents who may rely on AISA for support (e.g., Wirtz et al.,

2018, Huang and Rust, 2021). Accordingly, for the purposes of this study, we exclude feeling

AI (Huang and Rust, 2021) from the high (humanlike) touch – high tech paradigm.

In this study we use chatbots and virtual assistants as two types of AISA in the high

(humanlike) touch – high tech paradigm. Both chatbots and virtual assistants are AISA that can

autonomously support consumers or users. Chatbots are used by business to offer pre-defined

services in response to customer queries (e.g., customer support) using mainly text-based,

spoken language interactions (Whang and Im, 2021, Naveen, 2018). Virtual assistants provide

general, customized, contextualized personal assistance services in response to user questions

or commands (e.g., personal management) using natural voice interfaces (Youn and Jin, 2021,

Kidd, 2019) (see Table I for additional detail).

[Insert Table I here]

In the sections that follow, we discuss the nature of the service quality construct, before

providing an overview of service quality scale development research focusing on scale

development for service encounters in the high touch – low tech and low touch – high tech

paradigms. We conclude this section with a brief discussion of early AISA research, including

the work of Noor et al. (2021b) and the proposed qualitative AISAQUAL dimensions. Taken

together, these discussions form the basis for our development of the AISAQUAL scale

focusing on chatbots and virtual assistants which will contribute to the emerging, but under-

researched high (humanlike) touch – high tech paradigm.

12
Nature of service quality

Service quality is generally considered to be a long-term global judgement of service

performance by consumers operationalized at an attitude level (Parasuraman et al., 1994a,

Cronin Jr and Taylor, 1994). There has been extensive discussion and rigorous validation of

the service quality dimensions in various service settings. While some scholars have treated

service quality dimensions as antecedents (e.g., Dabholkar, 1996, Dabholkar et al., 2000), the

majority conceptualize these dimensions as components of the multi-dimensional service

quality construct (Brady and Cronin Jr, 2001).

Meanwhile the debate continues as to whether the construct is reflective as suggested

in the majority of service quality scale development literature or whether it contains formative

higher orders (e.g., Ladhari, 2009, Ladhari, 2010, Martínez and Martínez, 2010, Parasuraman

et al., 2005). Many scholars however, have advised caution in considering the formative

specification (Hair et al., 2018, Caro and Garcia, 2008).

In this study, we argue that measures of AISA service quality constitute more of a

reflective view and are latent and not an index (Hair et al., 2018). This is because the

dimensions of the AISA service quality construct would be “expressions of the complexity”

(Caro and Garcia, 2008, p. 716) of AISA service performance as perceived by consumers.

Consistent with prior service quality measurement development studies, a series of service

quality factors (e.g., availability, anthropomorphism) contribute to the overall AISA service

quality perception and the change in any of the factors reflects the change in the overall service

quality perception (Collier and Bienstock, 2009).

Another issue that has been debated in the service marketing literature concerns the

empirical advantages of a performance-only assessment of service quality over the

performance-expectations comparison approach (Cronin Jr and Taylor, 1992, Cronin Jr and

Taylor, 1994, Parasuraman et al., 1993, Parasuraman et al., 1994b). In a longitudinal study,

13
Dabholkar et al. (2000) concluded that a performance-only measurement of service quality is

better than performance-expectation measurements, and is more suitable when the objective is

to determine factors contributing to service quality. Similarly, in a meta-analysis study of 17

empirical service quality studies spanning 17 years, Carrillat et al. (2007) found no significant

advantage to the use of performance-expectation indicators in determining overall service

quality.

In addition, they highlighted the advantage of the performance-only method requiring

half as many items as the performance-expectation approach. By contrast, the longer and more

time-consuming disconfirmation measurement method is more suitable for a gap analysis

(Dabholkar et al., 2000), and offers more in-depth diagnostics of service quality (Carrillat et

al., 2007, Parasuraman et al., 1993). As such, the perception-based measurement method was

adopted for our study as it is most suitable for our purpose of developing AISA service quality

scale dimensions (Dabholkar et al., 2000).

Prior research has reached the consensus that service quality perceptions often affect

service outcomes such as satisfaction, perceived value and loyalty (Zeithaml et al., 1996b,

Cronin Jr et al., 2000, Prentice et al., 2020). For example, in the context of self-service, it was

found that self-service quality drives customer satisfaction and loyalty in e-retailing (Ding et

al., 2011). Similarly, customers’ perception of self-service technology quality determines their

behavioral intentions (Lin and Hsieh, 2011).

Service quality scales: research developments and limitations

Since its early conceptualization by Grönroos (1984), many service quality scales,

predominantly grounded on the high (human) touch – low tech paradigm, have been proposed

to help understand service quality (Seth et al., 2005). Parasuraman et al. (1988)’s SERVQUAL

model established a scale to measure service quality (Parasuraman et al., 1991, Parasuraman

14
et al., 1988, Parasuraman et al., 1994a) and has gained much popularity in the service

marketing literature. This scale consists of five dimensions – tangibles, reliability,

responsiveness, assurance and empathy – on which consumers base their assessment of the

service quality of human service agents. The model has been validated and shown to be robust

across many service industries, including education, healthcare, insurance, hotel, library, bank

and retail services (Ladhari, 2009).

Since the development of SERVQUAL, technology has emerged and continued to play

a growing, integral role in service delivery. The nature of the role of technology, however,

began to substantially change from one where technology was used to support service

employees to one where technology is used by service providers to replace them. Periodic leaps

in technological innovation have resulted in the introduction of new technology-based SST

service agents (Rust, 2020). These technologies include self-service machines, such as bank

ATMs, vending machines (Fitzsimmons, 2003), and the internet with many online services

(Yang et al., 2004). Changes to the service environment due to technology-based service agents

with unique interface designs and service delivery processes have affected the nature of service

encounters including both the service and how service quality is perceived by consumers (Rust

and Oliver, 1993), consequentially bringing about a shift to the low (human) touch – high tech

service paradigm (Parasuraman, 2000, Bitner et al., 2000).

There are different combinations of service quality dimensions applicable to different

service environments, although some dimensions are more universal than others in terms of

their perceived role by consumers in service quality. For example, responsiveness appears

frequently as a key dimension for both human (Mittal and Lassar, 1996, Brady and Cronin Jr,

2001) and technology-based service agents (Ladhari, 2010). However, dimensions such as

security and privacy are more salient for technology-based services such as websites (e.g.,

Yang et al., 2004), mobile services (e.g., Huang et al., 2015) and self-service technologies (e.g.,

15
Lin and Hsieh, 2011). Accordingly, scholars have developed different service quality scales

for different contexts over the years including generic SSTQUAL scales for SSTs (Iqbal et al.,

2018, Considine and Cormican, 2016, Boon-itt, 2015, Lin and Hsieh, 2011, Ganguli and Roy,

2011). By contrast, other scales were designed more specifically to measure service quality for

certain applications and environments, such as online shopping websites (e.g., E-S-QUAL)

(Parasuraman et al., 2005), technology-mediated transactions processing (eTransQual) (Bauer

et al., 2006), e-retailing self-service (e-SELFQUAL) (Ding et al., 2011, Rita et al., 2019),

mobile service quality (MS-QUAL) (Huang et al., 2015), and telematics applications

(TeleServQ) (He et al., 2017).

Taken together, service quality scale studies develop and validate scales that are

anchored on high (human) touch – low tech or low (human) touch – high tech paradigms,

featuring interactions of customers with human service personnel and with different

technology-based SST service agents operating in different contexts and environments. This

suggests that available scales are bounded by the nature of the context and technology, which

makes their direct applicability to high (humanlike) touch – high tech settings questionable.

For example, some scholars have attempted to assess the applicability of existing, established

scales to AISA. For example, Morita et al. (2020) and Meyer-Waarden et al. (2020) found that

the SERVQUAL scale was not suitable for measuring AISA service quality. AISA’s inherent

idiosyncrasies and differences from traditional service agents are likely to provide novel

experiences and provoke unique reactions from consumers and consequentially also affect their

perceptions of service quality, thereby affecting the relevance of existing service quality scales

for AISA (Wirtz et al., 2018, Bock et al., 2020). This criticism is common of scale development

studies that are developed for specific contexts, using specific applications (Ladhari, 2009,

Ladhari, 2010).

16
Early AISA service quality research

There is limited research pertaining to AISA service quality. Two notable studies include

Prentice et al. (2020) and Noor et al. (2021b). Focusing on the hospitality and hotel industry

Prentice et al. (2020) adopt five constructs representing a suite of hotel services including

services from concierge robots, digital assistance, voice-activated services, travel-experience

enhancers and automatic data processing to measure AI service quality. The constructs and 15

items are sourced from Makadia (2018), without evidence of rigorous development and

validation.

By contrast, Noor et al. (2021b) propose 12 dimensions which they argue represent the

perceived service quality of AISA. Based on prior research into SERVQUAL dimensions

development, they adopt a two-stage approach inclusive of a comprehensive review of service

quality and information systems literatures followed by a qualitative validation stage. The

second stage aimed to both validate identified dimensions and identify new dimensions that

might be applicable to AISA service quality and included users of chatbots and virtual

assistants, as common, popular types of AISA in the high (humanlike) touch – high tech

paradigm. The 12 dimensions identified by Noor et al. (2021b) are summarized in Table II.

[Insert Table II here]

Of the 12 proposed dimensions representing the service quality of AISA, 10 are based

on scales developed to capture service quality in high (human) touch – low tech and in low

(human) touch – high tech settings. For instance, dimensions and items related to “assurance”

(Parasuraman et al., 1988, Burgers et al., 2000) and “privacy” (Huang et al., 2015) were

adapted to describe the degree of security that AISA is perceived to provide to consumers.

17
However, other dimensions such as attitude (of service employees) from Brady and Cronin Jr

(2001) were excluded as unsuitable.

Two of the proposed dimensions, namely proactiveness and anthropomorphism, were

new to the service quality literature. Specifically, proactiveness refers to the extent AISA can

anticipate or predict consumers’ future needs, and prompt service beyond what is explicitly

required by a customer (Noor et al., 2021b). Proactiveness is grounded in the customer service

literature on the construct of proactive behavior of employees which is characterized by

personal initiative criteria (Frese and Fay, 2001) including being “self-started, long-term-

oriented, and persistent behavior that is organizationally function and goal directed” (Rank et

al, 2007, p. 366). AISA can be said to be proactive when it manifests the capability of

anticipating and going beyond a straightforward reaction to commands, that is, alerting

consumers about tasks they may have overlooked or of which they may not be aware due to

limited familiarity with the service context (Rank et al., 2007), including assisting consumers

with advice concerned with alternative courses of action (Tan and Chou, 2008).

Anthropomorphism is another key characteristic distinguishing AISA from other non-

AI technology-based service agents. It describes the attribution of human capacities to non-

human agents (Troshani et al., 2020, Bartneck et al., 2009, Moussawi, 2016, Epley et al.,

2007). Prior research suggests that a consumer who interacts with AISA exhibiting

anthropomorphic capacities, including humanlike features, may develop perceptions of a social

presence that reduce privacy concerns (Benlian et al., 2020) and increase trust in the use of

AISA (Qiu and Benbasat, 2009, Troshani et al., 2020) and potentially enhance service quality

perceptions (Noor et al., 2021b). Consumers may also experience negative feelings when they

realize the service they have received, also known as “counterfeit service” was from AISA

rather than human service employees (Robinson et al., 2020). Additionally, increasing

18
anthropomorphism beyond a certain point is likely to lead to feelings of discomfort and

uneasiness (Duffy, 2003, Noor et al., 2021b, Troshani et al., 2020, Kim et al., 2019).

The current study builds on the work of Noor et al. (2021b). To the best of our

knowledge, Noor et al. (2021b) are the only available study making important inroads into

AISA scale development research by qualitative validation. Accordingly, we adopt their

proposed dimension (Table II) to develop AISAQUAL. We discuss the scale development

process that we adopted in the sections that follow.

Scale development

Consistent with the psychometric procedure for scale development advocated by marketing

scholars, our scale development process consists of domain definition, item generation, scale

development, and scale validation (Churchill Jr., 1979, Netemeyer et al., 2003). The first phase

involves defining the domain and phenomenon to be measured based on extant literature

followed by the generation of a pool of items through literature reviews, interviews, and

domain experts’ input, in conjunction with an assessment of content validity of the items. The

scale development step requires selecting and categorizing appropriate items to establish

desirable reliability and validity. The final step involves scale validation in which we evaluate

the scale to ensure desirable psychometric properties. The process culminates with six

AISAQUAL dimensions comprised of 26 items which, whilst grounded in extant related scale

validation and emerging AI literature, are also customized specifically for the AISA context.

We frame AISAs in the novel high (humanlike) touch – high tech paradigm and the domain in

which they operate is the broader area of service delivery where they trigger consumer service

quality perceptions. So the scale development process capitalizes on the rich base of service

quality literature and adaptation for AISA. A summary of the process we followed in the study

is shown in Figure 1. Each step is elaborated in the sections that follow.

19
 [Insert Figure 1 here]

Step 1: phenomenon and domain definition

Following Zeithaml (1988), we define AISAQUAL as the extent to which AISA facilitate an

overall perception of excellence or superiority as perceived by consumers. It is a global

assessment and attitude consisting of consumer judgements of AISA service performance. We

adopt the dimensions proposed by Noor et al. (2021b) for which we identify measure items at

the perceptual level to effectively capture the abstract nature of service quality comparisons

which consumers make across categories (Zeithaml, 1988). These service quality perceptions

can be formed through AISA usage, and in turn affect various outcomes such as customer

satisfaction and loyalty (Cronin Jr et al., 2000).

Following Noor et al. (2021b), we also use chatbots and virtual assistants in this study

as common forms of AISA used by service providers. We argue that this choice is appropriate

due to the popularity and wide availability of these types of AISA, which is important in

facilitating the recruitment of participants with experience in the use of these types of AISA,

therefore providing access to meaningful responses for the subsequent stages of scale

development and validation.

Step 2: item generation

To generate the item pool for Noor et al. (2021b)’s 12 conceptual dimensions, we adopted key

established scales from the literature. As shown in Figure 2, measure items from service quality

scales in human service contexts were found to generally capture half of the 12 conceptual

AISAQUAL dimensions (i.e., reliability, responsiveness, availability, aesthetics,

20
personalization and security), whereas item measures from technology-based service quality

scales were able to tap into all dimensions except proactiveness and anthropomorphism.

[Insert Figure 2 here]

As proactiveness and anthropomorphism represented two new dimensions (Noor et al.,

2021b), a logical way to analyze the quality of AISA service is to examine what proactiveness

and anthropomorphism entail which are mostly discussed in information and system quality

literature. Thus, we turned to non-service quality scale studies which contained similar

constructs that may capture these dimensions. Specifically, item measures for proactiveness

were adapted from Rank et al. (2007), whereas those for anthropomorphism were from

Bartneck et al. (2009), Han and Yang (2018) and Moussawi (2016). Further, based on (Noor

et al., 2021b)’s qualitative evidence, additional measure items were introduced to better capture

the intended dimensions (Netemeyer et al., 2003). For example, one of the items “the AISA

uses its own ‘judgment’ to complete a task” was added from (Noor et al., 2021b)’s findings.

After screening for irrelevant, redundant, ambiguous and double-barrelled statements, an initial

battery of 85 items was produced.

To assess the item statements, we recruited a panel of six senior expert academics with

an extensive publication track record in service and in emerging AI in business research. These

academics are active researchers based in top-ranked universities in Australia (3), New Zealand

(1), Singapore (1), and the USA (1). To maintain confidentiality, we do not disclose the identity

of the academics.

Through an online questionnaire sent to each expert via Qualtrics, item statements were

assessed for their representativeness (i.e., content validity) and appearance to be relevant (i.e.,

face validity) to the target construct (Netemeyer et al., 2003). The following scale was used:

21
1=“not representative”, 2=“somewhat representative”, 3=“clearly representative” (Bearden et

al., 2001). Only items which scored 2 (i.e., “somewhat representative”) or 3 (i.e., “clearly

representative”) by at least 80 percent of the panel were retained in the item pool (Lin and

Hsieh, 2011). We also reviewed the qualitative suggestions that experts made to improve item

wording for inclusion in the item pool. In addition, we assessed the remaining items for

redundancy and added measures to ensure a sufficient item pool per dimension in the

subsequent scale refinement process (Netemeyer et al., 2003). This content and face validation

process reduced the number of items from 85 to 75.

Step 3: scale refinement

The next phase of scale development involved the testing of the preliminary 75-item

AISAQUAL scale. A self-administered questionnaire was constructed and consisted of two

sections. The first section contained demographic and AISA usage questions. The second

section contained statements for 75 items to measure respondents’ perceptions of AISA

performance. Items for the second section were measured using a seven-point Likert scale

anchored from 1 = strongly disagree to 7 = strongly agree, which is an established practice for

scale reliability and validity (Netemeyer et al., 2003).

Surveys were distributed by the online panel company Qualtrics using purposive

sampling to users of chatbots and virtual assistants. A condition of participation in the survey

was to have used these service agents in the three months prior to the survey. We asked

respondents to choose the AISA type (i.e., chatbots or virtual assistants) with which they were

most familiar. To clarify our definitional differences between chatbots and virtual assistants

and establish a common language to ensure applicability of responses, examples of each type

of AISA were provided in the survey introduction, including illustrating images, basic

definitions, and textual descriptions of common uses. A US sample was deliberately used since

22
the US represents one of the top 10 countries with a significant number of AISA users (PwC,

2018), and is expected to continue to occupy the largest global market share of chatbots and

virtual assistants (AMR, 2020).

The sample consisted of 211 respondents with an almost even gender split within each

category of chatbot (male=50.5%, female=49.5%) and virtual assistant (male=50.9%,

female=49.1%) users. The sample size of 211 complied with requirements of approximately

200 for an initial test stage of a new scale (Clark and Watson, 1995, Parasuraman et al., 1988).

Almost two-thirds of respondents (67.3%) were aged 25 to 44. Table III summarizes the

profiles of respondents for the scale refinement phase.

[Insert Table III here]

Exploratory Factor Analysis

As recommended by Churchill Jr. (1979), to better prepare the core items for the exploratory

factor analysis, we first categorized the 75 item measures into the 12 a priori conceptual

dimensions of AISA service quality before examining the reliability Cronbach Alpha score for

each dimension. We then inspected items with low individual reliabilities (<.50) (e.g., Bagozzi

and Yi, 1988) and removed one from the security dimension and the other from

anthropomorphism.

Next, a principal components analysis with oblimin rotation (Kaiser normalization) was

conducted on the remaining 73 items to empirically identify the underlying dimensions.

Consistent with extant service quality scales (e.g., Parasuraman et al., 1988, Parasuraman et

al., 2005, Ding et al., 2011), we used the oblique rotation – oblimin – to allow for correlations

between factors in order to obtain interpretable components. We note that several factor

23
correlations after oblique rotation exceeded the suggested threshold of 0.32 by Tabachnick et

al. (2007).

The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy (.95 > .50) and

Bartlett’s sphericity test (p < .001) were significant and indicated the suitability of using

exploratory factor analysis for our data (Field, 2013). To determine the number of components,

the Kaiser criterion of eigenvalues greater than one was used (Netemeyer et al., 2003). Next,

items were dropped using a minimum loading criterion of .40 (Ford et al., 1986) or those

exhibiting cross-loadings over .40 on two or more components (Hair et al., 2018). Remaining

items were again subjected to factor analysis.

The above process was done iteratively using SPSS 25 till all items and dimensions

satisfied the required minimum thresholds. After six extractions, 34 items remained and loaded

distinctly onto six dimensions D1 to D6 as shown in Table IV. The six components accounted

for 66.9% of the variance (Hair et al., 2018) and indicated good internal consistency among

items with reliability coefficients ranging from .85 to .90 (Nunnally, 1978).

[Insert Table IV here]

As shown in Table IV, dimensions D2, D4 and D5 remained reflective of the three

original dimensions of security, enjoyment and contact respectively. An exception was the

security item SEC1 “A clear privacy policy is accessible before I use the AISA” which loaded

with contact items in D5. Conceptually, consumers may relate the availability of such a privacy

policy as originating from contact with a human service professional. Removing SEC1 would

also marginally lower the coefficient alpha of D5. Thus, at this stage, SEC1 was retained with

the other contact items in D5.

24
 The remaining nine dimensions from the original 12 were collapsed into three. D1 –

containing several items for reliability, responsiveness, aesthetics and control – was found to

be similar in its conceptual item composition to the dimension of efficiency in E-S-QUAL

(Parasuraman et al., 2005) and functionality in SSTQUAL (Lin and Hsieh, 2011). D3 – mostly

containing items for availability – included an original responsiveness item RES3 which can

be conceptually related to the perceived availability of AISA to launch quickly when required.

In relation to the loading of an aesthetics item AES1 in D3, conceptually an innovative interface

design may signal the competency of AISA as being capable of service anytime. Similar to

SEC1, the removal of RES3 and AES1 would lower the coefficient alpha of D3. Thus, we kept

these items for further empirical scrutiny in the subsequent CFA stage. Finally, D6 contained

items related to anthropomorphism, personalization and proactiveness. This composition of

human and intelligence performance traits expected by AISA consumers is significant given

that AISA is a non-human service agent and past scales involving technology service agents

have not featured a service quality dimension of a similar nature.

Confirmatory Factor Analysis

A confirmatory factor analysis (CFA) of the 34-item, six-dimension model was next conducted

to further verify the model. A sample of 275 chatbot and virtual assistant users was used to

conduct this analysis. Just over half of the respondents are female (52%). The majority of the

respondents are from the age group of 25-34 (33.0%) and 35-44 (33.4%). The profile of the

sample can be found in Table III.

Using indices recommended by Bagozzi and Yi (1988), initial CFA results indicated a

significant chi-square value (χ2 = 1077.43, p < .001); with a Root Mean Square Error of

Approximation (RMSEA) = .09 (recommended RMSEA ≤ .07), Tucker and Lewis Index (TLI)

25
= .91 (recommended TLI ≥ .90), Comparative Fit Index (CFI) = .87 (recommended CFI ≥ .90)

and the Standardized Root Mean Square Residual (SRMR) = .10 (recommended SRMR ≤ .08).

To improve the model fit, we first looked at the item-to-factor loadings and removed

items with loading values below .70 (Hair et al., 2018). Two iterations removed items CTL3,

SEC1, PRO5 and RES2 which had item-to-factor loading values of .69, .65, .52 and .69,

respectively. Next, we inspected the standardized residual covariance matrices. Although the

standardized residual values were less than 2.5, there were observable patterns of fairly large

standardized residual loadings across several variables that were worthy of closer inspection

(Hair et al., 2018). Accordingly, we removed residuals greater than 2 (Anderson and Gerbing,

1988, Bagozzi and Yi, 1988). This iterative process deleted items AES1, REL7, SEC5 and

RES3, and resulted in an acceptable model fit (Bagozzi and Yi, 1988). The final confirmatory

model contained six factors and 26 items (see Table IV for items in bold) with values of χ2 =

835.01, p < .001; RMSEA = .07 (≤ .07), TLI = .91 (≥ .90), CFI = .90 (≥ .90) and SRMR = .08

(≤ .08). Based on the content of the items in each dimension, six labels and definitions were

chosen below. Of these, we kept five from the initial 12 conceptual dimensions with efficiency

combining several attributes.

1. Efficiency: AISA’s ease of use and speed.

2. Security: Perceived safety of the AISA from intrusion of privacy, fraud and loss of
personal information.
3. Availability: the extent to which AISA is ready for use anywhere, anytime.
4. Enjoyment: Extent to which using the AISA is perceived to be enjoyable regardless of
consequential performance.
5. Contact: Access to human assistance is available via AISA.
6. Anthropomorphism: AISA display human-like characteristics, motivations, intentions,
or emotions.

In comparison with existing human service agents’ service quality scales, the

dimensions of security and availability are found in both human- as well as technology-based

26
service quality scales. The dimensions of efficiency, enjoyment and contact are not found in

human-based service quality scales but instead in extant technology-based service quality

scales. The remaining anthropomorphism dimension represents a new service quality

dimension unique to AISA service quality.

Step 4: scale validation

Additional empirical research was conducted to confirm the reliability and validity of the 26-

item AISAQUAL scale. A self-administered questionnaire includes the 26 AISAQUAL items

and service outcome variables. The new sample consisted of 304 respondents which was larger

than the pilot sample size (n=211) for the scale refinement phase and satisfied the requirements

for scale validation (Clark and Watson, 1995). The sample profile is similar to the sample used

in the scale refinement phase (MacKenzie et al., 2011) and consists of US residents who had

used chatbots and virtual assistants in the three months prior to the survey. There was an almost

even gender split within each category of chatbot (male=50.3%, female=49.7%) and virtual

assistant (male=49.7%, female=50.3%) users. Almost an equal portion used their AISA on a

daily (31.3%) or weekly (31.9%) basis. In addition, the majority of respondents (47.4%) had

used their AISA for two to three years. In terms of usage context, most respondents used their

chatbots for services related to the retail trade (19.9%). This differs from those in the scale

refinement phase who predominantly interacted with chatbots for electricity, gas and waste

services (20.0%). For virtual assistants, similar to respondents from our scale refinement phase,

Google Assistant, Alexa and Siri were the most popular. This is also representative of the

overall US customer adoption for virtual assistants (Olson and Kemery, 2019). Table V

summarizes the profiles of respondents for the scale validation phase.

[Insert Table V here]

27
 The descriptive statistics of the dimensions have been provided in Table VI. In addition,

we conducted t-tests to detect if the mean scores of the dimensions differ between virtual

assistants and chatbots. The results suggest invariance of AISAQUAL dimensions across the

two types of AISA. To assess the proposed scale’s construct validity, we deployed several

methods. First, we conducted a CFA (χ2 = 562.94, p < .001; RMSEA = .06, TLI = .94, CFI =

.95 and SRMR = .04). All coefficient alphas are above the .70 level (Hair et al., 2018) except

for item EFF1 “The AISA works correctly at first attempt” which had a loading of .69. Upon

inspection, EFF1 contributed to the content validity of its latent variable and its removal did

not result in an increase in the composite reliability (CR) score of the efficiency dimension (see

Table VI) (Hair et al., 2011). Its item loading also fell within the acceptable range of between

.50 to .90 (Bagozzi and Yi, 1988) and was not below the absolute threshold value of .50 (Hair

et al., 2018). Thus, EFF1 displayed sufficient indicator reliability and was retained in the scale.

To assess convergent validity, the composite reliability (CR) of all six dimensions of

AISAQUAL was found to be between .80 and .90 which is above the recommended value of

.70 (Bagozzi and Yi, 1988). In addition, the average variance extracted (AVE) of the six factors

also ranged from .57 to .68 which is above the recommended level of .50 (Bagozzi and Yi,

1988), indicating high levels of convergence among the items in measuring their respective

constructs.

[Insert Table VI here]

Following the recommendations of Voorhees et al. (2016), the Hetero-Trait Mono-Trait

(HTMT) Ratio of the correlations was used to test for discriminant validity (Henseler et al.,

2015). All ratios were found to meet the conservative cut-off of .85 (Hair et al., 2018) except

for those between Efficiency and Enjoyment and between Efficiency and Anthropomorphism

28
(see Table VII). Their correlation ratio of .86 was within the acceptable threshold of .90 (Zheng

et al., 2020) and supported for conceptually similar constructs (Hair et al., 2018). In addition,

we examined the discriminant validity by comparing the correlation between the dimensions

and the AVE of each dimension (Fornell and Larcker, 1981). As shown in Table VIII, the

square roots of AVE ranges between .75 and .82, exceeding all the correlations squared

coefficients. This shows that AISAQUAL has satisfactory discriminant validity (Boudreau et

al., 2001).

[Insert Table VII here]

[Insert Table VIII here]

To assess the nomological validity of AISAQUAL, we test the relationship between the

six dimensions of AISAQUAL and three theoretically related variables of customer

satisfaction, perceived value and customer loyalty. This method is used to demonstrate the

proposed scale’s practical value and indicates its ability to explain and predict other dependent

variables (Arnold and Reynolds, 2003). Service quality has been shown to affect customer

satisfaction (Ding et al., 2011, Caruana, 2002) and perceived value (Parasuraman et al., 2005),

and that all three constructs work together to affect the behavioral outcome of customer loyalty

(Cronin Jr et al., 2000, Oh, 1999). To assess these relationships in the AISA context, five

loyalty intention items were adapted from Zeithaml et al. (1996a), three customer satisfaction

items from Bodet (2008), and two perceived value items from Tam (2004) and Cronin Jr et al.

(2000). AISAQUAL was modelled as an exogenous variable by aggregating its six dimensions

into six indicators using the average score of items per dimension (Ding et al., 2011, Lin and

Hsieh, 2011).

29
 The structural model illustrated in Figure 3 shows a good model fit (χ2 = 260.87, p <

.001; RMSEA = .07, TLI = .96, CFI = .97 and SRMR = .03). All loadings of these paths were

found to be significant at p < .001. The effect of AISAQUAL was strongest on perceived value

(β=.82, p < .001) followed by satisfaction (β=.34, p < .001) and loyalty intentions (β=.17, p <

.001). These path-strength patterns echo findings from extant research in both services

marketing (e.g., Cronin Jr et al., 2000) and IS (e.g., Kuo et al., 2009). In addition, the results

indicate that all extracted service quality dimensions have significant effects on customer

satisfaction, perceived value and customer loyalty. AISA service quality accounts for 41% of

the variance in customer satisfaction, 38% of the variance in perceived value, and 23% of the

variance in customer loyalty, indicating good external validity. This pattern of evidence

suggests that AISAQUAL demonstrates nomological validity.

[Insert Figure 3 here]

In addition to our theoretical reasoning regarding AISAQUAL’s reflective model

structure, we empirically tested our model specification using a confirmatory tetrad analysis

(CTA-PLS) (Hair et al., 2019). The CTA-PLS has been used in recent marketing studies (e.g.,

Nath, 2020) to assess if the difference between pairs of covariances among construct indicators

(i.e. tetrads) is significantly different from zero which would indicate a formative construct. A

reflective model would result in tetrads having a value of zero. Using the recommendations

outlined by Bollen and Ting (2000) and Gudergan et al. (2008), we find that all tetrad results

were non-significant (i.e. confidence intervals include zero), providing empirical support for

our view that AISAQUAL is a reflective construct.

Finally, we conducted a multi-group analysis (Henseler, 2012, Henseler et al., 2009) to

check if the proposed scale varies across the two types of AISA. Results for the difference

30
between group-specific path coefficients are non-significant at the 5 percent probability of error

level (see Table IX), suggesting that AISAQUAL demonstrates sufficient invariance across

chatbots and virtual assistants.

[Insert Table IX here]

Concluding discussion and implications

Theoretical implications

The growing research interest in services enabled by AISA and the continued use of AISA in

service sectors makes urgent the development of a suitable AISA service quality scale (Lu et

al., 2020, Bock et al., 2020). Our study is a direct response to this need. Specifically, our

development of AISAQUAL appears to be the first AISA service quality scale for AI-based

applications in the emergent high (humanlike) touch – high tech paradigm, validated via

rigorous, established scale validation processes using two popular, common AISA, namely

chatbots and virtual assistants. Consisting of six dimensions and 26 item measures,

AISAQUAL fills an important gap and extends current understanding of consumer service

quality evaluations for different AISA service environments using validated and generalizable

scale instruments.

Our findings suggest that anthropomorphism is a key dimension driving AISA service

quality (β=.89, p < .001), thereby providing support for the emerging research underscoring

the importance of further considering anthropomorphism in the improvement of user

experiences with AISA (Wirtz et al., 2018, Benlian et al., 2020, Troshani et al., 2020, Sheehan

et al., 2020).

Our findings also suggest that the hedonic element of AISA, i.e., enjoyment, is

important to consumers in the evaluation of service quality (β=.87, p < .001). This is consistent

31
with the findings of Lin and Hsieh (2011) and supports the role enjoyment plays in service

quality evaluations of AISA beyond factors that help fulfil customers’ utilitarian needs (Davis

et al., 1992). Importantly, we also identified efficiency as an important dimension of

AISAQUAL (β=.91, p < .001) which is consistent with and further strengthens the view that

AISA provide utilitarian value to consumers (Meyer-Waarden et al., 2020, Noor et al., 2021b).

The presence of both hedonic and utilitarian dimensions also support the strategic importance

of AISA in optimising quality and relationships in its service of consumers (Huang and Rust,

2021).

Taken together, the AISAQUAL dimensions capture what consumers perceive to be

important when using services provided by AISA. Prentice et al. (2020) have argued that the

disconnect, i.e., customer gap, between consumers’ AISA service needs and expectations and

their perceptions of the service quality actually delivered by AISA can explain the findings in

research that consumers often prefer human service employees to AISA. Our AISAQUAL

scale extends current understanding of the key factors that might contribute towards the

customer gap in the context of AISA-based services.

Additionally, these findings add to the emerging research stream of AISA and validate

relationships between service quality and customer satisfaction, perceived value and loyalty

intentions in the AISA context. The findings of the current study suggest how the service

quality of AISA can be determined, and as such will facilitate further theory development

through the use of AISAQUAL (Ranjan and Read, 2016). This is important as it extends the

link between service quality perception and positive service outcomes in the AISA context.

Managerial implications

The growing popularity of novel AI-based applications such as AISA (Davenport and Ronanki,

2018) increases the onus on service providers to effectively design AISA that customers will

32
use for service. Our contribution of the AISAQUAL scale has practical and managerial

relevance. It helps to improve service providers’ understanding of consumer expectations when

using AISA. The factors may also be useful to inform further efforts of service providers to

improve AISA design, putting them in a stronger position to ensure that the manner in which

the AISA they use to provide service are consistent with consumers’ service quality

expectations (Prentice et al., 2020).

For example, given the significance of the anthropomorphism and enjoyment

dimensions, companies should take these factors into consideration in the design phase of their

AISA interface with their customers. One way to do this is to test different designs and

interaction modes (e.g., speech versus gesture) of the AISA with their target audience at

multiple stages (Kepuska and Bohouta, 2018). Our study also highlights the prevailing

relevance of contact with human service agents in the eyes of consumers. Specifically,

managers should ensure that human service agents are an available option for consumers during

their AISA interaction (Shell and Buell, 2019). On AISA security and governance (Shepardson,

2020), companies need to foster greater trust and transparency with users (Bandara et al., 2020)

by being forthcoming with regards to their data privacy and protection policies as they continue

to access increasing personal data through AISA. Finally, developers can increase the

availability of their AISA within the wider service ecosystem as device interconnectivity

matures through the Internet of Things (IoT) (Huang and Rust, 2017).

Consistent with prior service quality research, AISAQUAL validates the importance of

service quality for AISA consumers as this can lead to perceived value, customer satisfaction

and loyalty. AISAQUAL can thus serve as a diagnostic tool to improve current AISA service

performance. As the AISAQUAL measurement is parsimonious, it helps service managers to

better understand customer perceptions and address service quality concerns in a systematic

way.

33
Limitations and future research

Our study contributes to the rich service quality literature by developing a robust service quality

scale with good psychometric properties to accommodate the new AISA service environment.

However, as with any scale development study, several caveats should be noted which also

represent opportunities for further research.

First, AISAQUAL is developed as both a first and second order reflective construct

based on underlying theoretical considerations (MacKenzie et al., 2011) and empirical testing.

While we have established this position in our study, future research can explore the

implications of an alternative formative model (Collier and Bienstock, 2009, Theodosiou et al.,

2019) which requires additional reflective indicators to be tested against the proposed

formative constructs (Jarvis et al., 2003).

Second, at the time of this study, the AISA types most widely used by consumers are

chatbots and virtual assistants. Accordingly, these software applications were used as suitable

representative technology types for the construction of our AISAQUAL scale. An inspection

of the measure items of the six dimensions of AISAQUAL suggests that the item measures

might also be applicable to other AISA types including social robots (Noor et al., 2021a). We

recommend that the AISAQUAL be adapted and tested for these other AISA types, which have

more physical features than chatbots and virtual assistants, as they become available in the

market.

Third, as previously indicated, anthropomorphism is a key dimension in our

AISAQUAL scale. Our evidence clearly suggests the important role that anthropomorphism

plays in consumers’ AISA service quality perceptions and evaluations. Whilst existing

literature has associated anthropomorphism of AI with the uncanny valley phenomenon, we did

34
not find evidence to indicate existence of this phenomenon in the context of AISA.

AISAQUAL indicates that consumers perceive anthropomorphism as a favorable dimension

rather than one which may cause uneasiness when interacting with AISA. Our findings both

raise questions concerning the extent of applicability of uncanny valley to AISA applications,

and also underscore the need for further research to enhance the context- and application-

specific understanding of anthropomorphism.

Fourth, the focus of this study was to produce AISAQUAL, a service quality scale for

improving current understanding of the key factors that influence consumers’ perception of

AISA service quality. The scale does not differentiate between customers’ AISA service

expectations and perceptions. Accordingly, further research is needed to identify the key

factors that influence the AISA customer gap (Bitner et al., 2010) including specific

measurements of what consumers expect for services provided by AISA and their perceptions

pertaining to services that are delivered by AISA.

Fifth, key steps in the scale development and validation process were carried out by

using respondents that were recruited via the Qualtrics platform. This platform has become

increasingly popular in recent years for respondent recruitment in research. Further research

may further validate our findings across other types of platforms.

Finally, the effects of AISAQUAL on the outcome variables used in this study was

based on a cross-sectional view. Additionally, given our focus on scale development, we have

not considered moderating factors. As AISA are expected to serve consumers in the long run,

understanding possible shifts in attitudes over time (Hussain et al., 2019) and whether these

may lead to more positive or negative outcomes is worth further investigation. Hence, we

recommend future longitudinal studies assess how the ongoing service performance of AISA

can change consumer outcomes including consideration of moderating factors in relation to the

AISA application (e.g., the form and type), the service (e.g. types, criticality), and the customer

35
(e.g., demographics, cultural values, usage situation, orientation, and technology savviness and

proneness).

References

AMR (2020) Intelligent Virtual Assistant (IVA) Market to grow at 33% CAGR during forecast
period (2020-2025) - Insights on Growth Drivers, Size and Share Analysis, Key Trends,
Leading Players, and Business Opportunities: Adroit Market Research. Available:
https://www.globenewswire.com/news-release/2020/02/24/1988963/0/en/Intelligent-
Virtual-Assistant-IVA-Market-to-grow-at-33-CAGR-during-forecast-period-2020-
2025-Insights-on-Growth-Drivers-Size-and-Share-Analysis-Key-Trends-Leading-
Players-and-Busin.html. Accessed April 19th, 2021. Dallas, TX: Adroit Market
Research.
Anderson, J.C. and Gerbing, D.W. (1988), "Structural equation modeling in practice: A review
and recommended two-step approach". Psychological Bulletin, Vol. 103 No. 3, pp. 411.
Arnold, M.J. and Reynolds, K.E. (2003), "Hedonic shopping motivations". Journal of
Retailing, Vol. 79 No. 2, pp. 77-95.
Bagozzi, R.P. and Yi, Y. (1988), "On the evaluation of structural equation models". Journal of
the academy of marketing science, Vol. 16 No. 1, pp. 74-94.
Bandara, R., Fernando, M. and Akter, S. (2020), "Managing consumer privacy concerns and
defensive behaviours in the digital marketplace". European Journal of Marketing, Vol.
55 No. 1, pp. 219-246.
Bartneck, C. et al. (2009), "Measurement instruments for the anthropomorphism, animacy,
likeability, perceived intelligence, and perceived safety of robots". International
journal of social robotics, Vol. 1 No. 1, pp. 71-81.
Bauer, H.H., Falk, T. and Hammerschmidt, M. (2006), "eTransQual: A transaction process-
based approach for capturing service quality in online shopping". Journal of Business
Research, Vol. 59 No. 7, pp. 866-875.
Bearden, W.O., Hardesty, D.M. and Rose, R.L. (2001), "Consumer self-confidence:
Refinements in conceptualisation and measurement". Journal of Consumer Research,
Vol. 28 No. 1, pp. 121-134.
Beck, M. and Libert, B. (2017), "The rise of AI makes emotional intelligence more important.
Available: https://hbr.org/2017/02/the-rise-of-ai-makes-emotional-intelligence-more-
important. Accessed April 20th, 2021". Harvard Business Review, Vol. February No.
Benlian, A., Klumpe, J. and Hinz, O. (2020), "Mitigating the intrusive effects of smart home
assistants by using anthropomorphic design features: A multimethod investigation".
Information Systems Journal, Vol. 30 No. 6, pp. 1010-1042.
Bharadwaj, S.G., Varadarajan, P.R. and Faly, J. (1993), "Sustainable competitive advantage in
service industries: a conceptual model and research propositions". Journal of
Marketing, Vol. 57 No. 4, pp. 83-99.
Bitner, M.J., Brown, S.W. and Meuter, M.L. (2000), "Technology infusion in service
encounters". Journal of the Academy of marketing Science, Vol. 28 No. 1, pp. 138-149.
Bitner, M.J., Ostrom, A.L. and Meuter, M.L. (2002), "Implementing successful self-service
technologies". Academy of Management Executive, Vol. 16 No. 4, pp. 96-109.
Bitner, M.J., Zeithaml, V.A. and Gremler, D.D. (2010) 'Technology’s impact on the gaps
model of service quality'. In: P., M., Kieliszewski C. and J., S. (eds.) Handbook of

36
 Service Science: Research and Innovations in the Service Economy. Boston, MA:
Springer, pp. 197-218.
Black, J.S. and van Esch, P. (2020), "AI-enabled recruiting: what is it and how should a
manager use it". Business Horizons, Vol. 63 No. 2, pp. 215-226.
Bock, D.E., Wolter, J.S. and Ferrell, O. (2020), "Artificial intelligence: disrupting what we
know about services". Journal of Services Marketing, DOI 10.1108/JSM-1101-2019-
0047.
Bodet, G. (2008), "Customer satisfaction and loyalty in service: Two concepts, four constructs,
several relationships". Journal of Retailing and Consumer Services, Vol. 15 No. 3, pp.
156-162.
Bollen, K.A. and Ting, K.F. (2000), "A Tetrad Test for Causal Indicators". Psychological
Methods, Vol. 5 pp. 3-22.
Boon-itt, S. (2015), "Managing self-service technology service quality to enhance e-
satisfaction". International Journal of Quality and Service Sciences, Vol. 7 No. 4, pp.
373-391.
Boudreau, M.-C., Gefen, D. and Straub, D.W. (2001), "Validation in information systems
research: A state-of-the-art assessment". MIS Quarterly, Vol. 25 No. 1, pp. 1-16.
Brady, M.K. and Cronin Jr, J.J. (2001), "Some new thoughts on conceptualizing perceived
service quality: a hierarchical approach". Journal of marketing, Vol. 65 No. 3, pp. 34-
49.
Burgers, A. et al. (2000), "Customer expectation dimensions of voice‐to‐voice service
encounters: a scale‐development study". International Journal of Service Industry
Management, Vol. 11 No. 2, pp. 142-161.
Caro, L.M. and Garcia, J.A.M. (2008), "Developing a multidimensional and hierarchical
service quality model for the travel agency industry". Tourism Management, Vol. 29
No. 4, pp. 706-720.
Carrillat, F.A., Jaramillo, F. and Mulki, J.P. (2007), "The validity of the SERVQUAL and
SERVPERF scales: A meta-analytic view of 17 years of research across five
continents". International Journal of Service Industry Management, Vol. 18 No. 5, pp.
472-490.
Caruana, A. (2002), "Service loyalty: The effects of service quality and the mediating role of
customer satisfaction". European Journal of Marketing, Vol. 36 No. 7/8, pp. 811-828.
Chi, C. (2021) 12 of the Best AI Chatbots for 2021. Available:
https://blog.hubspot.com/marketing/best-ai-chatbot. Accessed April 12th, 2021.
Chui, M. et al. (2020) The state of AI in 2020. Available: https://www.mckinsey.com/business-
functions/mckinsey-analytics/our-insights/global-survey-the-state-of-ai-in-2020#.
Accessed April 21st, 2021. San Francisco: McKinsey & Company.
Churchill Jr., G.A. (1979), "A paradigm for developing better measures of marketing
constructs". Journal of Marketing Research, Vol. 16 No. 1, pp. 64-73.
Clark, L.A. and Watson, D. (1995), "Constructing validity: Basic issues in objective scale
development". Psychological Assessment, Vol. 7 No. 3, pp. 309-319.
Colback, L. (2020) The impact of AI on business and society. Available:
https://www.ft.com/content/e082b01d-fbd6-4ea5-a0d2-05bc5ad7176c. Accessed:
April 14th, 2021. Financial Times.
Collier, J.E. and Bienstock, C.C. (2009), "Model misspecification: contrasting formative and
reflective indicators for a model of e-service quality". Journal of Marketing Theory and
Practice, Vol. 17 No. 3, pp. 283-293.
Considine, E. and Cormican, K. (2016), "Self-service technology adoption: an analysis of
customer to technology interactions". Procedia Computer Science, Vol. 100 pp. 103-
109.

37
Cronin Jr, J.J., Brady, M.K. and Hult, G.T.M. (2000), "Assessing the effects of quality, value,
and customer satisfaction on consumer behavioral intentions in service environments".
Journal of Retailing, Vol. 76 No. 2, pp. 193-218.
Cronin Jr, J.J. and Taylor, S.A. (1992), "Measuring service quality: a reexamination and
extension". Journal of marketing, Vol. 56 No. 3, pp. 55-68.
Cronin Jr, J.J. and Taylor, S.A. (1994), "SERVPERF versus SERVQUAL: reconciling
performance-based and perceptions-minus-expectations measurement of service
quality". Journal of marketing, Vol. 58 No. 1, pp. 125-131.
Dabholkar, P. (1996), "Consumer evaluations of new technology-based self-service options:
An investigation of alternative modes of service quality". International Journal of
Research in Marketing, Vol. 13 No. 1, pp. 29-51.
Dabholkar, P.A., Shepherd, C.D. and Thorpe, D.I. (2000), "A comprehensive framework for
service quality: an investigation of critical conceptual and measurement issues through
a longitudinal study". Journal of Retailing, Vol. 76 No. 2, pp. 139-173.
Davenport, T. et al. (2020), "How artificial intelligence will change the future of marketing".
Journal of the Academy of Marketing Science, Vol. 48 No. 1, pp. 24-42.
Davenport, T. and Ronanki, R. (2018), "Artificial intelligence for the real world". Harvard
Business Review, Vol. 96 No. 1, pp. 108-116.
Davis, F.D. (1989), "Perceived usefulness, perceived ease of use, and user acceptance in
information technology". MIS Quarterly, Vol. 13 No. 3, pp. 319-339.
Davis, F.D., Bagozzi, R.P. and Warshaw, P.R. (1992), "Extrinsic and intrinsic motivation to
use computers in the workplace". Journal of Applied Social Psychology, Vol. 22 No.
14, pp. 1111-1132.
De Keyser, A. et al. (2019), "Frontline service technology infusion: conceptual archetypes and
future research directions". Journal of Service Management, Vol. No.
Ding, D.X., Hu, P.J.-H. and Sheng, O.R.L. (2011), "e-SELFQUAL: A scale for measuring
online self-service quality". Journal of Business Research, Vol. 64 No. 5, pp. 508-515.
Duffy, B.R. (2003), "Anthropomorphism and the social robot". Robotics and autonomous
systems, Vol. 42 No. 3-4, pp. 177-190.
Dwivedi, Y.K. et al. (2021), "Artificial Intelligence (AI): Multidisciplinary perspectives on
emerging challenges, opportunities, and agenda for research, practice and policy".
International Journal of Information Management, Vol. 57 No. Article ID 101994.
Elliot, C. (2018) Chatbots Are Killing Customer Service. Here's Why. Available:
https://www.forbes.com/sites/christopherelliott/2018/08/27/chatbots-are-killing-
customer-service-heres-why/?sh=27678f2a13c5. Accessed April 16th, 2021. Forbes.
Epley, N., Waytz, A. and Cacioppo, J.T. (2007), "On seeing human: a three-factor theory of
anthropomorphism". Psychological review, Vol. 114 No. 4, pp. 864.
Fassnacht, M. and Koese, I. (2006), "Quality of electronic services: conceptualizing and testing
a hierarchical model". Journal of Service Research, Vol. 9 No. 1, pp. 19-37.
Field, A. (2013) Discovering statistics using IBM SPSS statistics. London: Sage.
Fitzsimmons, J.A. (2003), "Is self‐service the future of services?". Managing Service Quality:
An International Journal, Vol. 13 No. 6, pp. 443-444.
Ford, J.K., MacCallum, R.C. and Tait, M. (1986), "The application of exploratory factor
analysis in applied psychology: A critical review and analysis". Personnel Psychology,
Vol. 39 No. 2, pp. 291-314.
Fornell, C. and Larcker, D.F. (1981), "Structural equation models with unobservable variables
and measurement error: Algebra and statistics". Sage Publications Sage CA: Los
Angeles, CA.
Fountaine, T., McCarthy, B. and Saleh, T. (2021) Reimagining your business for AI. Available:
https://www.mckinsey.com/business-functions/mckinsey-analytics/our-

38
 insights/reimagining-your-business-for-ai#. Accessed April 17, 2021. McKinsey &
Company.
Frese, M. and Fay, D. (2001), "4. Personal initiative: An active performance concept for work
in the 21st century". Research in organizational behavior, Vol. 23 No. 133-187.
Ganguli, S. and Roy, S. (2011), "Generic technology-based service quality dimensions in
banking: impact on customer satisfaction and loyalty.". International Journal of Bank
Marketing, Vol. 29 No. 2, pp. 168-189.
Grewal, D. and Levy, M. (2009), "Emerging issues in retailing research". Journal of retailing,
Vol. 85 No. 4, pp. 522-526.
Grönroos, C. (1984), "A service quality model and its marketing implications". European
Journal of Marketing, Vol. 18 No. 4, pp. 36-44.
Gudergan, S.P. et al. (2008), "Confirmatory tetrad analysis in PLS path modeling". Journal of
Business Research, Vol. 61 pp. 1238-1249.
Hair, J.F. et al. (2018) Multivariate Data Analysis. Cengage.
Hair, J.F., Ringle, C.M. and Sarstedt, M. (2011), "PLS-SEM: Indeed a silver bullet". Journal
of Marketing Theory and Practice, Vol. 19 No. 2, pp. 139-152.
Hair, J.F. et al. (2019), "When to use and how to report the results of PLS-SEM". European
Business Review, Vol. 31 pp. 2-24.
Han, S. and Yang, H. (2018), "Understanding adoption of intelligent personal assistants".
Industrial Management & Data Systems, Vol. No.
He, Z. et al. (2017), "Measuring service quality in telematics service: development and
validation of multidimensional TeleServQ scale". Total Quality Management &
Business Excellence, Vol. 28 No. 9-10, pp. 1166-1182.
Henseler, J. (2012) 'PLS-MGA: A non-parametric approach to partial least squares-based
multi-group analysis'. Challenges at the interface of data analysis, computer science,
and optimization. Springer, pp. 495-501.
Henseler, J., Ringle, C.M. and Sarstedt, M. (2015), "A new criterion for assessing discriminant
validity in variance-based structural equation modeling". Journal of the Academy of
Marketing Science, Vol. 43 No. 1, pp. 115-135.
Henseler, J., Ringle, C.M. and Sinkovics, R.R. (2009), "The use of partial least squares path
modeling in international marketing". Advances in International Marketing, Vol. 20
No. 277-319.
Hoy, M.B. (2018), "Alexa, Siri, Cortana, and more: an introduction to voice assistants".
Medical reference services quarterly, Vol. 37 No. 1, pp. 81-88.
Huang, E.Y., Lin, S.-W. and Fan, Y.-C. (2015), "MS-QUAL: Mobile service quality
measurement". Electronic Commerce Research and Applications, Vol. 14 No. 2, pp.
126-142.
Huang, M.-H. and Rust, R.T. (2017), "Technology-driven service strategy". Journal of the
Academy of Marketing Science, Vol. 45 No. 6, pp. 906-924.
Huang, M.-H. and Rust, R.T. (2018), "Artificial intelligence in service". Journal of Service
Research, Vol. 21 No. 2, pp. 155-172.
Huang, M.-H. and Rust, R.T. (2021), "Engaged to a Robot? The Role of AI in Service". Journal
of Service Research, Vol. 24 No. 1, pp. 30-41.
Hussain, K. et al. (2019), "The dynamic outcomes of service quality: a longitudinal
investigation". Journal of Service Theory and Practice, Vol. 29 No. 4, pp. 513-536.
Iqbal, M.S., Hassan, M.U. and Habibah, U. (2018), "Impact of self-service technology (SST)
service quality on customer loyalty and behavioral intention: The mediating role of
customer satisfaction". Cogent Business & Management, Vol. 5 No. 1, pp. 1423770.

39
Jarvis, C.B., MacKenzie, S.B. and Podsakoff, P.M. (2003), "A critical review of construct
indicators and measurement model misspecification in marketing and consumer
research". Journal of Consumer Research, Vol. 30 No. 2, pp. 199-218.
Kepuska, V. and Bohouta, G. (2018) Published. 'Next-generation of virtual personal assistants
(microsoft cortana, apple siri, amazon alexa and google home)'. 2018 IEEE 8th Annual
Computing and Communication Workshop and Conference (CCWC), 2018. IEEE,
pp.99-103.
Kidd, C. (2019) Chatbot vs Virtual Agent: What’s The Difference? Available:
https://www.bmc.com/blogs/chatbot-vs-virtual-agent/ Accessed April 12, 2021. BMC.
Kim, S.Y., Schmitt, B.H. and Thalmann, N.M. (2019), "Eliza in the uncanny valley:
Anthropomorphizing consumer robots increases their perceived warmth but decreases
liking". Marketing Letters, Vol. 30 No. 1, pp. 1-12.
Krehl, E.-H. (2020) 'How Technology is changing Employees’ Roles in the Service Encounter–
A Skill-based Analysis'. Automatisierung und Personalisierung von Dienstleistungen.
Springer, pp. 227-248.
Kuo, Y.-F., Wu, C.-M. and Deng, W.-J. (2009), "The relationships among service quality,
perceived value, customer satisfaction, and post-purchase intention in mobile value-
added services". Computers in Human Behavior, Vol. 25 No. 4, pp. 887-896.
Ladhari, R. (2009), "A review of twenty years of SERVQUAL research". International journal
of quality and service sciences, Vol. 1 pp. 172-198.
Ladhari, R. (2010), "Developing e-service quality scales: a literature review". Journal of
Retailing and Consumer services, Vol. 17 pp. 464-477.
Le, N.T., Rao Hill, S. and Troshani, I. (2020), "Perceived control and perceived risk in self-
service technology recovery". Journal of Computer Information Systems,
https://doi.org/10.1080/08874417.08872020.01756533.
Lin, J.-S.C. and Hsieh, P.-L. (2011), "Assessing the self-service technology encounters:
development and validation of SSTQUAL scale". Journal of retailing, Vol. 87 No. 2,
pp. 194-206.
Lu, V.N. et al. (2020), "Service robots, customers and service employees: what can we learn
from the academic literature and where are the gaps?". Journal of Service Theory and
Practice, Vol. 30 No. 3, pp. 361-391.
MacKenzie, S.B., Podsakoff, P.M. and Podsakoff, N.P. (2011), "Construct measurement and
validation procedures in MIS and behavioral research: Integrating new and existing
techniques". MIS Quarterly, Vol. No. pp. 293-334.
Makadia, M. (2020) Can Artificial Intelligence Improve Your Customer Service? Available:
https://www.business2community.com/customer-experience/can-artificial-
intelligence-improve-your-customer-service-02336788. Accessed April 19th, 2021.
Business 2 Community.
Martínez, J.A. and Martínez, L. (2010), "Some insights on conceptualizing and measuring
service quality". Journal of Retailing and Consumer Services, Vol. 17 No. 1, pp. 29-
42.
Meuter, M.L. et al. (2000), "Self-service technologies: understanding customer satisfaction
with technology-based service encounters". Journal of marketing, Vol. 64 No. 3, pp.
50-64.
Meyer-Waarden, L. et al. (2020), "How service quality influences customer acceptance and
usage of robots". Journal of Service Management Research, Vol. 4 No. 1, pp. 35-51.
Mittal, B. and Lassar, W.M. (1996), "The role of personalization in service encounters".
Journal of Retailing, Vol. 72 No. 1, pp. 95-109.
Mogaji, E. and Erkan, I. (2019), "Insight into consumer experience on UK train transportation
services". Travel Behaviour and Society, Vol. 14 No. 21-33.

40
Mogaji, E., Soetan, T.O. and Kieu, T.A. (2020), "The implications of artificial intelligence on
the digital marketing of financial services to vulnerable customers". Australasian
Marketing Journal (AMJ), Vol. No.
Morita, T. et al. (2020), "Evaluation of a multi-robot cafe based on service quality dimensions".
The Review of Socionetwork Strategies, Vol. 14 No. 1, pp. 55-76.
Moussawi, S. (2016) Investigating personal intelligent agents in everyday life through a
behavioral lens. PhD Thesis City University of New York, Ann Arbor.
Nath, P. (2020), "Taking measure: the link between metrics and marketing’s exploitative and
explorative capabilities". European Journal of Marketing, Vol. 54 No. 1549-1580, pp.
Naveen, J. (2018) Yes, Chatbots And Virtual Assistants Are Different! Available:
https://www.forbes.com/sites/cognitiveworld/2018/12/23/yes-chatbots-and-virtual-
assistants-are-different/?sh=35542e2a6d7d. Accessed April 12th, 2021. Forbes.
Netemeyer, R.G., Bearden, W.O. and Sharma, S. (2003) Scaling procedures: Issues and
applications. London: Sage Publications.
Neuhofer, B., Magnus, B. and Celuch, K. (2020), "The impact of artificial intelligence on event
experiences: a scenario technique approach". Electronic Markets, Vol. No. 1-17.
Noor, N., Rao Hill, S. and Troshani, I. (2021a), "Artificial Intelligence Service Agents: Role
of Parasocial Relationship". Journal of Computer Information Systems, Vol.
https://doi.org/10.1080/08874417.2021.1962213 No. 1-15.
Noor, N., Rao Hill, S. and Troshani, I. (2021b), "Recasting service quality for AI-based
service". Australasian Marketing Journal, Vol. No.
https://doi.org/10.1177/18393349211005056.
Nunnally, J.C. (1978) Psychometric Theory. New York: McGraw-Hill.
Oh, H. (1999), "Service quality, customer satisfaction, and customer value: A holistic
perspective". International Journal of Hospitality Management, Vol. 18 No. 1, pp. 67-
82.
Olson, C. and Kemery, K. (2019), "Voice report: From answers to action: customer adoption
of voice technology and digital assistants". Microsoft Search and Market Intelligence,
Tech. Rep, Vol. No.
Oosthuizen, K. et al. (2020), "Artificial intelligence in retail: The AI-enabled value chain".
Australasian Marketing Journal, Vol. No.
https://doi.org/10.1016/j.ausmj.2020.1007.1007.
Orel, F.D. and Kara, A. (2014), "Supermarket self-checkout service quality, customer
satisfaction, and loyalty: Empirical evidence from an emerging market". Journal of
Retailing and Consumer services, Vol. 21 No. 2, pp. 118-129.
Overgoor, G. et al. (2019), "Letting the computers take over: Using AI to solve marketing
problems". California Management Review, Vol. 61 No. 4, pp. 156-185.
Paluch, S. and Wirtz, J. (2020), "Artificial intelligence and robots in the service encounter".
Journal of Service Management Research, Vol. 4 No. 1, pp. 3-8.
Parasuraman, A. (2000), "Technology readiness index (TRI): a multiple-item scale to measure
readiness to embrace new technologies". Journal of Service Research, Vol. 2 No. 4, pp.
307-320.
Parasuraman, A., Berry, L.L. and Zeithaml, V.A. (1991), "Understanding customer expectation
of service". Sloan Management Review, Vol. 32 No. 3, pp. 39-48.
Parasuraman, A., Berry, L.L. and Zeithaml, V.A. (1993), "More on improving service quality
measurement". Journal of Retailing, Vol. 69 No. 1, pp. 140-147.
Parasuraman, A., Zeithaml, V.A. and Berry, L. (1988), "SERVQUAL: A multiple-item scale
for measuring consumer perceptions of service quality". 1988, Vol. 64 No. 1, pp. 12-
40.

41
Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1985), "A conceptual model of service
quality and its implications for future research". Journal of Marketing, Vol. 49 No. 1,
pp. 41-50.
Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1994a), "Alternative scales for measuring
service quality: a comparative assessment based on psychometric and diagnostic
criteria". Journal of retailing, Vol. 70 No. 3, pp. 201-230.
Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1994b), "Reassessment of expectations as a
comparison standard in measuring service quality: implications for further research".
Journal of marketing, Vol. 58 No. 1, pp. 111-124.
Parasuraman, A., Zeithaml, V.A. and Malhotra, A. (2005), "ES-QUAL: A multiple-item scale
for assessing electronic service quality". Journal of service research, Vol. 7 No. 3, pp.
213-233.
Park, S.S., Tung, C.D. and Lee, H. (2021), "The adoption of AI service robots: a comparison
between credence and experience service settings". Psychology & Marketing, Vol. 38
No. 691-703, pp.
Paschen, J., Wilson, M. and Ferreira, J.J. (2020), "Collaborative intelligence: how human and
artificial intelligence create value along the B2B sales funnel". Business Horizons, Vol.
63 No. 3, pp. 403-414.
PATResearch (2021) Top 22 intelligent personal assistants or automated personal assistants.
Available: https://www.predictiveanalyticstoday.com/top-intelligent-personal-
assistants-automated-personal-assistants/. Accessed April 12, 2021. Predictive
Analytics Today.
Poole, D.L. and Mackworth, A.K. (2010) Artificial Intelligence: foundations of computational
agents. Camridge: Cambridge University Press.
Prentice, C., Lopes, S.D. and Wang, X. (2020), "The impact of artificial intelligence and
employee service quality on customer satisfaction and loyalty". Journal of Hospitality
Marketing & Management, Vol. 29 No. 7, pp. 739-756.
PwC (2018) Artificial intelligence: Touchpoints with consumersAvailable
https://www.pwc.com/co/es/industrias/Retail/insights/gcis-ai_vf.pdf. Accessed April
24th 2021. PwC.
Qiu, L. and Benbasat, I. (2009), "Evaluating anthropomorphic product recommendation agents:
A social relationship perspective to designing information systems". Journal of
management information systems, Vol. 25 No. 4, pp. 145-182.
Ranjan, K.R. and Read, S. (2016), "Value co-creation: concept and measurement". Journal of
the Academy of Marketing Science, Vol. 44 No. 3, pp. 290-315.
Rank, J. et al. (2007), "Proactive customer service performance: Relationships with individual,
task, and leadership variables". Human Performance, Vol. 20 No. 4, pp. 363-390.
Rita, P., Oliveira, T. and Farisa, A. (2019), "The impact of e-service quality and customer
satisfaction on customer behavior in online shopping". Heliyon, Vol. 5 No. 10, pp.
e02690.
Robinson, S. et al. (2020), "Frontline encounters of the AI kind: An evolved service encounter
framework". Journal of Business Research, Vol. 116 No. 366-376.
Russell, S. and Norvig, P. (2018) Artificial intelligence: a modern approach. Boston, MA:
Pearson.
Rust, R.T. (2020), "The future of marketing". International Journal of Research in Marketing,
Vol. 37 No. 1, pp. 15-26.
Rust, R.T. and Huang, M.-H. (2014), "The service revolution and the transformation of
marketing science". Marketing Science, Vol. 33 No. 2, pp. 206-221.
Rust, R.T. and Oliver, R.L. (1993) Service quality: New directions in theory and practice.
London: Sage Publications.

42
Salomann, H. et al. (2007), "Self-service Revisited: How to Balance High-tech and High-touch
in Customer Relationships". European Management Journal, Vol. 25 No. 4, pp. 310-
319.
Schniter, E., Shields, T.W. and Sznycer, D. (2020), "Trust in humans and robots: Economically
similar but emotionally different". Journal of Economic Psychology, Vol. 78 No.
https://doi.org/10.1016/j.joep.2020.102253.
Seth, N., Deshmukh, S. and Vrat, P. (2005), "Service quality models: a review". International
journal of quality & reliability management, Vol. 22 No. 9, pp. 913-949.
Shankar, V. (2018), "How artificial intelligence (AI) is reshaping retailing". Journal of
Retailing, Vol. 94 No. 4, pp. vi-xi.
Sheehan, B., Jin, H.S. and Gottlieb, U. (2020), "Customer service chatbots: Anthropomorphism
and adoption". Journal of Business Research, Vol. 115 No. 14-24.
Shell, M.A. and Buell, R.W. (2019), "How Necessary Is the Human Touch?". Harvard
Business Review, Vol. 97 No. 4, pp. 28-28.
Shepardson, D. (2020) White House proposes regulatory principles to govern AI use.
Available: https://www.reuters.com/article/us-tech-ces-ai-white-house-
idUSKBN1Z60GL. Accessed April 24, 2021. Reuters.
Shewan, D. (2021) 10 of the Most Innovative Chatbots on the Web. Available:
https://www.wordstream.com/blog/ws/2017/10/04/chatbots Accessed April 12th, 2021.
Stock, R.M. and Merkle, M. (2018) Published. 'Can humanoid service robots perform better
than service employees? A comparison of innovative behavior cues. Available:
https://scholarspace.manoa.hawaii.edu/bitstream/10125/50020/1/paper0133.pdf'.
Proceedings of the 51st Hawaii International Conference on System Sciences, 2018
Waikoloa Village, HI, January 3-6. pp.1056-1065.
Tabachnick, B.G., Fidell, L.S. and Ullman, J.B. (2007) Using multivariate statistics. Pearson
Boston, MA.
Tam, J.L. (2004), "Customer satisfaction, service quality and perceived value: an integrative
model". Journal of Marketing Management, Vol. 20 No. 7-8, pp. 897-917.
Tan, F.B. and Chou, J.P. (2008), "The relationship between mobile service quality, perceived
technology compatibility, and users' perceived playfulness in the context of mobile
information and entertainment services". Intl. Journal of Human–Computer
Interaction, Vol. 24 No. 7, pp. 649-671.
Theodosiou, M. et al. (2019), "A comparison of formative versus reflective approaches for the
measurement of electronic service quality". Journal of Interactive Marketing, Vol. 47
No. 53-67.
Troshani, I. et al. (2020), "Do We Trust in AI? Role of Anthropomorphism and Intelligence".
Journal of Computer Information Systems, Vol. No.
Verhagen, T. et al. (2014), "Virtual customer service agents: Using social presence and
personalization to shape online service encounters". Journal of Computer-Mediated
Communication, Vol. 19 No. 3, pp. 529-545.
Voorhees, C.M. et al. (2016), "Discriminant validity testing in marketing: an analysis, causes
for concern, and proposed remedies". Journal of the academy of marketing science,
Vol. 44 No. 1, pp. 119-134.
Whang, C. and Im, H. (2021), "" I Like Your Suggestion!" the role of humanlikeness and
parasocial relationship on the website versus voice shopper's perception of
recommendations". Psychology & Marketing, Vol. 38 No. 4, pp. 581-595.
Wirtz, J. et al. (2018), "Brave new world: service robots in the frontline". Journal of Service
Management, Vol. 29 No. 5, pp. 776-808.

43
Yang, Z., Jun, M. and Peterson, R.T. (2004), "Measuring customer perceived online service
quality". International Journal of Operations & Production Management, Vol. 24 No.
11, pp. 1149-1174.
Youn, S. and Jin, S.V. (2021), "In AI we trust?” The effects of parasocial interaction and
technopian versus luddite ideological views on chatbot-based customer relationship
management in the emerging “feeling economy". Computers in Human Behavior, Vol.
119 No. 106721.
Zarouali, B. et al. (2018), "Predicting consumer responses to a chatbot on Facebook".
Cyberpsychology, Behavior, and Social Networking, Vol. 21 No. 8, pp. 491-497.
Zeithaml, V.A. (1988), "Consumer perceptions of price, quality and value: a means-end model
and synthesis of evidence". Journal of Marketing, Vol. 52 No. 1, pp. 2-22.
Zeithaml, V.A., Berry, L.L. and Parasuraman, A. (1988), "Communication and Control
Processes in the Delivery of Service Quality". Journal of Marketing, Vol. 52 No. 2, pp.
35-48.
Zeithaml, V.A., Berry, L.L. and Parasuraman, A. (1996a), "The behavioral consequences of
service quality". The Journal of Marketing, Vol. No. 31-46.
Zeithaml, V.A., Berry, L.L. and Parasuraman, A. (1996b), "The behavioural consequences of
service quality". Journal of Marketing, Vol. 60 No. 1, pp. 31-46.
Zeithaml, V.A., Parasuraman, A. and Malhotra, A. (2002), "Service quality delivery through
web sites: a critical review of extant knowledge". Journal of the Academy of Marketing
Science, Vol. 30 No. 4, pp. 362-375.
Zendesk (2020) The business impact of customer service on customer lifetime value. Available:
https://www.zendesk.com/blog/customer-service-and-lifetime-customer-value/.
Published April 8th, 2013. Updated October 6th, 2020. Accessed April 19th, 2021.
Zendesk.
Zheng, X. et al. (2020), "Role of technology attraction and parasocial interaction in social
shopping websites". International Journal of Information Management, Vol. 51 No.
102043.

44
APPENDIX: Measures of Constructs
Respondents were asked to rate the following statements using a seven-point Likert scale anchored from 1 =
strongly disagree to 7 = strongly agree. Items appeared in random order in the survey.

AISAQUAL

Efficiency
EFF1 The AISA works correctly at first attempt.
EFF2 I can get my task done with the AISA in a short time.
EFF3 The AISA interface design provides information clearly.
EFF4 The AISA adequately meets my requirements.

Security
SEC1 There is no risk of loss associated with disclosing personal information to the AISA.
SEC2 I feel secure in providing sensitive information to the AISA.
SEC3 I believe that information that the AISA has about me is protected.
SEC4 I trust that my personal information with the AISA will not be misused.

Availability
AVA1 The AISA is always available.
AVA2 The AISA is never too busy to respond to my requests.
AVA3 The AISA is always accessible.

Enjoyment
ENJ1 Using the AISA is fun.
ENJ2 Using the AISA is enjoyable.
ENJ3 Using the AISA is interesting.
ENJ4 Using the AISA is entertaining.

Contact
CON1 Human assistants are available to contact via the AISA.
CON2 Follow-up services with human assistants are available to me when necessary.
CON3 I can speak to a human assistant via the AISA.
CON4 Human assistance is easy to access via the AISA.
CON5 The AISA provides detailed contact information when I need human assistance.

Anthropomorphism
ANT1 The AISA has humanlike features.
ANT2 The AISA has personality.
ANT3 The AISA gradually gets to know me.
ANT4 The AISA is able to behave like a human.
ANT5 The AISA responds in ways that are personalized.
ANT6 The AISA is able to communicate like a human.

Satisfaction
SAT1 I am satisfied with my decision to use the AISA.
SAT2 I think that I did the right thing by using the AISA.
SAT3 My choice to use the AISA was a wise one.

Perceived Value
VAL1 Overall, the AISA gives me good value.
VAL2 The time I spent on the AISA was worthwhile.

Loyalty Intentions
LOY1 I will say positive things about the AISA to other people.
LOY2 I will recommend the AISA to someone who seeks my advice.
LOY3 I will encourage friends and others to use the AISA.
LOY4 I will consider the AISA to be my first choice for future tasks.
LOY5 I will use the AISA more in the coming months.

45
 Table I.
Chatbots and Virtual assistants

Chatbot Virtual assistant

Common Chatbots are rule-based AI applications that are Virtual assistants are software agents that act as
functionality and typically used in business to facilitate customer personal assistant to users for a wide range of tasks
uses interaction. Chatbot are also known as in daily activities, both work-related and personal,
‘conversational agents’ as typical interaction such as personal communications management such
with customers is in the form for a structured as email and SMS, productivity management such
dialogue or chat via spoken language text-based as scheduling of meetings, or performing general
conversational interface, although there are also tasks such as playing music, setting alarms. Virtual
chatbots that support audio and images assistants are designed to understand natural
(Shewan, 2021). Typically, consumers ask language voice commands and questions including
specific questions, while the chatbot provides implied meaning, user emotions, language slangs,
live pre-defined responses. Companies have and dialects. Virtual assistants can learn from prior
been using chatbots to engage with consumers interactions with their users, consequently improve
in a range of roles such as consumer service and ability to contextualize and customize interactions
support such as order customization, scheduling overtime.
(e.g., product deliveries, or travel bookings) or
refunds.
Usage channel Websites, support portals, messaging channels Mobile phones, laptop computers, smart speakers
and mobile applications. and interactive devices.
Common IBM’s Watson Assistant, SalesForce’s Eistein Apple’s Siri, Amazon’s Alexa, Microsoft’s Cortana
examples (Chi, 2021). (PATResearch, 2021)

46
 Table II.
Dimensions representing service quality of AISA

Dimension Definition Key supporting service quality

scale studies
Reliability Ability of the AISA to perform the service dependably Parasuraman et al. (1988)
and accurately
Responsiveness Prompt response of the AISA to customer requests and Yang et al. (2004)
the speed in resolving customer problems
Availability Ability of the AISA to be ready for use anytime, (Lin and Hsieh, 2011),
anywhere Parasuraman et al. (2005)
Aesthetics Appeal and clarity associated with the AISA interface Dabholkar (1996)
design
Personalization Ability of the AISA to meet the customer’s individual He et al. (2017)
preferences
Security Perceived safety of the AISA from intrusion, fraud and He et al. (2017)
loss of personal information and privacy
Control Degree of control that the customer feels over the Dabholkar (1996)
process or outcome of the service encounter with the
AISA
Ease of Use Degree to which using the AISA would be free of Davis (1989)
effort
Enjoyment Extent to which using the AISA is perceived to be Davis et al. (1992)
enjoyable in its own right, apart from any performance
consequences that may be anticipated
Contact Access to human assistance Parasuraman et al. (2005)
Proactiveness Extent to which AISA is able to predict and anticipate Noor et al. (2021b)
customers’ future needs and acts beyond explicitly
prescribed commands.
Anthropomorphism Extent to which AISA shows human-like Noor et al. (2021b)
characteristics, motivations, intentions, or emotions.

47
 Table III.
Profile of respondents for scale refinement phase

Category EFA (N=211) CFA (N=275)

Gender
Male 107 (50.7%) 132 (48%)
Female 104 (49.3%) 143 (52%)
Total 211 275

Age
18-24 21 (10.0%) 26 (9.5%)
25-34 70 (33.2%) 91 (33.0%)
35-44 72 (34.1%) 92 (33.4%)
45-54 30 (14.2%) 49 (17.8%)
55-64 11 (5.2%) 13 (4.7%)
65 and above 7 (3.3%) 4 (1.5%)
Total 211 275

Chatbot usage context

Accommodation and food services 8 (7.6%) 11 (7.9%)
Administrative and support services 7 (6.7%) 9 (6.4%)
Arts and recreation services 5 (4.8%) 7 (5%)
Education and training 4 (3.8%) 5 (3.6%)
Electricity, gas, water and waste services 21 (20%) 26 (18.6%)
Financial and insurance services 15 (14.3%) 18 (12.9%)
Health care and social assistance 5 (4.8%) 8 (5.7%)
Information media and telecommunications 16 (15.2%) 17 (12.1%)
Professional, scientific and technical services 7 (6.7%) 9 (6.4%)
Public administration and safety 3 (2.9%) 6 (4.3%)
Rental, hiring and real estate services 3 (2.9%) 4 (2.9%)
Retail trade 4 (3.8%) 9 (6.4%)
Transport, postal and warehousing 7 (6.7%) 12 (8.6%)
Total 105 140

Virtual assistant
Alexa 31 (29.2%) 40 (29.6%)
Bixby 2 (1.9%) 3 (2.2%)
Google Assistant 42 (39.6%) 43 (31.9%)
Google Home Mini 4 (3.8%) 8 (5.9%)
Siri 27 (25.5%) 41 (30.4%)
Total 106 135

48
 Table IV.
EFA and CFA results and final AISAQUAL scale
Dimension Itema Original EFA CFAb Final Label
after EFA Identifierc Loading Loading
D1 The AISA provides the service as expected. REL7 .74 Efficiency
(Cronbach’s The AISA works correctly at first attempt.a REL5 .73 .72
α = .84) I can get my task done with the AISA in a short time. a RES1 .66 .74
The AISA can perform the task quickly. RES2 .65
The AISA interface design provides information clearly. a AES5 .63 .73
I know how long it takes to complete the task with the AISA. CTL3 .60
The AISA adequately meets my requirements.a REL3 .52 .82

D2 I trust that my personal information with the AISA is safe. SEC5 .83 Security
(Cronbach’s There is no risk of loss associated with disclosing SEC9 .73 .77
α = .88) personal information to the AISA.a
I feel secure in providing sensitive information to the SEC3 .71 .76
AISA.a
I believe that information that the AISA has about me is SEC7 .65 .88
protected.a
I trust that my personal information with the AISA will SEC6 .59 .80
not be misused.a

D3 The AISA is always available.a AVA2 .84 .71 Availability

(Cronbach’s The AISA is never too busy to respond to my requests.a AVA4 .71 .87
α = .80) The AISA launches right away. RES3 .70
The AISA is always accessible.a AVA1 .58 .79
The AISA interface design is innovative. AES1 .46

D4 Using the AISA is fun.a ENJ5 .67 .74 Enjoyment

(Cronbach’s Using the AISA is enjoyable.a ENJ3 .52 .87
α = .89) Using the AISA is interesting.a ENJ6 .46 .79
Using the AISA is entertaining.a ENJ4 .43 .84

D5 Human assistants are available to contact via the AISA. a CTC2 .81 .78 Contact
(Cronbach’s Follow-up services with human assistants are available to CTC5 .71 .71
α = .88) me when necessary.a
I can speak to a human assistant via the AISA.a CTC3 .65 .78
Human assistance is easy to access via the AISA.a CTC1 .64 .82
The AISA provides detailed contact information when I CTC4 .60 .80
need human assistance.a
A clear privacy policy is accessible before I use the AISA. SEC1 .45

D6 The AISA has humanlike features.a ANT1 .85 .76 Anthropomorphism

(Cronbach’s The AISA has personality.a ANT5 .77 .74
α = .90) The AISA gradually gets to know me.a PER9 .72 .75
The AISA is able to behave like a human.a ANT4 .64 .84
The AISA responds in ways that are personalized.a PER8 .62 .84
The AISA is able to communicate like a human.a ANT3 .61 .78
The AISA uses its own ‘judgment’ to complete a task. PRO5 .47
a
final AISAQUAL items are shown in bold
χ = 835.01, p < .001; RMSEA = .07, TLI = .91, CFI = .90, SRMR = .08.
b 2
c
REL=Reliability, RES=Responsiveness, AVA=Availability, AES=Aesthetics, PER=Personalization, SEC=Security, CTL=Control,
EAS=Ease of Use, ENJ=Enjoyment, CTC=Contact, PRO=Proactiveness, ANT=Anthropomorphism

49
 Table V.
Profile of respondents for scale validation phase
Category Frequency Percentage Category Frequency Percentage

Gender AISA usage frequency

Male 152 50.0 Daily 95 31.3
Female 152 50.0 Weekly 97 31.9
Total 304 100.0 Every 2-3 weeks 43 14.1
Monthly 33 10.9
Age Every 2-3 months 14 4.6
18-24 54 17.8 Every 4-6 months 10 3.3
25-34 79 26.0 Once a year 12 3.9
35-44 72 23.7 Total 304 100.0
45-54 44 14.5
55-64 35 11.5
65 and above 20 6.6
Total 304 100.0

Highest education AISA usage experience

Less than high school 2 0.7 Less than 1 year 85 28.0
High school 50 16.4 2-3 years 144 47.4
Vocational training 13 4.3 4-5 years 51 16.8
Some college 86 28.3 6-7 years 13 4.3
Bachelor’s degree 95 31.3 8 years and above 11 3.6
Postgraduate degree 58 19.1 Total 304 100.0
Total 304 100.0

Work industry Chatbot usage context

Accommodation and food services 10 3.3 Accommodation and food services 10 6.6
Administrative and support 20 6.6 Administrative and support 19 12.6
services services
Arts and recreation services 16 5.3 Arts and recreation services 6 4.0
Construction 8 2.6 Construction 2 1.3
Education and training 26 8.6 Education and training 11 7.3
Electricity, gas, water and waste 8 2.6 Electricity, gas, water and waste 5 3.3
services services
Financial and insurance services 23 7.6 Financial and insurance services 19 12.6
Health care and social assistance 34 11.2 Health care and social assistance 12 7.9
Information media and 18 5.9 Information media and 18 11.9
telecommunications telecommunications
Manufacturing 7 2.3 Professional, scientific and 5 3.3
Professional, scientific and 24 7.9 technical services
technical services Public administration and safety 2 1.3
Public administration and safety 9 3.0 Rental, hiring and real estate 4 2.6
Rental, hiring and real estate 5 1.6 services
services Retail trade 30 19.9
Retail trade 33 10.9 Transport, postal and warehousing 4 2.6
Transport, postal and warehousing 19 6.3 Others 4 2.6
Other Industries 7 2.3 Total 151 100.0
Retired 19 6.3
Unemployed 18 5.9
Total 304 100.0

Personal annual income (USD) Virtual assistant

Less than $25,000 62 20.4 Alexa 60 39.2
$25,000 to $49,999 75 24.7 Bixby 5 3.3
$50,000 to $74,999 60 19.7 Google Assistant 36 23.5
$75,000 to $99,999 45 14.8 Google Home Mini 18 11.8
$100,000 and more 62 20.4 Siri 34 22.2
Total 304 100.0 Total 153 100.0

50
 Table VI.
CFA results of AISAQUAL for scale validation phase

Dimension Item Mean S.D. CFA CR AVE

Loading
Efficiency The AISA works correctly at first attempt. 4.77 1.38 .69 .84 .57
(Cronbach’s α = I can get my task done with the AISA in a short time. 5.12 1.37 .79
.84) The AISA interface design provides information clearly. 5.20 1.27 .72
The AISA adequately meets my requirements. 5.10 1.40 .82

Security There is no risk of loss associated with disclosing personal 3.99 1.69 .78 .89 .67
(Cronbach’s α = information to the AISA.
.89) I feel secure in providing sensitive information to the AISA. 4.16 1.71 .81
I believe that information that the AISA has about me is protected. 4.46 1.63 .86
I trust that my personal information with the AISA will not be 4.43 1.65 .81
misused.

Availability The AISA is always available. 5.42 1.37 .75 .80 .58
(Cronbach’s α = The AISA is never too busy to respond to my requests. 5.38 1.53 .74
.80) The AISA is always accessible. 5.40 1.38 .80

Enjoyment Using the AISA is fun. 5.02 1.60 .85 .89 .68
(Cronbach’s α = Using the AISA is enjoyable. 5.05 1.48 .80
.89) Using the AISA is interesting. 5.17 1.44 .79
Using the AISA is entertaining. 4.91 1.61 .84

Contact Human assistants are available to contact via the AISA. 4.59 1.50 .76 .87 .58
(Cronbach’s α = Follow-up services with human assistants are available to me when 4.76 1.45 .76
.87) necessary.
I can speak to a human assistant via the AISA. 4.53 1.56 .76
Human assistance is easy to access via the AISA. 4.79 1.59 .81
The AISA provides detailed contact information when I need 4.77 1.49 .71
human assistance.

Anthropomorphism The AISA has humanlike features. 4.56 1.64 .74 .90 .60
(Cronbach’s α = The AISA has personality. 4.46 1.60 .79
.90) The AISA gradually gets to know me. 4.59 1.55 .74
The AISA is able to behave like a human. 4.40 1.65 .77
The AISA responds in ways that are personalized. 4.74 1.47 .80
The AISA is able to communicate like a human. 4.72 1.55 .83

χ2 = 562.94, p < .001; RMSEA = .06, TLI = .94, CFI = .95 and SRMR = .04.

51
 Table VII.
HTMT analysis of AISAQUAL

Constructs EFF SEC AVA ENJ CON ANT

Efficiency (EFF)
Security (SEC) .75
Availability (AVA) .78 .50
Enjoyment (ENJ) .86 .61 .68
Contact (CON) .80 .63 .53 .66
Anthropomorphism (ANT) .86 .79 .62 .84 .73

Table VIII.
Squared correlations between AISAQUAL dimensions.

Constructs EFF SEC AVA ENJ CON ANT

Efficiency (EFF) .75
Security (SEC) .65 .82
Availability (AVA) .64 .42 .76
Enjoyment (ENJ) .74 .53 .58 .82
Contact (CON) .68 .55 .44 .58 .76
Anthropomorphism (ANT) .74 .71 .53 .76 .65 .77
Notes: average variance extracted appears on the diagonal. All correlations are significant at p< .01.

Table IX.
Multigroup comparison test results of AISAQUAL

Path Chatbot Virtual Assistant Multigroup Analysis

Users (151) Users (153)
β t β t β p Result
Difference Difference
AISAQUAL -> Loyalty .17 2.62 .16 2.90 .01 .95 Rejected
Intentions
AISAQUAL -> Perceived .83 29.98 .81 23.85 .01 .81 Rejected
Value
AISAQUAL -> .37 4.81 .34 5.45 .03 .74 Rejected
Satisfaction
Perceived Value -> .33 3.76 .12 1.18 .22 .11 Rejected
Loyalty Intentions
Perceived Value -> .55 7.22 .62 9.73 -.07 .51 Rejected
Satisfaction
Satisfaction -> Loyalty .47 5.70 .69 6.15 -.23 .11 Rejected
Intentions
Note: Multigroup analysis based on 5000 bootstrap. Results are based on two tail test at 5% probability of error level.

52
 Step 1: phenomenon and domain definition
Defined domain of AISAQUAL based on extant literature.

Step 2: Item generation

Adopt AISAQUAL dimensions proposed by Noor et al. (2021b). Generate an item pool based on
literature and content and face validated feedback from six experts.

Step 3: Scale refinement

Administered pilot scale test to 211 chatbot and virtual assistant users from the USA. Developed
parsimonous scale through iterative purification process by examining coefficient alpha and
dimensionality using exploratory factor analysis (EFA). Assessed scale structure using confirmatory
factor analysis (CFA) on a new sample of 275 chatbot and virtual assistant users.

Step 4: Scale validation

Administered finalized scale to 304 chatbot and virtual assistant users from the USA. Assesed scale
reliability and and validity.

Figure 1. Research approach

53
Figure 2. Extant service quality scales used to form initial AISAQUAL item battery

54
 Figure 3. Model for AISAQUAL nomological validity assessment. All parameter estimates
are significant at the .001 level.

55

View publication stats