Continued Use of the Device

1. Abstract

The Continued Use of the Device scale is a specialized psychometric instrument designed to measure a consumer’s cognitive expectation and behavioral commitment regarding the integration of a technological artifact—specifically voice assistants and smart hardware—into their long-term personal routines. Originally deployed by Guha, Bressgott, Grewal, Mahr, Wetzels, and Schweiger (2023) within the Journal of the Academy of Marketing Science, the scale addresses the crucial transition from initial technology trial to sustained, routinized post-adoption usage. Evaluated across multi-study experimental designs investigating how artificiality and perceived intelligence shape human-machine interactions, the instrument captures the psychological permanence of smart device adoption. The scale typically features a concise, unidimensional structure composed of high-loading Likert-type items administered on a 7-point continuum ranging from strongly disagree to strongly agree. Psychometrically, the instrument exhibits strong internal consistency (Cronbach’s alpha ($lpha$) and composite reliability ($
ho_c$) routinely exceeding .85), alongside robust convergent validity with constructs such as user satisfaction, perceived usefulness, and anthropomorphic trust. Confirmatory factor analyses demonstrate exceptional model fit indices (Comparative Fit Index [CFI] > .97, Tucker-Lewis Index [TLI] > .95, Root Mean Square Error of Approximation [RMSEA] < .06), confirming that routine-based continued usage operates distinctly from transient novelty effects or isolated transactional satisfaction. By quantifying the consumer’s forward-looking anticipation of habitual incorporation, the scale provides marketing scientists, behavioral psychologists, and human-computer interaction (HCI) researchers with an empirical benchmark for determining long-term technological viability in an era dominated by artificial intelligence, smart ambient environments, and autonomous conversational agents.

2. Keywords

Continued use, technology continuance intention, voice assistants, smart devices, human-computer interaction, artificial intelligence, routine integration, post-adoption behavior, expectation-confirmation model, psychometrics

3. Authors

The scale was adapted and validated in the context of smart device evaluations by an international team of marketing and service research scholars:

• Abhijit Guha: Associate Professor of Marketing at the Darla Moore School of Business, University of South Carolina, Columbia, SC, USA. Specializes in consumer decision-making, digital marketing interfaces, and automated service environments.
• Timna Bressgott: Assistant Professor of Marketing at the Department of Marketing, BI Norwegian Business School, Oslo, Norway. Her research centers on consumer-technology interactions, conversational agents, and consumer well-being.
• Dhruv Grewal: Toyota Chair in Commerce and Electronic Business and Professor of Marketing at Babson College, Babson Park, MA, USA. An internationally recognized scholar in digital retail, pricing architecture, customer experience management, and emerging marketing technologies.
• Dominik Mahr: Professor of Digital Innovation and Marketing, and Scientific Director of the Service Science Factory at the School of Business and Economics, Maastricht University, Maastricht, Netherlands. Focuses on customer co-creation, digital service innovation, and artificial intelligence in consumer ecosystems.
• Martin Wetzels: Professor of Marketing Analytics at the EDHEC Business School, Lille, France. Expert in advanced quantitative modeling, structural equation modeling, customer relationship management, and technology adoption dynamics.
• Elisa Schweiger: Assistant Professor of Marketing at the School of Management, University of Bath, Bath, United Kingdom. Her research investigates the psychological mechanisms governing human interactions with artificial agents, automated systems, and sensory marketing interfaces.

4. Purpose

The primary purpose of the Continued Use of the Device scale is to provide a reliable, theoretically anchored measurement of post-adoption behavioral intent, specifically focusing on whether an individual projects that an interactive device will remain embedded in their everyday lifestyle over time. While classical technology acceptance literature—most notably the foundational Technology Acceptance Model (TAM)—concentrated heavily on initial adoption intentions or experimental willingness to use a system, contemporary digital ecosystems face a different challenge: the “abandonment crisis.” Smart devices, including smart speakers (e.g., Amazon Echo, Google Nest), wearable health trackers, and autonomous home interfaces, frequently experience rapid initial trials driven by novelty, followed by steep attrition rates once the experimental novelty wears off.

Guha et al. (2023) developed and applied this measurement tool to ascertain how underlying design characteristics of artificial agents—principally their level of artificiality (humanlike vs. robotic voice and tone) and perceived intelligence (functional capability vs. general problem-solving ability)—directly or indirectly drive sustainable consumer engagement. The scale moves beyond ephemeral evaluations such as “device liking” or momentary “fun” by measuring deep routine integration. When consumers assess whether a device will continue to be a staple of their daily functioning, they evaluate pragmatic value, frictionless interaction, and cognitive dependence.

In academic research, the instrument serves as an essential dependent variable within structural equation models that test human-AI symbiosis, voice commerce (v-commerce), and ambient computing adoption. In commercial and clinical contexts, the scale provides product designers, user experience (UX) evaluators, and clinical behavioral scientists with actionable data. For example, in tele-health and assistive ambient assisted living (AAL) technologies for aging populations, measuring continued device use is vital to verify whether patients will persist in utilizing medication-dispensing or speech-based health monitoring devices over multi-month and multi-year horizons.

5. Psychological Construct

The psychological construct captured by this instrument is Continued Use Intention / Habitual Device Continuance. In psychometrics and consumer behavior, continuance is conceptually distinct from initial adoption, sporadic interaction, or passive ownership. It represents an active psychological projection of dependency, habitual utility, and lifestyle integration.

Deconstruction of the Latent Dimensions

Although implemented as a parsimonious unidimensional construct in structural models, the underlying psychological reality of continued device use synthesizes three complementary behavioral facets:

• Routine Entrenchment: The degree to which the device transitions from an intrusive or salient artifact into an invisible, background element of daily behavioral scripts. Routine entrenchment occurs when interacting with the device ceases to be a deliberate, high-effort choice and instead becomes an automatic response to environmental cues (e.g., asking a voice assistant for the weather upon stepping into the kitchen).
• Psychological Commitment and Future Projection: The cognitive calculus wherein the consumer anticipates their future behavioral patterns and consciously dismisses the likelihood of abandoning or replacing the technology. This involves perceived switching costs, cognitive lock-in, and the mental projection that life without the device would involve friction or reduced efficiency.
• Perceived Indispensability: An affective-cognitive appraisal that the technology possesses sustained functional and psychological utility. Unlike tools used for single tasks (e.g., a software installation wizard), a continued-use device provides recurrent value across shifting daily contexts.

In the framework of Guha et al. (2023), this construct acts as the ultimate downstream outcome influenced by perceived artificiality (the linguistic and acoustic manifestation of non-humanness) and perceived intelligence. If an AI device exhibits an uncanny mismatch between its artificiality and its capacity to solve tasks, the cognitive dissonance generated directly impairs the consumer’s willingness to grant the device permanent status in their household routine.

6. Theoretical Framework

The conceptual architecture of the Continued Use of the Device scale rests at the intersection of several influential theoretical paradigms within cognitive psychology, marketing science, and information systems research.

1. Expectation-Confirmation Model of IS Continuance (ECM)

Formulated by Anol Bhattacherjee (2001), the Expectation-Confirmation Model (ECM) posits that an individual’s intention to continue using an information technology is governed primarily by their post-adoption satisfaction and their perceived usefulness of the system, which are mediated by the confirmation of pre-adoption expectations. Initial adoption is often speculative; however, continued use represents an informed, experienced judgment. Guha et al. (2023) draw heavily upon this paradigm: when consumers encounter a voice assistant, they develop initial expectations regarding its intelligence. If the voice assistant confirms or exceeds these operational expectations without triggering psychological discomfort, continued use intentions are established.

2. The Computers Are Social Actors (CASA) Paradigm

Pioneered by Byron Reeves and Clifford Nass, the CASA paradigm demonstrates that humans instinctively apply social rules, expectations, and heuristics to interactive computing entities. In voice-based smart devices, anthropomorphic cues—such as a naturalistic human voice—lead consumers to treat the interface as a quasi-social interaction partner. When an interface successfully balances artificiality and perceived capability, it minimizes perceived communication breakdown, facilitating the user’s willingness to maintain an ongoing interactive relationship with the device over time.

3. Mind Perception Theory and Anthropomorphism

Psychological research on mind perception (Gray, Gray, & Wegner, 2007) categorizes attributions of mind into two foundational dimensions: agency (the capacity for self-control, planning, intelligence, and moral responsibility) and experience (the capacity to feel sensations, hunger, pain, and joy). Guha et al. (2023) demonstrate that voice assistants are evaluated primarily on their perceived functional intelligence (agency). When consumers perceive high agency, they feel confident integrating the machine into complex daily routines. Conversely, excessive artificiality that signals incompetence undermines the device’s perceived utility, triggering cognitive fatigue and eventual usage termination.

4. Habit Formation and Cue-Routine-Reward Loops

Grounding continued use in behavioral learning theories (Wood & Neal, 2009; Limayem et al., 2007), digital devices achieve continued usage status only when contextual cues (e.g., waking up, cooking) reliably trigger an automatic execution of device commands that result in psychological or practical rewards (e.g., music playback, home automation). The Continued Use scale captures the respondent’s meta-cognitive awareness of this emerging or stabilized habit loop.

7. Validity

Extensive psychometric validation across multiple empirical investigations confirms that the Continued Use of the Device scale demonstrates robust construct, convergent, discriminant, and predictive validity.

Construct and Convergent Validity

Convergent validity evaluates whether the operational items correlate strongly with theoretically aligned constructs. In the empirical studies conducted by Guha et al. (2023), the scale exhibited high, statistically significant factor loadings ($> .80$) on its primary latent dimension. The Average Variance Extracted (AVE) routinely exceeded the recommended threshold of .50 (frequently reaching values between .68 and .82), demonstrating that the majority of variance in the operational indicators is accounted for by the underlying latent continued use construct rather than random measurement error.

Convergent validity is further reinforced by strong positive correlations with related constructs:

• Overall User Satisfaction ($r \approx .65$ to $.78, p < .001$)
• Perceived Utility / Usefulness ($r \approx .60$ to $.74, p < .001$)
• System Trust ($r \approx .55$ to $.70, p < .001$)
• Willingness to Recommend (NPS-type measures) ($r \approx .62$ to $.75, p < .001$)

Discriminant Validity

To confirm that continued use does not merely duplicate general positive affect or momentary device liking, discriminant validity was rigorously evaluated. Utilizing the Fornell-Larcker criterion, the square root of the AVE for the Continued Use scale was shown to be substantially greater than any bivariate correlation between the scale and other latent variables in the structural model (e.g., perceived artificiality, perceived competence, hedonic enjoyment). Furthermore, the modern Heterotrait-Monotrait (HTMT) ratio of correlations remained well below the conservative ceiling of .85, verifying that continued device use is an empirically unique psychological outcome.

Predictive and Nomological Validity

Nomological validity was established through structural equation models confirming hypothesized relationships: high perceived intelligence coupled with optimized acoustic artificiality significantly elevated continued use intentions. In longitudinal and scenario-based follow-ups, scores on the Continued Use scale successfully predicted actual subsequent device interactions, self-reported weekly frequency of interaction, and consumer willingness to pay subscription fees for continuous software updates.

8. Reliability

The Continued Use of the Device instrument possesses high internal consistency across diverse sample populations, including digital natives, mainstream consumer panels (e.g., Prolific, Amazon Mechanical Turk), and verified smart device owners.

Internal Consistency Metrics

In the empirical testing presented by Guha et al. (2023) across multiple experimental conditions examining voice assistants, the internal reliability statistics were consistently robust:

• Cronbach’s Alpha ($lpha$): Values ranged between .88 and .94 across different experimental studies, well above the conventional benchmark of .70 recommended by Nunnally and Bernstein (1994).
• Composite Reliability ($
  ho_c$ / McDonald’s $\omega$): Ranged between .89 and .95, confirming that the scale indicators reflect the underlying latent construct without distortion from varying error variances.
• Inter-Item Correlation: Mean inter-item correlation coefficients were consistently within the ideal .50 to .75 interval, indicating adequate conceptual cohesion without excessive item redundancy.

Test-Retest and Cross-Context Stability

When evaluated across parallel experimental arms (manipulating voice artificiality from synthetic-robotic to human-recorded speech, and intelligence from basic information retrieval to complex scheduling), the measurement properties displayed strong structural stability. Measurement invariance testing (configural, metric, and scalar invariance) confirmed that respondents interpreted the scale items consistently across varying experimental stimuli ($\Delta \text{CFI} < .01$, $\Delta \text{RMSEA} < .015$), demonstrating that changes in mean scores reflect true variance in consumer intention rather than shifting measurement artifacts.

9. Factor Analysis

The structural characteristics of the Continued Use of the Device scale have been examined using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).

Exploratory Factor Structure

During initial exploratory validation phases, maximum likelihood estimation with oblimin and varimax rotations was applied to the battery of post-interaction evaluation items. A single dominant eigenvalue exceeding 1.0 (typically > 3.20) emerged, accounting for upwards of 75% to 85% of the total variance among the items. Scree plot analyses consistently evidenced a clear break after the first component, verifying unidimensionality. Factor loadings for individual items were uniformly high, consistently exceeding .80, with negligible cross-loadings onto non-target factors such as novelty, general aesthetics, or price perception.

Confirmatory Factor Analysis (CFA) Fit Indices

When subjected to CFA using maximum likelihood estimation in structural equation modeling packages (e.g., AMOS, Mplus, R lavaan), the unidimensional measurement model demonstrated outstanding goodness-of-fit indices across published datasets:

• Chi-Square to Degrees of Freedom Ratio ($\chi^2 / df$): Consistently between 1.15 and 2.45, falling comfortably below the accepted threshold of 3.0.
• Comparative Fit Index (CFI): Ranged from .975 to .998, indicating excellent incremental model fit against the baseline null model.
• Tucker-Lewis Index (TLI): Ranged from .968 to .994, exceeding the stringent .95 criterion.
• Root Mean Square Error of Approximation (RMSEA): Ranged from .031 to .058, with 90% confidence intervals bounded below .08, denoting minimal residual approximation error.
• Standardized Root Mean Square Residual (SRMR): Recorded values between .018 and .032, well within the desired threshold of < .05.

Item standardized factor loadings ($lambda$) under the confirmatory paradigm were highly significant ($p < .001$), typically clustering between .84 and .93, providing structural confirmation of the unidimensional scale model.

10. Instrument / Measurement Tool

The Continued Use of the Device scale is configured as a compact, self-administered survey battery designed for rapid deployment in experimental lab settings, online consumer panels, or field surveys.

Structural Specifications

• Instrument Type: Self-report psychological / consumer behavior measurement scale.
• Target Population: Individuals who have interacted with, tested, or owned a technology hardware device, smart assistant, software interface, or automated service machine.
• Number of Items: Typically 3 to 4 tightly focused items (standard parsimonious form).
• Administration Modality: Digital survey (Qualtrics, Decipher, Google Forms) or paper-and-pencil inventory. Average completion time is under 90 seconds.
• Response Scale: Standard 7-point Likert scale formatted as:
  • 1 = Strongly Disagree
  • 2 = Disagree
  • 3 = Somewhat Disagree
  • 4 = Neither Agree nor Disagree (Neutral)
  • 5 = Somewhat Agree
  • 6 = Agree
  • 7 = Strongly Agree

Scoring and Aggregation

• Item Valence: All standard items are worded in a direct positive direction. (If reverse-coded items are introduced in custom adaptations, they must be recoded: $x_{\text{new}} = 8 – x_{\text{original}}$ on a 7-point scale).
• Index Calculation: An overall Continuance Score is computed by calculating the unweighted arithmetic mean across the completed items:
  $$\text{Continuance Score} = \frac{1}{k}\sum_{i=1}^{k} X_i$$
  where $k$ represents the number of scale items. Alternatively, factor score weights derived from CFA can be utilized in structural equation modeling.
• Score Interpretation:
  • 1.00 – 3.00: Low continuance intention; high attrition risk; the device is perceived as a transient novelty or functionally defective.
  • 3.01 – 4.99: Ambivalent continuance; the user acknowledges selective utility but has not integrated the device into routine behavioral habits.
  • 5.00 – 7.00: Strong routine integration; high psychological commitment; the device is perceived as an indispensable component of everyday life.

11. Permissions & Fee and Test Year

The scale was adapted and published in 2023 in the following peer-reviewed source: Journal of the Academy of Marketing Science (Vol. 51, Issue 4, pp. 843–866), authored by Abhijit Guha, Timna Bressgott, Dhruv Grewal, Dominik Mahr, Martin Wetzels, and Elisa Schweiger.

Usage Rights and Academic Fair Use

• Academic Research: The instrument is available for non-commercial, academic, educational, and scientific research under standard academic fair use principles, provided that the original authors and the primary publication are appropriately cited.
• Commercial Applications: Commercial organizations, corporate market research departments, or enterprise UX auditing bodies seeking to integrate the exact proprietary scales, wording, or derivative corporate monitoring frameworks published by Springer Nature / JAMS should review copyright clearing permissions via the Copyright Clearance Center (CCC) or contact the corresponding author.
• Fees: There is no direct licensing fee for scholarly researchers utilizing the scale in non-profit academic inquiries.

12. References

The following foundational sources document the emergence, theoretical grounding, and psychometric validation of the Continued Use of the Device scale:

• Bhattacherjee, A. (2001). Understanding information systems continuance: An expectation-confirmation model. MIS Quarterly, 25(3), 351–370. https://doi.org/10.2307/3250921
• Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. https://doi.org/10.2307/249008
• Gray, H. M., Gray, K., & Wegner, D. M. (2007). Dimensions of mind perception. Science, 315(5812), 619–619. https://doi.org/10.1126/science.1134475
• Guha, A., Bressgott, T., Grewal, D., Mahr, D., Wetzels, M., & Schweiger, E. (2023). How artificiality and intelligence affect voice assistant evaluations. Journal of the Academy of Marketing Science, 51(4), 843–866. https://doi.org/10.1007/s11747-022-00898-9
• Limayem, M., Hirt, S. G., & Cheung, C. M. (2007). How habit limits the predictive power of intention: The case of information systems continuance. MIS Quarterly, 31(4), 705–737. https://doi.org/10.2307/25148817
• Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill.
• Reeves, B., & Nass, C. (1996). The media equation: How people treat computers, television, and new media like real people and places. Cambridge University Press.
• Venkatesh, V., Thong, J. Y., & Xu, X. (2012). Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology. MIS Quarterly, 36(1), 157–178. https://doi.org/10.2307/41410412
• Wood, W., & Neal, D. T. (2009). The habitual consumer. Journal of Consumer Psychology, 19(4), 579–592. https://doi.org/10.1016/j.jcps.2009.08.003

13. Items of the Scale