CoChat
Sign Up
Cited by 965Technology Adoption and User BehaviourAI in Service InteractionsHuman-Automation Interaction and Safety

Trust in AI and Its Role in the Acceptance of AI Technologies

Open access
Choung, Hyesun, David, Prabu, Ross, Arun·2023·International Journal of Human–Computer Interaction
DOI 10.1080/10447318.2022.2050543
Open this paper in CoChat — free accountOpen PDF

Abstract

As AI-enhanced technologies become common in a variety of domains, there is an increasing need to define and examine the trust that users have in such technologies. Given the progress in the development of AI, a correspondingly sophisticated understanding of trust in the technology is required. This paper addresses this need by explaining the role of trust in the intention to use AI technologies. Study 1 examined the role of trust in the use of AI voice assistants based on survey responses from college students. A path analysis confirmed that trust had a significant effect the on intention to use AI, which operated through perceived usefulness and participants’ attitude toward voice assistants. In Study 2, using data from a representative sample of the U.S. population, different dimensions of trust were examined using exploratory factor analysis, which yielded two dimensions: human-like trust and functionality trust. The results of the path analyses from Study 1 were replicated in Study 2, confirming the indirect effect of trust and the effects of perceived usefulness, ease of use, and attitude on intention to use. Further, both dimensions of trust shared a similar pattern of effects within the model, with functionality-related trust exhibiting a greater total impact on usage intention than human-like trust. Overall, the role of trust in the acceptance of AI technologies was significant across both studies. This research contributes to the advancement and application of the TAM in AI-related applications and offers a multidimensional measure of trust that can be utilized in the future study of trustworthy AI.

AI snapshot

Written from the paper’s full text.

TL;DR

Two studies extend the technology acceptance model (TAM) by adding trust in AI, finding that trust indirectly affects intention to use AI through perceived usefulness and attitude. Study 2 identifies two trust dimensions—human-like trust and functionality trust—with functionality trust showing a greater total impact on usage intention than human-like trust.

Key findings

  • In Study 1 (N = 312 college students), the path model explained 52% of variance in usage intention (R² = .52); attitude was the strongest direct predictor of behavioral intention (β = .62, p < .001), and trust had a significant indirect effect through perceived usefulness and attitude. Total effects on usage intention were perceived ease of use β = .63, perceived usefulness β = .46, and trust β = .32 (all p < .001).
  • In Study 2 (N = 640 representative U.S. adults), exploratory factor analysis yielded two trust dimensions: human-like trust in AI (six items) and functionality trust in AI (five items). The TAM results replicated: attitude predicted intention (β = .65, p < .001), perceived ease of use had total effects of β = .71 in both models, and perceived usefulness had total effects of β = .57 and β = .56 (all p < .001).
  • Both trust dimensions had significant indirect effects on intention to use smart technologies, but functionality-related trust had a greater total impact on usage intention (β = .42, p < .001) than human-like trust (β = .36, p < .001).

Methods

  • Study 1: online survey conducted October–December 2020 with 312 undergraduate students from a large Midwest university (56% female; ages 18–29, median 21; 96% had used at least one AI voice assistant). Path analysis was conducted in R using lavaan with maximum likelihood estimation and bootstrapped 95% confidence intervals (k = 1,000).
  • Study 2: online survey conducted April 2021 using a national Qualtrics panel with quota sampling for gender, age, and race based on U.S. Census data; 640 respondents participated (50.1% female, mean age = 46.43, SD = 17.83; 65.9% White, 12% Black or African American, 12% Hispanic, 5.9% Asian, 4.4% other). Trust items were factor analyzed, yielding two dimensions, and two separate path models tested human-like trust and functionality trust using the same analytic approach as Study 1.
  • Measures: perceived ease of use, perceived usefulness, attitude, and behavioral intention were adapted from TAM scales; Study 1 trust in AI voice assistants used four created items; Study 2 trust items were constructed around benevolence/helpfulness, integrity/reliability, and competence/functionality.

Limitations

  • Stated in the paper: the studies focused on consumer AI products (voice assistants and smart technologies) associated with lower risk, so findings may not generalize to high-stakes AI such as self-driving cars or healthcare AI; trust is dynamic and may differ across stages such as initial versus ongoing trust; the path models are correlational and cannot guarantee causal effects; and online panel respondents may not be truly representative because they have computer and internet access, skewing toward higher technology acceptance.
  • Evident from the text: Study 1 used a convenience sample of college students, limiting generalizability; both studies relied on cross-sectional self-report survey data, so causal direction among trust, perceived usefulness, attitude, and intention cannot be established from these data alone.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1080/10447318.2022.2050543

Cites

62 of 62 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

0 of 33 cited sources checked · 0 supported · 0 flagged

  • Findingown data
    § [Figure 2 About Here]

    The perceived usefulness of an AI voice assistant positively influenced users' attitude toward it (β=.42, p<.001).

    “The perceived usefulness of an AI voice assistant also positively influenced users' attitude toward it ( 𝛽 = .42, p < .001)”
    Quote in the paperSubstantial claim
  • Findingown data
    § Discussion

    The validated model presents a high explanatory power, with 52% of variance explained (R2=.52).

    “The validated model presents a high explanatory power, with 52% of variance explained ( R 2 = .52).”
    Quote in the paperNot a substantial claim
  • Findingown data
    § Results

    The path model for AI voice assistant usage intention demonstrated a good fit with the data (χ2(1)=.32, p=.58; RMSEA=.00 [.00,.12]; SRMR=.01; CFI=1.00; TLI=1.01).

    “the path model demonstrated a good fit with the data: 𝜒 2 (1) = .32, p = .58; RMSEA = .00 [.00, .12]; SRMR = .01; CFI = 1.00; TLI = 1.01.”
    Quote in the paperSubstantial claim
  • Findingown data
    § [Figure 2 About Here]

    Perceived ease of use was positively associated with trust perception of a voice assistant (β=.39, p<.001).

    “trust perception ( 𝛽 = .39, p < .001)”
    Quote in the paperSubstantial claim
  • Findingown data
    § [Figure 2 About Here]

    Trust positively predicted perceived usefulness of AI voice assistants.

    “Moreover, trust positively predicted perceived usefulness.”
    Quote in the paperSubstantial claim
  • May combine two claims. 38
  • Classified as probably not a substantial claim. 39
  • The quote only partly states this claim. 3

  • May combine two claims.

    Findingown data
    § Results

    All other measures of fit (CFI above .95, SRMR below .05, and RMSEA below .08) showed that the models were adequate in reproducing covariances among the variables.

    “All other measures of fit (CFI above .95, SRMR below .05, and RMSEA below .08) showed that the models were adequate in reproducing covariances among the variables.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Results

    The models for human trust in AI and technology-related trust in AI fit the data well (human trust: χ2(1)=4.92, p=.03; RMSEA=.08 [.02,.15]; SRMR=.01; CFI=1.00; TLI=.98; technology-related trust: χ2(1)=4.85, p=.03; RMSEA=.08 [.02,.15]; SRMR=.01; CFI=1.00; TLI=.98).

    “The models fit the data well: human trust in AI, 𝜒 2 (1) = 4.92, p = .03; RMSEA = .08 [.02, .15]; SRMR = .01; CFI = 1.00; TLI = .98; technology-related trust in AI, 𝜒 2 (1) = 4.85, p = .03 RMSEA = .08 [.02, .15]; SRMR = .01; CFI = 1.00; TLI = .98.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Abstract

    The path analysis results from Study 1 were replicated in Study 2, confirming the indirect effect of trust and the effects of perceived usefulness, ease of use, and attitude on intention to use.

    “The results of the path analyses from Study 1 were replicated in Study 2, confirming the indirect effect of trust and the effects of perceived usefulness, ease of use, and attitude on intention to use.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § Analytic Approach

    The constructs in Study 1 are moderately correlated.

    “The values in Table 2 indicate that the constructs are moderately correlated.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § [Table 4 Near Here]

    Trust predicted positive attitudes, which in turn influenced usage intention (H4b).

    “Trust also predicted positive attitudes, which in turn influenced usage intention (H4b).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Table 4 Near Here]

    Perceived ease of use had the largest total effect on usage intention (β=.63; p<.001), followed by perceived usefulness (β=.46; p<.001) and trust perception (β=.32; p<.001).

    “perceived ease of use had the largest total effect on usage intention ( 𝛽 = .63; p < .001), followed by perceived usefulness ( 𝛽 = .46; p < .001) and trust perception ( 𝛽 = .32; p < .001).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Discussion

    Trust can influence and is influenced by the factors included in the TAM.

    “In addition, we found that trust can influence and is influenced by the factors included in the TAM.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Table 4 Near Here]

    Trust was associated with increased perceived usefulness, which in turn increased usage intention (H4a).

    “Trust was associated with increased perceived usefulness, which in turn increased usage intention (H4a).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Measures

    Factor analysis of trust items yielded a two-factor structure: items measuring benevolence and integrity were grouped together, and competence was treated as another dimension.

    “Items from the three constructs were factor analyzed, which yielded a two-factor structure, based on a principal components exploratory factor analysis (EFA) with an Oblimin rotation. The items measuring benevolence and integrity were grouped together, and competence was treated as another dimension.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Figure 2 About Here]

    Perceived ease of use, perceived usefulness, and attitude were positive predictors of AI voice assistant usage intention.

    “Perceived ease of use, perceived usefulness, and attitude were positive predictors of usage intention.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Table 4 Near Here]

    All mediated paths depicted in the model were statistically significant, including the hypothesized H5 paths.

    “The results of indirect effects revealed that all mediated paths depicted were statistically significant, including the hypothesized (H5) paths (see Table 3).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Abstract

    Both dimensions of trust shared a similar pattern of effects within the model, with functionality-related trust exhibiting a greater total impact on usage intention than human-like trust.

    “Further, both dimensions of trust shared a similar pattern of effects within the model, with functionality-related trust exhibiting a greater total impact on usage intention than human-like trust.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Table 4 Near Here]

    Trust predicted perceived usefulness, which in turn increased positive attitudes and usage intention (H4c).

    “Moreover, trust predicted perceived usefulness, which in turn increased positive attitudes and usage intention (H4c).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Tables 9 and 10 About Here]

    Participants' intention to use a smart technology was predicated on its perceived ease of use (total effect β=.71; p<.001; β=.71; p<.001) and perceived usefulness (total effect β=.57; p<.001; β=.56; p<.001).

    “In both path models, participants' intention to use a smart technology was predicated on its perceived ease of use (total effect 𝛽 = .71; p < .001; 𝛽 = .71; p < .001) and perceived usefulness (total effect 𝛽 = .57; p < .001; 𝛽 = .56; p < .001).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data
    § [Tables 7 and 8 About Here]

    Perceived ease of use was positively associated with both trust dimensions (β=.56, p<.001; β=.51, p<.001).

    “both trust dimensions ( 𝛽 = .56, p < .001; 𝛽 = .51, p < .001)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § Discussion

    The two dimensions of trust in AI-the human-like trust dimension and the functionality dimension-significantly predict perceived usefulness and positive attitude toward smart technology, which in turn predict greater usage intention.

    “Additionally, the results indicate that the two dimensions of trust in AI-the human-like trust dimension and the functionality dimension -significantly predict perceived usefulness and positive attitude toward smart technology, which in turn, predict greater usage intention.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Tables 9 and 10 About Here]

    The two types of trust were associated with perceived usefulness, which predicted greater usage intention (H4a).

    “The two types of trust were associated with perceived usefulness, which predicted greater usage intention (H4a).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Tables 9 and 10 About Here]

    Trust dimensions predicted perceived usefulness, which in turn increased positive attitudes and usage intention (H4c).

    “Finally, trust dimensions predicted perceived usefulness, which in turn increased positive attitudes and usage intention (H4c).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Tables 9 and 10 About Here]

    Trust predicted positive attitudes, which in turn affected usage intention (H4b).

    “Trust also predicted positive attitudes, which in turn affected usage intention (H4b).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § General Discussion

    In Study 1 and Study 2, perceived ease of use and perceived usefulness were essential determinants for the use of AI technologies, although perceived ease of use had a consistently greater impact on the acceptance of these technologies.

    “In Study 1 and Study 2, we found that perceived ease of use and perceived usefulness were essential determinants for the use of AI technologies, although perceived ease of use had a consistently greater impact on the acceptance of these technologies.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § General Discussion

    The human-like trust dimension was a significant predictor of the acceptance of technology, as was the functionality dimension.

    “This dimension was a significant predictor of the acceptance of technology, as was the functionality dimension.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data
    § [Tables 9 and 10 About Here]

    Both the human-like and the functionality dimensions of trust had statistically significant indirect effects on the intention to use AI smart technologies.

    “both the human-like and the functionality dimensions of trust had statistically significant indirect effects on the intention to use AI smart technologies, thus supporting H5.”
    Quote in the paperSubstantial claim

  • The quote only partly states this claim.

    Findingown data
    § Tables and Figures

    In Study 2, exploratory factor analysis of the Trust in AI questionnaire yielded two factors: human-like trust in AI and functionality trust in AI.

    “Factor 1: Human-like trust in AI”
    • Also stated in: § Tables and Figures · plain
      “Factor 2: Functionality trust in AI”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation
    § [Table 1 About Here]

    Trust in technology is a dynamic concept that is still evolving and variable based on the context and the characteristics of the trusted agent.

    “This finding suggests that trust in technology is a dynamic concept that is still evolving and variable based on the context and the characteristics of the trusted agent.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation
    § Discussion

    Current voice assistants are designed to be more human-like than previous attempts, and many individuals are communicating with voice assistants as part of their everyday life in the same way they interact with other humans.

    “Also, current voice assistants are designed to be more human-like than previous attempts, and many individuals are communicating with voice assistants as part of their everyday life in the same way they interact with other humans (Sundar et al., 2017), making trust more relevant.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Sundar et al., 2017

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Premisereasoning
    § [Table 1 About Here]

    Unlike traditional technologies that rely on user input and the execution of rules programmed by a human, AI implies certain levels of autonomy.

    “Unlike traditional technologies that rely on user input and the execution of rules programmed by a human, AI implies certain levels of autonomy.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

    Premisenone given
    § [Table 1 About Here]

    Trust is vital to understanding interpersonal interaction, as it contributes to perceived reliability and integrity.

    “Trust is vital to understanding interpersonal interaction, as it contributes to perceived reliability and integrity.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisenone given
    § [Table 1 About Here]

    The two main dimensions of trust in AI are human-like trust in AI and functionality trust in AI.

    “In the current study, building on previous work regarding trust, we focus on the two main dimensions of trust in AI: 1) human-like trust in AI, and 2) functionality trust in AI.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation
    § Trust in AI and its Role in the Acceptance of AI Technologies

    Trust has been examined as a key factor in the acceptance of emerging technologies.

    “trust has been examined as a key factor in the acceptance of emerging technologies (K. Wu et al., 2011).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites K. Wu et al., 2011

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation
    § Discussion

    The finding that trust perception with perceived ease of use and perceived usefulness predicts attitudes and usage intention is in line with studies that have tested the impact of these constructs on the adoption of various forms of technology.

    “The finding is in line with studies that have tested the impact of these constructs on the adoption of various forms of technology (e.g., Beldad & Hegner, 2018; Choi & Ji, 2015).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Beldad & Hegner, 2018

      Not checked in the free run.

    • cites Choi & Ji, 2015

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisereasoning
    § [Table 1 About Here]

    Functionality trust in AI pertains to the competency and expertise of the technological features.

    “The latter dimension pertains to the competency and expertise of the technological features.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation
    § Measures

    Perceived ease of use and perceived usefulness were measured with five items each adapted from the original TAM scale.

    “The constructs 'perceived ease of use' and 'perceived usefulness' were measured with five items, each adapted from the original TAM scale (Davis, 1989).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Davis, 1989

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning
    § Study 1

    The anthropomorphic nature of voice assistants could make trust more critical.

    “The anthropomorphic nature of voice assistants could make trust more critical.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning
    § Method Participants and Procedure

    To ensure representativeness of participants, a quota sampling method was adopted that set quotas for gender, age, and race based on U.S. Census data.

    “To ensure the representativeness of participants, we adopted a quota sampling method that set quotas for gender, age, and race based on U.S. Census data.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning
    § Discussion

    To motivate individuals to adopt or continue to use an AI voice assistant, they should find the technology easy to use and useful.

    “This finding suggests that to motivate individuals to adopt or continue to use an AI voice assistant, they should find the technology easy to use and useful.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § [Table 1 About Here]

    Past studies have applied interpersonal trust to human technology relationships, especially when the technology has human-like characteristics.

    “Past studies have applied interpersonal trust to human technology relationships (Calhoun et al., 2019), especially when the technology has human-like characteristics (Gillath et al., 2021).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Calhoun et al., 2019

      Not checked in the free run.

    • cites Gillath et al., 2021

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § [Table 1 About Here]

    For human-like trust in AI, trust is in the agent, not in the specific actions and operations of the agent.

    “For human-like trust in AI, trust is in the agent, not in the specific actions and operations of the agent (J. D. Lee & See, 2004).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites J. D. Lee & See, 2004

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § Method Participants and Procedure

    Ninety-six percent of participants reported that they had used at least one type of AI voice assistant.

    “Ninety-six percent of participants reported that they had used at least one type of AI voice assistant.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § [Table 1 About Here]

    Trust in automation is summarized in three categories: performance, process, and purpose.

    “Similar to trust in people and trust in technology, trust in automation is summarized in three categories: 1) performance, 2) process, and 3) purpose.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lee and See (2004)

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § Study 1

    Voice assistants are designed to be more human-like, with their voice feature encouraging users to adopt the same social responses that they habitually use in their interpersonal relationships.

    “Voice assistants are designed to be more human-like (McLean & Osei-Frimpong, 2019), with their voice feature encouraging users to adopt the same social responses that they habitually use in their interpersonal relationships (Reeves & Nass, 1996).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites McLean & Osei-Frimpong, 2019

      Not checked in the free run.

    • cites Reeves & Nass, 1996

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data
    § Discussion

    Trust is pivotal in improving user perception of the utility of voice assistants and in building positive attitudes toward them.

    “Although trust does not directly affect usage intention, the construct appears to be pivotal in improving user perception of the utility of voice assistants and in building positive attitudes toward them.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductcitation
    § Measures

    Behavioral intention to use or continue to use an AI voice assistant was measured with three items from the TAM 2 scale.

    “Behavioral intention to use or continue to use an AI voice assistant was measured with three items from the TAM 2 scale (Venkatesh & Davis, 2000).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Venkatesh & Davis, 2000

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisenone given
    § [Table 1 About Here]

    Human-like trust in AI pertains to the social and cultural values of the algorithms and the values and ethics that undergird the design of the AI technology.

    “The former dimension pertains to the social and cultural values of the algorithms and the values and ethics that undergird the design of the AI technology.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation
    § [Table 1 About Here]

    In the context of AI, both trust in people and trust in technology are relevant, since AI technologies are often depicted as capable of human qualities, including reasoning and motivations, which can induce high expectations and initial trust.

    “We consider that in the context of AI, both trust in people and trust in technology are relevant, since AI technologies are often depicted as being capable of human qualities, including reasoning and motivations, which can induce high expectations and initial trust (Glikson & Woolley, 2020).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Glikson & Woolley, 2020

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § [Table 1 About Here]

    Trust in technology is qualitatively different from trust in people.

    “McKnight et al. (2011) explained that trust in technology is qualitatively different from trust in people.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites McKnight et al. (2011)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § Trust in AI and its Role in the Acceptance of AI Technologies

    The TAM is a theoretical framework for explaining technology usage behavior that has been validated in different technologies across various populations.

    “The TAM is a theoretical framework for explaining the technology usage behavior that has been validated in different technologies across various populations”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Venkatesh & Davis

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data
    § Method Participants and Procedure

    The sample consisted of 640 respondents (50.1% female, age M=46.43, SD=17.83, 65.9% White, 12% Black or African American, 12% Hispanic, 5.9% Asian, 4.4% other).

    “A total of 640 respondents (50.1% female, age M = 46.43, SD = 17.83, 65.9% White, 12% Black or African American, 12% Hispanic, 5.9% Asian, 4.4% other) participated in the study.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Discussion

    Glikson and Woolley (2020) organized the literature on trust in AI based on the factors that shaped users' cognitive and emotional trust.

    “For instance, Glikson and Woolley (2020) organized the literature on trust in AI based on the factors that shaped users' cognitive and emotional trust.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Glikson & Woolley, 2020

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § References

    Trust in automation has three bases: performance, process, and purpose.

    “| Basis of trust in people Mayer et al., (1995) | Basis of trust in technology Lankton et al. (2015) | Basis of trust in automation Lee & Moray (1992) |”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lee & Moray (1992)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given
    § Discussion

    Human-like trust may explain emotional trust and emotional attachment to technology.

    “Further, human-like trust may explain emotional trust and emotional attachment to technology.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data
    § General Discussion

    Our findings provide further support for the utility of the TAM in explaining the acceptance of AI technologies and the added contribution of trust within the model.

    “Our findings provide further support for the utility of the TAM in explaining the acceptance of AI technologies and the added contribution of trust within the model.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation
    § Discussion

    The distinctive perceptions of human-like trust and functionality trust align with studies that propose multiple dimensions and factors of trust in AI and automation.

    “The distinctive perceptions of human-like trust and functionality trust align with studies that propose multiple dimensions and factors of trust in AI and automation (Glikson & Woolley, 2020; Hoff & Bashir, 2015; J. D. Lee & See, 2004).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Glikson & Woolley, 2020

      Not checked in the free run.

    • cites Hoff & Bashir, 2015

      Not checked in the free run.

    • cites J. D. Lee & See, 2004

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § General Discussion

    Trust in AI can be built through embedding specific principles in the design and governance of AI, as suggested in recent publications using the fairness, accountability, transparency, explainability framework.

    “As past research has already confirmed, trust in AI can be built through embedding specific principles in the design and governance of AI, as suggested in recent publications using the fairness, accountability, transparency, explainability framework (Shin, 2020b; Shin & Park, 2019).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Shin, 2020b

      Not checked in the free run.

    • cites Shin & Park, 2019

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § Limitations and Future Research Directions

    An opposite direction of the relationship between trust and perceived usefulness was also found to be valid in another study.

    “Still, an opposite direction of the relationship was also found to be valid in another study (Shin, 2020a).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Shin, 2020a

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation
    § Limitations and Future Research Directions

    Trust was proposed as an antecedent of perceived usefulness based on the findings from the past TAM literature.

    “We proposed trust as an antecedent of perceived usefulness based on the findings from the past TAM literature (K. Wu et al., 2011).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites K. Wu et al., 2011

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Study design and conductnone given
    § Limitations and Future Research Directions

    The study focused on AI technology applications in the form of consumer products (i.e., voice assistants and smart technologies), which tend to be associated with lower risks.

    “First, we focused on AI technology applications in the form of consumer products (i.e., voice assistants and smart technologies), which tend be associated with lower risks.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § General Discussion

    Hoff and Bashir (2015) suggested three layers of trust in automation (dispositional trust, situational trust, and learned trust) based on their review of trust in AI.

    “Hoff and Bashir (2015) suggested three layers of trust in automation (dispositional trust, situational trust, and learned trust) based on their review of trust in AI.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Hoff & Bashir, 2015

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning
    § Discussion

    The duality of AI as both a functional and social technology must be considered when designing trustworthy AI.

    “It appears the duality of AI as both a functional and social technology must be considered when designing trustworthy AI.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § Discussion

    The cognitive construct involves the rational evaluation of AI while emotional trust is typically influenced by irrational factors.

    “The cognitive construct involves the rational evaluation of AI while emotional trust is typically influenced by irrational factors.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Glikson & Woolley, 2020

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationreasoning
    § General Discussion

    The consistently greater impact of perceived ease of use on acceptance may be because consumers tend to believe that AI technologies are complicated and difficult to use.

    “This impact may be because consumers tend to believe that AI technologies are complicated and difficult to use.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § References

    Trust in people has three bases: competence/ability, integrity, and benevolence.

    “| Basis of trust in people Mayer et al., (1995) | Basis of trust in technology Lankton et al. (2015) | Basis of trust in automation Lee & Moray (1992) |”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Mayer et al., (1995)

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data
    § General Discussion

    When people consider an AI technology to be easy to use, they are more likely to trust the technology and consider it useful.

    “The model explains that when people consider an AI technology to be easy to use, they are more likely to trust the technology and consider it useful.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation
    § Discussion

    These findings are consistent with Study 1 and prior literature.

    “These findings are consistent with Study 1 and prior literature (Gefen et al., 2003; I.-L. Wu & Chen, 2005).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Gefen et al., 2003

      Not checked in the free run.

    • cites I.-L. Wu & Chen, 2005

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § References

    Trust in technology has three bases: functionality, reliability, and helpfulness.

    “| Basis of trust in people Mayer et al., (1995) | Basis of trust in technology Lankton et al. (2015) | Basis of trust in automation Lee & Moray (1992) |”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Lankton et al. (2015)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The human-like trust in smart technologies scale (Study 2) had a Cronbach's alpha of .92.

    “N/A | 𝛼 = .92 |”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The attitude toward AI technologies scale had a Cronbach's alpha of .89 in Study 1 and .90 in Study 2.

    “𝛼 = .89 | 𝛼 = .90”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The perceived usefulness scale had a Cronbach's alpha of .92 in Study 1 and .90 in Study 2.

    “𝛼 = .92 | 𝛼 = .90”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The behavioral intention of non-users scale had a Cronbach's alpha of .98 in Study 1 and .91 in Study 2.

    “𝛼 = .98 | 𝛼 = .91”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The trust in the voice assistant scale (Study 1) had a Cronbach's alpha of .81.

    “𝛼 = .81 | N/A”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The perceived ease of use scale had a Cronbach's alpha of .90 in Study 1 and .91 in Study 2.

    “𝛼 = .90 | 𝛼 = .91”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Study design and conductown data
    § Results of Path Analysis (Study 1)

    The functionality trust in smart technologies scale (Study 2) had a Cronbach's alpha of .91.

    “N/A | 𝛼 = .91 |”
    Quote in the paperNot a substantial claim
71 of 71

  • The paper relies on this without support.

    Perception of fairness and justice is associated with human trust in AI.

    “Moreover, the perception of fairness and justice will be associated with human trust in AI.”

  • The paper relies on this without support.

    Trust is relevant throughout the lifecycle of a technology.

    “Indeed, trust is relevant throughout the lifecycle of a technology.”

  • The paper relies on this without support.

    Users who participate in an experience with technology build ideas about whether it is useful and easy to use based on their experience of use.

    “Users who participate in an experience with technology build ideas about whether it is useful and easy to use based on their experience of use.”

  • The paper relies on this without support.

    Trust is a dynamic concept that varies depending on the stage of technology.

    “Second, trust is a dynamic concept that varies depending on the stage of technology.”

  • The paper relies on this without support.

    Participants in an online survey are likely to have greater access to and experience with various forms of information technology.

    “Participants in an online survey are likely to have greater access to and experience with various forms of information technology.”
5 of 5

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Trust significantly influences reliance intentions.

    “Regarding reliance, Choung et al. (2023) demonstrated that trust significantly influences reliance intentions and Kahr et al. (2024) found that trust and reliance on AI advice were significantly higher when the AI model was highly accurate and that trust increased over repeated interactions.”

    From Investigating appropriate reliance on AI-Based decision support systems: the role of expertise, trust, and self-confidence (2025) · cites this paper as Choung et al., 2023

    The results near Table 4 report a statistically significant total effect of trust perception on intention to use AI technologies. The paper measures usage intention, not specifically intention to rely on AI advice, and its Limitations section says the correlational models cannot establish a causal effect. (LLM, conf=0.85)

    Experimental — accuracy not yet measured

  • Trust has a significant impact on perceived ease of use, perceived usefulness, and attitudes toward GenAI.

    “These findings underscore that trust has a significant impact on perceived ease of use, perceived usefulness, and attitudes toward GenAI.”

    From Artificial intelligence and communication technologies in academia: faculty perceptions and the adoption of generative AI (2025) · cites this paper as Choung et al., 2022

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Not confirmedχ²(1) = 0.32reported p = .58 · recomputed p = .572

    “χ 2 (1) = .32, p = .58”

  • Consistentχ²(1) = 4.92reported p = .03 · recomputed p = .027

    “χ 2 (1) = 4.92, p = .03”

  • Consistentχ²(1) = 4.85reported p = .03 · recomputed p = .028

    “χ 2 (1) = 4.85, p = .03”

3 of 3

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Study 2 gives incompatible factor-analysis methods

    Major

    The methods describe principal components extraction with Oblimin rotation, whereas the note to Table 5 specifies principal axis factoring with promax rotation.

  • Study 1 trust-mediated effect is inconsistent with its path coefficients

    Major

    The reported effect of .03 for TRU → PU → ATT → BI is outside its own 95% confidence interval of .06-.16. Multiplying the corresponding standardized path coefficients gives about .10, not .03; this also changes the reported indirect impact of trust.

    Reported 0.03 · recomputed 0.101556

  • One functionality-trust item appears twice with different loadings

    Minor

    Table 5 lists the same questionnaire item twice among the five functionality-trust items, assigning it different factor loadings and communalities. The appendix instead gives a distinct competence item for that position.

  • Trust is said both to have and not to have a direct effect in Study 1

    Minor

    The results say all four predictors have significant direct and indirect effects on usage intention, but the discussion explicitly says trust has no direct effect on usage intention.

4 of 4

Cites

Flagged first, then by citations

Research works printed but not linked: 3

  1. AI HLEG (2019). Ethics guidelines for trustworthy AI. European Commission. https://data.europa.eu/doi/10.2759/346720 · DOI
  2. Shin D, 2021, New Media & Society, 1
  3. Wu J., 2007, Journal of Electronic Commerce Research, 8, 128

Cited by

Sorted by citations

Related

Authors
Choung, H., David, P., & Ross, A.
Affiliations
Michigan State University (US)
Corresponding author
Choung, Hyesun
ORCID
Venue
International Journal of Human–Computer Interaction
Publisher
Informa UK Limited
Published
2022-04-20
Volume
39
Issue
9
Pages
1727-1739
ISSN
  • 1044-7318 (ISSN-L)
  • 1532-7590
DOI
10.1080/10447318.2022.2050543
OpenAlex
W4221014470
Semantic Scholar
f281ad6573ed8ea7fb3116a3747b0e80ef522ea9
Cited by
963
References
62
Citations by year
2026: 402 · 2025: 360 · 2024: 142 · 2023: 54 · 2022: 5
FWCI
117.34
Type
journal article
Language
en
Open access
Locations
  • International Journal of Human-Computer Interaction (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Social Sciences › Decision Sciences › Information Systems and Management › Technology Adoption and User Behaviour
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in Service Interactions
  • Social Sciences › Psychology › Social Psychology › Human-Automation Interaction and Safety
Keywords
trust in AI, technology acceptance, intention to use, voice assistants, perceived usefulness, perceived ease of use, attitudes toward technology, trustworthy AI, technology acceptance model

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-06.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-06.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review