*Important notice: This news reports on an unedited version of an accepted paper and is awaiting final editing. Therefore, the paper should not be regarded as conclusive or treated as established information.
Through a risk-based lens, researchers have examined trust in human-machine collaboration amid growing concerns that technological risks have triggered a trust crisis in service-industry machines.
Study: A risk-based view of trust in human–machine collaboration on cooking robots. Image Credit: SWKStock/Shutterstock.com
Using survey data from 201 respondents in China and partial least squares structural equation modeling (PLS-SEM) analysis, the research identifies antecedents of trust and models how trust, perceived risk, adoption intention, and use behavior relate, with implications for designing trustworthy cooking robots. Their findings were published in Scientific Reports.
Emerging Trust Issues in Automation
Digitalization has reshaped the role of human workers, intensifying the potential competition between humans and machines while opening
possibilities for collaboration that frees humans from routine tasks.
However, adopting emerging technologies has triggered a crisis of trust, in turn raising concerns about deploying collaborative machines in services such as healthcare, finance, and recommendation.
Because of the complexity of artificial intelligence (AI) models, user data may lack transparency and explainability, which fosters suspicion about potentially unethical business models behind these technologies.
Prior work examines trust from an interactional perspective, but has limited applicability to human-machine collaboration involving deep integration of AI and robotics. This study therefore investigated trust in cooking-robot collaboration from a risk-based view.
Modeling Trust in Cooking Robots
This study develops a risk-based model of trust in human-machine collaboration, using cooking robots as its context.
The authors argue that while a risk-based view of trust is already common in technology-adoption research, human-machine collaboration emphasizes win-win outcomes and wellbeing, extending beyond earlier technological boundaries. Therefore, they argue that past efficiency-focused survey models may not fully capture how trust forms in a specific application context.
The model draws on the dichotomous structure of individual and relational trust attributes, further subdividing these into affective trust (empathy), cognitive trust (competence), social trust (reciprocity), and institutional trust (assurance).
Four hypotheses predict that empathy, competence, reciprocity, and assurance are positively related to trust. Further hypotheses link trust positively to adoption intention, adoption intention positively to use behavior, trust negatively to perceived risk, and perceived risk negatively to adoption intention.
Download a free copy of this page here!
Anthropomorphism was included as a control variable to rule out an alternative explanation, not as a focal antecedent. The survey instruments were adapted from prior validated studies and modified for the cooking-robot context, with double translation and a pilot test of 20 participants.
Data came from an online survey of 201 respondents in China, recruited through online communities with support from Chunmi, targeting individuals with prior cooking-robot experience. Analysis used PLS-SEM, chosen because the context is relatively unexamined and the method suits predictive models without strict distributional assumptions.
Empathy, Competence, Reciprocity, and Assurance
Because the data came from a single source, the authors took steps against common method bias: predictor, criterion, and control items were separated and randomized; participants were assured of anonymity; and a statistical test showed the largest factor explained only about 30% of variance, with collinearity values well below the threshold.
The measurement model proved to be valid and reliable. Internal consistency, composite reliability, and average variance extracted all exceeded recommended thresholds, and factor loadings, inter-construct correlations, and a ratio-based test all supported discriminant validity, with no concerning collinearity.
For the structural model, the four antecedents were each statistically positively associated with trust, namely, empathy, competence, reciprocity, and assurance, supporting the first four hypotheses.
Trust was positively associated with adoption intention and negatively associated with perceived risk, and perceived risk was negatively associated with adoption intention, supporting the three hypotheses.
The path from adoption intention to self-reported use behavior was also significant, supporting the final hypothesis. A mediation analysis showed perceived risk partially mediated the association between trust and adoption intention.
Among control variables, anthropomorphism was significantly positively associated with trust and adoption intention, but not with perceived risk or use behavior.
How Trust Shapes Adoption Decisions
The study found that empathy, competence, reciprocity, and assurance are each positively associated with trust in collaborative machines.
Empathy matters because machines that convey care may be trusted as service partners rather than just tools, while competence is important because users judge functionality subjectively. Reciprocity reflects perceived responsiveness within collaborative relationships, and assurance captures institutional and regulatory support, while also warning against blind trust.
Trust was negatively associated with perceived risk, and elevated perceived risk was negatively associated with adoption intention. Perceived risk significantly mediated the link between trust and adoption intention, supporting the risk-based view that trust is tied to adoption both directly and indirectly, through reduced perceptions of uncertainty.
Anthropomorphism was associated with trust and adoption intention, but not with perceived risk or actual use behavior, suggesting favorable attitudes toward human-like design may not translate into behavior.
Building Trust in Collaborative Robotics
This study offers empirical support for a risk-based view of trust in human-machine collaboration, using cooking robots as a case study. Trust in collaborative machines is shaped by individual attributes (empathy and competence) and relational ones (reciprocity and assurance).
Trust is tied to adoption intention both directly and indirectly, through reduced perceived risk, while perceived risk itself discourages adoption. Notably, anthropomorphic design was associated with trust and intention to adopt, but not with actual use behavior, suggesting that favorable attitudes may not translate into action.
These findings carry implications for designing trustworthy cooking robots and for understanding trust formation in service-oriented human-machine collaboration. However, further research involving larger and more diverse populations is needed to establish whether these associations extend to other users and service environments.
Journal Reference
Yu, Y., et al. (2026). A risk-based view of trust in human–machine collaboration on cooking robots. Scientific Reports. DOI:10.1038/s41598-026-72253-z. https://www.nature.com/articles/s41598-026-72253-z.
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.