Trustworthiness judgments and pupil-size in individuals with schizophrenia.

Individuals with schizophrenia show aberrant processing of social cues. In the current study, we (1) compared trustworthiness ratings of faces between patients with schizophrenia and healthy controls, (2) compared pupillary reactivity between patients and controls (3) examined whether trustworthiness judgments in schizophrenia are related to pupil reactivity, (4) and examined associations between trustworthiness judgements and symptom severity, specifically paranoia. Patients with schizophrenia spectrum disorders (N = 48) and healthy controls (N = 33) completed a Trustworthiness Task, during which their pupil size was measured via an eye-tracking device. The mean baseline-corrected pupil size was calculated from 24 pictures of real neutral faces, each presented for 2500ms. Self-reported psychotic experiences were measured by Community Assessment of Psychic Functioning (CAPE-42), and symptom severity was rated by Brief Psychiatric Rating Scale (BPRS). No group differences were found in trustworthiness ratings or pupil reactivity parameters during trustworthiness judgments. Separately, among patients, absolute difference in pupil-size change and dilation after reaching minimum size were related to more severe positive symptoms and self-reported paranoia. Our results did not show social cognitive biases in the stable outpatients with schizophrenia, or the role of pupil reactivity in trustworthiness judgments. Future studies should use longer stimuli for pupillary reactivity and control the type and dosage of utilized antipsychotic medication. Further studies are required to explore relationships in larger and more symptomatic groups of patients.

When Trustworthiness Meets Face: Facial Design for Social Robots.

As a technical application in artificial intelligence, a social robot is one of the branches of robotic studies that emphasizes socially communicating and interacting with human beings. Although both robot and behavior research have realized the significance of social robot design for its market success and related emotional benefit to users, the specific design of the eye and mouth shape of a social robot in eliciting trustworthiness has received only limited attention. In order to address this research gap, our study conducted a 2 (eye shape) × 3 (mouth shape) full factorial between-subject experiment. A total of 211 participants were recruited and randomly assigned to the six scenarios in the study. After exposure to the stimuli, perceived trustworthiness and robot attitude were measured accordingly. The results showed that round eyes (vs. narrow eyes) and an upturned-shape mouth or neutral mouth (vs. downturned-shape mouth) for social robots could significantly improve people's trustworthiness and attitude towards social robots. The effect of eye and mouth shape on robot attitude are all mediated by the perceived trustworthiness. Trustworthy human facial features could be applied to the robot's face, eliciting a similar trustworthiness perception and attitude. In addition to empirical contributions to HRI, this finding could shed light on the design practice for a trustworthy-looking social robot.

Why does one trust? A 360-degree perspective on the role of position power in weighting trustworthiness factors.

The purpose of the current paper is to explore the influence of the perceived trustworthiness factors of ability, benevolence and integrity on general and specific trust across supervisors, subordinates and peers with the same focal referent. This study used a 360-degree approach to examine how positional power asymmetries influence the relationship between trustworthiness and trust for a general and situation-specific referent. Data were obtained from military supervisors (N = 200), peers (N = 123), and subordinates (N = 85). Measures of trustworthiness and trust of a common military officer were obtained. Supervisors and subordinates differed in their relative weighting of trustworthiness factors (i.e. ability, benevolence, and integrity) when evaluating general and specific trust. Peers evidenced no difference in the relative weighting of trustworthiness factors. The relationship between benevolence and specific trust was stronger for subordinates than for supervisors. One implication of our findings is that trust can develop differently in bottom-up versus top-down organizational relationships. This study provides evidence that supervisors and subordinates emphasize different aspects of trustworthiness when evaluating their trust of a focal officer, and this process was different for general versus specific trust referents. The study also extends previous research by replicating previous findings across raters.

Assessing reliability and validity of food safety culture assessment tools.

Since its recognition as a plausible direction to assure food safety, food safety culture research has evolved with several commercial and scientific assessment tools developed to evaluate the food safety culture in food businesses. However, existing research does not specify the validity and reliability checks required to demonstrate rigor in the tool development process and there is no unified methodology to confirm robustness of the tools to ensure trustworthiness and usefulness of findings and inferences generated. The purpose of the study was to develop a method to evaluate food safety culture assessment tools and to assess the reliability and validity of existing food safety culture assessment tools using the developed method. Eleven elements were found to be key in validating food safety culture assessment tools. Of the eight tools assessed, only one tool (CT2) was validated on each of the elements. The depth of validation strategies differed for each tool. Three out of the five commercial tools published peer reviewed publications that demonstrated the validation checks that were done. Face validation, and pilot testing were evident and appeared to be done the most. Whilst content, ecological, and cultural validity were the least validated for scientific tools, factor analysis and reliability checks were the least evaluated for commercial tools. None of the tools were assessed for postdictive validity, concurrent validity and the correlation coefficient relating to construct validity. Having an established science-based approach is key as it provides a way to determine the trustworthiness of established assessment tools against accepted methods.

How the EU AI Act Seeks to Establish an Epistemic Environment of Trust.

With focus on the development and use of artificial intelligence (AI) systems in the digital health context, we consider the following questions: How does the European Union (EU) seek to facilitate the development and uptake of trustworthy AI systems through the AI Act? What does trustworthiness and trust mean in the AI Act, and how are they linked to some of the ongoing discussions of these terms in bioethics, law, and philosophy? What are the normative components of trustworthiness? And how do the requirements of the AI Act relate to these components? We first explain how the EU seeks to create an epistemic environment of trust through the AI Act to facilitate the development and uptake of trustworthy AI systems. The legislation establishes a governance regime that operates as a socio-epistemological infrastructure of trust which enables a performative framing of trust and trustworthiness. The degree of success that performative acts of trust and trustworthiness have achieved in realising the legislative goals may then be assessed in terms of statutorily defined proxies of trustworthiness. We show that to be trustworthy, these performative acts should be consistent with the ethical principles endorsed by the legislation; these principles are also manifested in at least four key features of the governance regime. However, specified proxies of trustworthiness are not expected to be adequate for applications of AI systems within a regulatory sandbox or in real-world testing. We explain why different proxies of trustworthiness for these applications may be regarded as 'special' trust domains and why the nature of trust should be understood as participatory.

Implicit bias and experience influence overall but not relative trustworthiness judgment of other-race faces.

Impressions of trustworthiness are formed quickly from faces. To what extent are these impressions shared among observers of the same or different races? Although high consensus of trustworthiness evaluation has been consistently reported, recent studies suggested substantial individual differences. For instance, negative implicit racial bias and low contact experience towards individuals of the other race have been shown to be related to low trustworthiness judgments for other-race faces. This pre-registered study further examined the effects of implicit social bias and experience on trustworthiness judgments of other-race faces. A relatively large sample of White (N = 338) and Black (N = 299) participants completed three tasks: a trustworthiness rating task of faces, a race implicit association test, and a questionnaire of experience. Each participant rated trustworthiness of 100 White faces and 100 Black faces. We found that the overall trustworthiness ratings for other-race faces were influenced by both implicit bias and experience with individuals of the other-race. Nonetheless, when comparing to the own-race baseline ratings, high correlations were observed for the relative differences in trustworthiness ratings of other-race faces for participants with varied levels of implicit bias and experience. These results suggest differential impact of social concepts (e.g., implicit bias, experience) vs. instinct (e.g., decision of approach-vs-avoid) on trustworthiness impressions, as revealed by overall vs. relative ratings on other-race faces.

Uncertainty quantification via localized gradients for deep learning-based medical image assessments.

Objective.Deep learning models that aid in medical image assessment tasks must be both accurate and reliable to be deployed within clinical settings. While deep learning models have been shown to be highly accurate across a variety of tasks, measures that indicate the reliability of these models are less established. Increasingly, uncertainty quantification (UQ) methods are being introduced to inform users on the reliability of model outputs. However, most existing methods cannot be augmented to previously validated models because they are not post hoc, and they change a model's output. In this work, we overcome these limitations by introducing a novel post hoc UQ method, termedLocal Gradients UQ, and demonstrate its utility for deep learning-based metastatic disease delineation.Approach.This method leverages a trained model's localized gradient space to assess sensitivities to trained model parameters. We compared the Local Gradients UQ method to non-gradient measures defined using model probability outputs. The performance of each uncertainty measure was assessed in four clinically relevant experiments: (1) response to artificially degraded image quality, (2) comparison between matched high- and low-quality clinical images, (3) false positive (FP) filtering, and (4) correspondence with physician-rated disease likelihood.Main results.(1) Response to artificially degraded image quality was enhanced by the Local Gradients UQ method, where the median percent difference between matching lesions in non-degraded and most degraded images was consistently higher for the Local Gradients uncertainty measure than the non-gradient uncertainty measures (e.g. 62.35% vs. 2.16% for additive Gaussian noise). (2) The Local Gradients UQ measure responded better to high- and low-quality clinical images (p< 0.05 vsp> 0.1 for both non-gradient uncertainty measures). (3) FP filtering performance was enhanced by the Local Gradients UQ method when compared to the non-gradient methods, increasing the area under the receiver operating characteristic curve (ROC AUC) by 20.1% and decreasing the false positive rate by 26%. (4) The Local Gradients UQ method also showed more favorable correspondence with physician-rated likelihood for malignant lesions by increasing ROC AUC for correspondence with physician-rated disease likelihood by 16.2%.Significance. In summary, this work introduces and validates a novel gradient-based UQ method for deep learning-based medical image assessments to enhance user trust when using deployed clinical models.

Teachers' Punishment Intensity and Student Observer Trust: A Moderated Mediation Model.

During social interactions, people decide whether to trust an actor based on their punitive behaviour. Several empirical studies have indicated that punishment intensity impacts observer trust, yet the underlying mechanism remains to be elucidated. This study included 242 junior high school students and was conducted to investigate the relationship between teachers' punishment intensity and levels of student bystander trust. Additionally, the mediating role of trustworthiness and the moderating role of group relationships were explored. The results showed that the relationship between punishment intensity and observer trust follows an inverted U-shaped pattern. In addition, mild punishment boosts observer trust by improving perceived trustworthiness (ability and integrity) compared to no punishment, while harsh punishment reduces observer trust more than mild punishment by diminishing perceived trustworthiness (ability, benevolence, and integrity). More importantly, group relationships positively moderate the relationship between punishment intensity and observer trust. Specifically, compared to mild or no punishment, harsh punishment decreases trustworthiness (ability, benevolence, and integrity) in close teacher-student relationships but has less impact on neutral relationships. The above findings demonstrate that guiding educators in developing appropriate disciplinary concepts contributes to enhancing student observer trust.

Factors affecting consumers' repurchase intention toward skin care cosmetics: A cross - sectional study in Vietnam.

The purpose of this study is to investigate the factors affecting consumers' repurchase intentions of skincare cosmetics in Can Tho city. By examining consumers' knowledge, behaviors, and factors impacting their intention of repeatedly buying skincare products, manufacturers and sales businesses are able to more effectively comprehend their clients' demands. Data were collected from 531 consumers residing in the area of Can Tho city, Vietnam, who had bought skincare products for the second time or more. The survey using the cross-sectional method and a set of questionnaires was based on the theory of planned behavior (TPB) and developed on the 5-point Likert scale. The factors impacting repurchase intentions for skincare merchandise involve attitude, reliability of signal quality, and retailer credibility, with attitude being the most powerful influence, followed by retailer credibility and reliability of signal quality respectively. These factors accounted for 27.5 % of behavior variance. The clarity element of signal quality reliability does not affect behavioral intention. Cosmetic companies can develop policies, improve the quality of their products in order to suit the customer needs, retain potential clients, and enhance their competitive capability.

Visual analysis of trustworthiness studies: based on the Web of Science database.

Trustworthiness is the most significant predictor of trust and has a significant impact on people's levels of trust. Most trustworthiness-related research is empirical, and while it has a long history, it is challenging for academics to get insights that are applicable to their fields of study and to successfully transfer fragmented results into practice. In order to grasp their dynamic development processes through the mapping of network knowledge graphs, this paper is based on the Web of Science database and uses CiteSpace (6.2.R4) software to compile and visualize the 1,463 publications on trustworthy studies over the past 10 years. This paper aims to provide valuable references to theoretical research and the practice of Trustworthiness. The findings demonstrate that: over the past 10 years, trustworthiness-related research has generally increased in volume; trustworthiness research is concentrated in industrialized Europe and America, with American research findings having a bigger global impact; The University of California System, Harvard University, and Yale University are among the high-production institutions; the leading figures are represented by Alexander Todorov, Marco Brambilla, Bastian Jaeger, and others; the core authors are distinguished university scholars; however, the level of cooperation of the core author needs to be improved. The primary journal for publishing research on trustworthiness is the Journal of Personality and Social Psychology and Biology Letters. In addition, the study focuses on three distinct domains, involving social perception, facial clues, and artificial intelligence.

Influence of protective clothing and masks on facial trustworthiness in an investment game: insights from a Chinese population study.

Facial features are important sources of information about perceived trustworthiness. Masks and protective clothing diminish the visibility of facial cues by either partially concealing the mouth and nose or covering the entire face. During the pandemic, the use of personal protective equipment affected and redefined who trusts whom in society. This study used the classical investment game of interpersonal trust with Chinese participants to explore the impact of occlusion on interpersonal trust. Faces with moderate initial trustworthiness were occluded by a mask or protective clothing in Experiment 1 and were digitally occluded by a square in Experiment 2, and faces with three levels of initial trustworthiness were occluded by a mask in Experiment 3. Results showed that both undergraduates (Experiment 1a) and non-student adults (Experiment 1b) perceived the faces with protective clothing as more trustworthy than faces wearing standard masks and faces not wearing masks. Faces with the top halves showing were perceived as trustworthy as full faces, while faces with the bottom halves showing were perceived as less trustworthy. The effect of masks is weak and complex. Masks reduced participants' trust in faces with high initial trustworthiness, had no effect on faces with low and moderate initial trustworthiness, and only slightly increased the trust of undergraduates in faces with moderate initial trustworthiness. Our findings indicate that the lack of information caused by occlusion and the social significance associated with occlusion collectively affect people's trust behavior in Chinese society. We believe the findings of this study will be useful in elucidating the effects of personal protective equipment usage on perceptions of trustworthiness.

The time course of person perception from voices in the brain.

When listeners hear a voice, they rapidly form a complex first impression of who the person behind that voice might be. We characterize how these multivariate first impressions from voices emerge over time across different levels of abstraction using electroencephalography and representational similarity analysis. We find that for eight perceived physical (gender, age, and health), trait (attractiveness, dominance, and trustworthiness), and social characteristics (educatedness and professionalism), representations emerge early (~80 ms after stimulus onset), with voice acoustics contributing to those representations between ~100 ms and 400 ms. While impressions of person characteristics are highly correlated, we can find evidence for highly abstracted, independent representations of individual person characteristics. These abstracted representationse merge gradually over time. That is, representations of physical characteristics (age, gender) arise early (from ~120 ms), while representations of some trait and social characteristics emerge later (~360 ms onward). The findings align with recent theoretical models and shed light on the computations underpinning person perception from voices.

Misinformation does not reduce trust in accurate search results, but warning banners may backfire.

People rely on search engines for information in critical contexts, such as public health emergencies-but what makes people trust some search results more than others? Can search engines influence people's levels of trust by controlling how information is presented? And, how does the presence of misinformation influence people's trust? Research has identified both rank and the presence of misinformation as factors impacting people's search behavior. Here, we extend these findings by measuring the effects of these factors, as well as misinformation warning banners, on the perceived trustworthiness of individual search results. We conducted three online experiments (N = 3196) using Covid-19-related queries, and found that although higher-ranked results are clicked more often, they are not more trusted. We also showed that misinformation does not damage trust in accurate results displayed below it. In contrast, while a warning about unreliable sources might decrease trust in misinformation, it significantly decreases trust in accurate information. This research alleviates some concerns about how people evaluate the credibility of information they find online, while revealing a potential backfire effect of one misinformation-prevention approach; namely, that banner warnings about source unreliability could lead to unexpected and nonoptimal outcomes in which people trust accurate information less.

Loneliness is associated with more trust but worse trustworthiness expectations.

Subjective feelings of loneliness emerge due to unsatisfactory social relationships, representing a major risk for mental and physical well-being. Despite its social nature, evidence on how loneliness affects social behaviours and expectations is lacking. Using Bayesian analyses and economic games, we show in three different studies that lonelier individuals trusted their partners to a greater extent despite less favourable trustworthiness expectations, showing a greater discrepancy between their trusting behaviours and their expectations of others' trustworthiness. Such discrepancy was reversed in extravert individuals who also reported to be less lonely. These results provide evidence on two opposing effects of loneliness as a motivator for social connections and promoter of social withdrawal, and demonstrate the moderating role of personality traits. This work contributes to a better understanding of how loneliness impacts social behaviour and social expectations, with important downstream clinical implications for varying health conditions associated with heightened feelings of loneliness.

Atypical neural encoding of faces in individuals with autism spectrum disorder.

Individuals with autism spectrum disorder (ASD) experience pervasive difficulties in processing social information from faces. However, the behavioral and neural mechanisms underlying social trait judgments of faces in ASD remain largely unclear. Here, we comprehensively addressed this question by employing functional neuroimaging and parametrically generated faces that vary in facial trustworthiness and dominance. Behaviorally, participants with ASD exhibited reduced specificity but increased inter-rater variability in social trait judgments. Neurally, participants with ASD showed hypo-activation across broad face-processing areas. Multivariate analysis based on trial-by-trial face responses could discriminate participant groups in the majority of the face-processing areas. Encoding social traits in ASD engaged vastly different face-processing areas compared to controls, and encoding different social traits engaged different brain areas. Interestingly, the idiosyncratic brain areas encoding social traits in ASD were still flexible and context-dependent, similar to neurotypicals. Additionally, participants with ASD also showed an altered encoding of facial saliency features in the eyes and mouth. Together, our results provide a comprehensive understanding of the neural mechanisms underlying social trait judgments in ASD.

The influence of face mask color on perceptions of African American and white men.

Despite their widespread use during the COVID-19 pandemic, face masks hinder abilities to interpret facial expressions. Yet, they can also reduce the appearance of characteristics that are used to categorize individuals into racial groups, such as Afrocentric features. The color of a face mask might also promote associations with certain types of behavior and professional occupations (e.g., blue surgical mask connoting physician stereotypes; black masks potentially being associated with criminality). This study assessed whether the presence and color of a face mask impacted perceptions of a target male of varying race. White participants (N = 250) were presented with an African American or White male adult face from the Chicago Face Database (of equal age and attractiveness) wearing a blue or black surgical mask, or no mask (Photoshopped onto the face) and rated the man on emotions (happy, sad, angry) as well as how trustworthy, threatening, and attractive the target appeared. Targets wearing a blue surgical mask were judged as more trustworthy and attractive than those wearing no mask (perhaps due to association with medical professions), but these judgements were not qualified by race, despite the African American target's selection based on Afrocentric features. The color black on a face mask did not exacerbate negative perceptions of targets, perhaps suggesting a decline effect in previously demonstrated associations between this color and criminal actions. Unlike previous research performed at the beginning of the Covid-19 Pandemic on cloth masks shown to potentially exacerbate racial biases, surgical masks (pleated and made of polymeric materials), appear to lessen potential stereotyping of Black relative to White men.

The unmet promise of trustworthy AI in healthcare: why we fail at clinical translation.

Artificial intelligence (AI) has the potential to revolutionize healthcare, for example via decision support systems, computer vision approaches, or AI-based prevention tools. Initial results from AI applications in healthcare show promise but are rarely translated into clinical practice successfully and ethically. This occurs despite an abundance of "Trustworthy AI" guidelines. How can we explain the translational gaps of AI in healthcare? This paper offers a fresh perspective on this problem, showing that failing translation of healthcare AI markedly arises from a lack of an operational definition of "trust" and "trustworthiness". This leads to (a) unintentional misuse concerning what trust (worthiness) is and (b) the risk of intentional abuse by industry stakeholders engaging in ethics washing. By pointing out these issues, we aim to highlight the obstacles that hinder translation of Trustworthy medical AI to practice and prevent it from fulfilling its unmet promises.

Challenges and efforts in managing AI trustworthiness risks: a state of knowledge.

This paper addresses the critical gaps in existing AI risk management frameworks, emphasizing the neglect of human factors and the absence of metrics for socially related or human threats. Drawing from insights provided by NIST AI RFM and ENISA, the research underscores the need for understanding the limitations of human-AI interaction and the development of ethical and social measurements. The paper explores various dimensions of trustworthiness, covering legislation, AI cyber threat intelligence, and characteristics of AI adversaries. It delves into technical threats and vulnerabilities, including data access, poisoning, and backdoors, highlighting the importance of collaboration between cybersecurity engineers, AI experts, and social-psychology-behavior-ethics professionals. Furthermore, the socio-psychological threats associated with AI integration into society are examined, addressing issues such as bias, misinformation, and privacy erosion. The manuscript proposes a comprehensive approach to AI trustworthiness, combining technical and social mitigation measures, standards, and ongoing research initiatives. Additionally, it introduces innovative defense strategies, such as cyber-social exercises, digital clones, and conversational agents, to enhance understanding of adversary profiles and fortify AI security. The paper concludes with a call for interdisciplinary collaboration, awareness campaigns, and continuous research efforts to create a robust and resilient AI ecosystem aligned with ethical standards and societal expectations.

IMPA-Net: Interpretable Multi-Part Attention Network for Trustworthy Brain Tumor Classification from MRI.

Deep learning (DL) networks have shown attractive performance in medical image processing tasks such as brain tumor classification. However, they are often criticized as mysterious "black boxes". The opaqueness of the model and the reasoning process make it difficult for health workers to decide whether to trust the prediction outcomes. In this study, we develop an interpretable multi-part attention network (IMPA-Net) for brain tumor classification to enhance the interpretability and trustworthiness of classification outcomes. The proposed model not only predicts the tumor grade but also provides a global explanation for the model interpretability and a local explanation as justification for the proffered prediction. Global explanation is represented as a group of feature patterns that the model learns to distinguish high-grade glioma (HGG) and low-grade glioma (LGG) classes. Local explanation interprets the reasoning process of an individual prediction by calculating the similarity between the prototypical parts of the image and a group of pre-learned task-related features. Experiments conducted on the BraTS2017 dataset demonstrate that IMPA-Net is a verifiable model for the classification task. A percentage of 86% of feature patterns were assessed by two radiologists to be valid for representing task-relevant medical features. The model shows a classification accuracy of 92.12%, of which 81.17% were evaluated as trustworthy based on local explanations. Our interpretable model is a trustworthy model that can be used for decision aids for glioma classification. Compared with black-box CNNs, it allows health workers and patients to understand the reasoning process and trust the prediction outcomes.

Learned trustworthiness does not have the same influence on implicit responses measured via fast periodic visual stimulation as face trustworthiness.

Past research has demonstrated that it is possible to detect implicit responses to face trustworthiness using fast periodic visual stimulation (FPVS). Because people readily retrieve affective associations with faces, the current study investigated whether learned trustworthiness would yield similar responses to face trustworthiness as measured via FPVS. After learning to associate faces with untrustworthy or trustworthy behaviors, participants completed three separate tasks while electroencephalography (EEG) was recorded. In each of these tasks, participants viewed oddball sequences of faces where a single base face was presented repeatedly at a rate of 6 Hz and oddball faces with different identities were presented every fifth face (6 Hz/5 = 1.2 Hz). Providing evidence of learning, the oddball response at 1.2 Hz and its harmonics was stronger for the learned faces compared to novel faces over bilateral occipitotemporal cortex and beyond. In addition, reproducing previous findings with face trustworthiness, we observed a stronger response at 1.2 Hz and its harmonics for sequences with less trustworthy-looking versus trustworthy-looking oddball faces over bilateral occipitotemporal cortex and other sites. However, contrary to our predictions, we did not observe a significant influence of learned trustworthiness on the oddball response. These data indicate that impressions based on learning are treated differently than impressions based on appearance, and they raise questions about the types of design and stimuli that yield responses that are measurable via FPVS.