AI Instrumentality and Brand Credibility in Voice Assistant Retention–Model

Abstract

The AI Instrumentality and Brand Credibility in Voice Assistant Retention—Model was developed by Ghazali, Mutum, and Lun (2024) to rigorously investigate consumer continuance behavior concerning AI voice assistants (AIVAs). This comprehensive framework significantly extends the established Expectation-Confirmation Model (ECM) by incorporating critical user technology-related traits, specific AI instrumentality attributes, and the crucial factor of brand credibility. The instrument, a 63-item questionnaire adapted from validated psychological and technological scales, was assessed using structural equation modeling (SEM) on a sample of past AIVA users, confirming its strong reliability and validity for measuring retention factors.

Keywords

Anthropomorphism, Brand Expertise, Brand Trustworthiness, Confirmation, Continuance Intention, Discomfort, Info Accuracy, Info Completeness (Reliability), Info Up-to-Datedness (Recency), Innovativeness, Insecurity, Intelligence, Optimism, Satisfaction, System Flexibility, System Reliability, System Timeliness, Measurement Model, Consumer Behavior, Technology Acceptance.

Authors

Ghazali, Ezlika, Mutum, Dilip S., Lun, Na Kai

Purpose

The primary purpose of the model is to systematically evaluate the drivers of user retention and continued usage intention among consumers interacting with AI voice assistants (AIVAs). Specifically, it aims to determine the extent to which perceptions of brand expertise and brand trustworthiness influence post-use satisfaction, which, in turn, drives the intention for long-term use.

This research addresses a critical gap in technology acceptance literature by focusing on post-adoption behavior in the context of intelligent agents, providing a robust framework for understanding the interplay between technological performance (AI instrumentality) and consumer trust (brand credibility).

Construct

The model operationalizes several key psychological and technological constructs essential for understanding AIVA retention. The central constructs measured are Artificial Intelligence Instrumentality, which relates to the perceived usefulness and performance of the AI; Brand Credibility, encompassing both expertise and trustworthiness; and the ultimate outcome variable, Voice Assistant Retention (Continuance Intention).

The scale encompasses a wide array of subscales and factors, including technology readiness traits (Innovativeness, Optimism, Insecurity, Discomfort), core ECM components (Confirmation, Satisfaction, Continuance Intention), AI-specific features (Intelligence, Anthropomorphism), information quality measures (Info Accuracy, Reliability, Recency), system quality measures (System Flexibility, Reliability, Timeliness), and brand attributes (Brand Trustworthiness, Brand Expertise).

Validity

The validity of the instrument was rigorously assessed through standard psychometric procedures, ensuring the constructs accurately reflect the intended theoretical concepts. Convergent validity was established by confirming that the Average Variance Extracted (AVE) for all measured constructs exceeded the acceptable threshold of 0.5, demonstrating that items measuring the same construct were sufficiently related.

Furthermore, Discriminant validity was established using the Heterotrait-Monotrait Ratio of Correlations (HTMT) criterion. The majority of HTMT values were found to be below 0.75, confirming that each construct was empirically distinct from other constructs within the model.

Reliability

The reliability of the AI Instrumentality and Brand Credibility model was confirmed through measures of internal consistency. Both Composite Reliability (CR) and Cronbach’s Alpha coefficients were calculated for all scales, consistently exceeding the acceptable threshold of 0.7. This high level of internal consistency indicates that the items within each construct reliably measure the intended underlying factor across the consumer sample.

Factor Analysis

The primary analytical approach utilized was Structural Equation Modeling (SEM), which allowed for the simultaneous testing of the measurement model and the structural relationships hypothesized by the extended Expectation-Confirmation Model. The measurement model demonstrated strong factor loadings.

Initial analysis showed that most outer loadings exceeded the acceptable minimum of 0.7. However, six specific items (INS1, INS4, DIS1, SYT1, INA2, BT2) were removed due to insufficient loadings, optimizing the scale structure, resulting in a 63-item final inventory. Additionally, common method bias was assessed with a marker variable approach, revealing a minimal impact on the structural results (R² difference ranged only from 0.489 to 0.493).

Instrument

Test Type: Original

Format: Inventory/Questionnaire. Responses are collected using a 7-point scale ranging from “1 = Strongly Disagree” to “7 = Strongly Agree,” designed to assess user experiences with specific AIVA brands.

Language Available: English (Original research language).

Population Group: Human (Male; Female)

Age Group: Adulthood (18+ years), specifically focusing on Young Adulthood (18-29 years) and Thirties (30-39 years).

Population Details: Respondents were consumers located in Malaysia who had prior experience using AI voice assistants.

Test Methodology: The methodology included assessment of Convergent Validity, Discriminant Validity, Internal Consistency, Measurement Model analysis, and full Structural Equation Modeling.

Keywords

Brand Preferences, Consumer Attitudes, Consumer Behavior, Consumer Satisfaction, Anthropomorphism, Behavioral Intention, Consumer Measures, Technology Acceptance, Intelligent Personal Agents, AI Instrumentality, Voice Assistant Retention.

Authors

Author ORCID Identifier: Ezlika Ghazali: 0000-0001-7824-4433; Dilip S. Mutum: 0000-0002-9857-1164

Affiliation Email addresses: [email protected]

Correspondence Address: Dilip S. Mutum, Monash University Malaysia, School of Business, Subang Jaya, Malaysia, 47500

Permissions & Fee and Test Year

Test Year: 2024

Permissions and Fees: Information regarding specific permissions or licensing fees for academic or commercial use is not provided in the source material. Researchers should contact the corresponding author, Dilip S. Mutum, for usage permissions.

Reference’s

Ghazali, E., Mutum, D. S., & Lun, N. K. (2024). Expectations and beyond: The nexus of AI instrumentality and brand credibility in voice assistant retention using extended expectation-confirmation model. Middle Western Psychological Research Journal, 23(2), 655–675. DOI: https://doi.org/10.1002/cb.2228.

Items of the AI Instrumentality and Brand Credibility in Voice Assistant Retention–Model

IMPORTANT: The following scale items must be preserved in their original language and must not be changed in any way.

The final instrument consists of 63 items covering the following 17 constructs (subscales):

• Innovativeness
• Optimism
• Insecurity
• Discomfort
• Confirmation
• Satisfaction
• Continuance Intention
• Intelligence
• Anthropomorphism
• Info Accuracy
• Info Completeness (Reliability)
• Info Up-to-datedness (Recency)
• System Flexibility
• System Reliability
• System Timeliness
• Brand Trustworthiness
• Brand Expertise