Customer-facing AI agents are becoming frontline service interfaces, yet firms still face a consequential design question: should these agents look human-like or robot-like? This research examines how AI agent appearance shapes consumers' intentionality attributions in AI-mediated service encounters and how these inferences translate into satisfaction and future AI use intention. Drawing on attribution theory and anthropomorphism research, we propose that human-like and robot-like appearances do not have uniformly positive or negative effects; rather, their effects depend on consumers’ evaluations of service outcomes and their familiarity with AI. A 2 (AI agent appearance: human-like vs. robot-like) × 2 (service outcome: success vs. failure) between-subjects experiment was conducted with 350 U.S. participants in an airline customer-service chatbot context. Results reveal a reversal across outcome evaluations. When consumers experienced service success, robot-like AI elicited higher perceived intentionality than human-like AI, particularly among those with lower AI familiarity. When consumers experienced service failure, human-like AI elicited higher perceived intentionality than robot-like AI, although this effect weakened and reversed among consumers with higher AI familiarity. Perceived intentionality, in turn, increased satisfaction and future AI use intention, supporting a moderated mediation process. These findings reframe AI appearance as an attributional cue rather than a simple humanization strategy and demonstrate that consumers interpret the same AI appearance differently depending on how they evaluate the service encounter. Managerially, the results suggest that firms should not assume that increasing human likeness will universally improve customer responses. Instead, marketers should consider how different customer segments, particularly those varying in AI familiarity, are likely to interpret and respond to AI appearance cues across diverse service experiences.
Accurate assessment of structural resilience under time-varying deterioration and external loading requires explicit consideration of cumulative failure behavior and disruption-dependent system response. To address this need, this study makes two main contributions. First, a new time-dependent system-reliability-based resilience formulation is proposed, in which reliability and redundancy are defined using first-passage events. Unlike instantaneous indices, the proposed first-passage indices preserve the cumulative threshold-crossing behavior of component disruption and subsequent system collapse over a service-life horizon. Second, a damage-aware neural-operator acceleration framework is developed to enable efficient evaluation of the proposed indices. A boundary-emphasized loss is introduced to improve surrogate accuracy near limit-states, where small prediction errors can affect the identification of first-passage events. In addition, a Gaussian-mixture-model-based augmentation strategy is used to improve sample support for selected underrepresented disruption scenarios. The proposed approach is demonstrated using a Daniels load-sharing system and a truss bridge subjected to time-varying loading. The results demonstrate that the proposed formulation and acceleration strategy provide an efficient framework for time-dependent resilience assessment of deteriorating structural systems.
Dynamic pricing has become a widely adopted strategy in the hospitality and tourism industry, yet little is known about how consumers’ political identity influences their perceptions of its fairness. Drawing on data from a cross-sectional survey and a series of experimental studies, we examine how consumers’ political orientation shapes their responses to dynamic pricing practices and uncover that conservative consumers perceive dynamic pricing as fairer and react less negatively compared to their liberal counterparts. We further reveal that the government intervention moderates this effect: liberal consumers react more favorably when government intervention is present, whereas conservatives respond more negatively. These findings underscore the need for hospitality and tourism firms to account for consumers’ political identities in their target market when implementing dynamic pricing strategies, as these identities significantly shape attitudinal and behavioral responses across market segments.
The concordance cosmological model ΛCDM assumes dark energy to be a constant, consistent with early-time observations, evidenced by Planck-ΛCDM analyses. However, in the face of late-time tensions, the nature of dark energy remains a central open problem. Modern precision cosmology offers a potential new window into the nature of dark energy in the w(a)CDM framework, which provides a model-independent prescription for its unknown equation of state w(a). A confrontation of w(a)CDM with data generally constitutes a nonlinear inference problem. We find that w0wa estimates posterior to a fully non-linear w(a) analysis are stabilized by the Baryon Acoustic Oscillation (BAO) constraint on cM=Ωm,0h2, inherited for instance from Planck-ΛCDM analysis of the CMB. This implementation produces w0wa estimates that are invariant under constraint-preserving variations in Ωm,0. In contrast, the early-linearization w0waCDM shows pronounced correlation with Ωm,0 even when preserving cM. We quantify this correlation resulting from the non-commutativity of w0wa estimation and linearization in w(a)CDM. This discrepancy is demonstrated in controlled mock-data experiments. Applied to cosmic chronometer data, w0wa estimates from correlation-free late linearization of w(a)CDM analysis favor w0<−1, whereas w0waCDM favors w0>−1. If the correlation between w0 and Ωm,0 in w0waCDM is interpreted as arising from linearization effects rather than a physical origin, application to DESI DR2 may shift w0 downward, potentially extending to w0<−1, corresponding to increasing dark energy at the present epoch. Alternatively, if the correlation is of hitherto unseen physical origin, the w0waCDM parametrization is self-consistent and no such correction to the DESI inference may be required.
Sooting propensity of alkanes, especially, cycloalkanes and fuel mixtures with cycloalkanes is investigated experimentally over a wide range of carbon numbers from C5 to C16. Their smoke points are measured by both a standard method and a virtual smoke point method. The new improved extrapolation of exponential fitting with the smallest number of data is adopted to get more accurate smoke point than the linear fitting. And, the consistent dataset of smoke points for alkanes are obtained and collected. Threshold sooting indices (TSIs) are estimated by these comprehensive smoke points. All of n-alkanes, iso-alkanes, and cycloalkanes show generally increasing TSI with carbon number, but TSIs of highly branched iso-alkanes and multi-ring cycloalkanes deviate from the linear correlation. Inherent molecular structure of fused-ring cycloalkanes results in abrupt increase in TSI. From this observation, a novel predictive model of TSI for fuel mixtures is composed based on the QSPR regression model, where the fragments of a cyclic ring, (CH2)cyclo, and an aromatic ring, (CH)benzyl, are considered separately to reflect their independent contribution to sooting tendency. The new proposed model predicts accurately TSIs of cycloalkanes as well as the other alkanes. From an error analysis with L2-norm, the model provides better performance than the conventional model by 38% in predicting pure cycloalkanes. And, it predicts well TSIs of binary and ternary fuel mixtures with cycloalkanes and further, TSI of a surrogate fuel for a real fuel, e.g., Jet A-1.
