Make Trust Decisions Research Articles

To Authenticity, and Beyond! Building Safe and Fair Generative AI Upon the Three Pillars of Provenance.

Provenance facts, such as who made an image and how, can provide valuable context for users to make trust decisions about visual content. Against a backdrop of inexorable progress in generative AI for computer graphics, over two billion people will vote in public elections this year. Emerging standards and provenance enhancing tools promise to play an important role in fighting fake news and the spread of misinformation. In this article, we contrast three provenance enhancing technologies-metadata, fingerprinting, and watermarking-and discuss how we can build upon the complementary strengths of these three pillars to provide robust trust signals to support stories told by real and generative images. Beyond authenticity, we describe how provenance can also underpin new models for value creation in the age of generative AI. In doing so, we address other risks arising with generative AI such as ensuring training consent, and the proper attribution of credit to creatives who contribute their work to train generative models. We show that provenance may be combined with distributed ledger technology to develop novel solutions for recognizing and rewarding creative endeavor in the age of generative AI.

Trust, lies, and inequality

AbstractThe effect of endowment equality on trust may stem from outcome inequality aversion or changes in expected trustworthiness. Here, we measure trust as the expectation of honesty in a sender‐receiver game, where participants must make trust decisions without knowing the outcome. Our design enables us to isolate the effect of initial endowment inequality on trust. Our results show that endowment inequality reduces trust regardless of whether it favors the sender or the receiver. We further find that the frequency of lies is insensitive to endowment inequality. Our results amplify the importance of equal starting positions in promoting trust.

Social network position, trust behavior, and neural activity in young adolescents

Our social interactions take place within numerous social networks, in which our relationships with others define our position within these networks. In this study, we examined how the centrality of positions within social networks was associated with trust behavior and neural activity in 49 adolescents (Mage = 12.8 years, SDage = 0.4 years). The participants played a trust game with a cartoon animation as a partner, which showed adaptive behavior in response to the participant and was generally untrustworthy. Social network positions were obtained in secondary school classrooms where the participants and their classmates reported on who their friends were. Using social network analysis, a score was calculated that indicated the centrality of everyone's position within the friendship network. The results showed that more central social network positions were associated with higher levels of initial trust behavior, although no evidence was found for a relationship between network position and the adaptation of trust behavior. The results of the functional MRI analyses showed that the centrality of the network positions was positively associated with caudate activity when making trust decisions. Furthermore, the adolescents with more central network positions also showed stronger increases of caudate activity when the partner's return was processed compared to the adolescents with less central network positions. The current study provides initial evidence that social network positions in friendship networks relate to socio-cognitive behavior and neural activity in adolescents.

The effect of incorrect prior information on trust behavior in adolescents

During adolescence, social cognition and the brain undergo major developments. Social interactions become more important, and adolescents must learn that not everyone can be trusted equally. Prior knowledge about the trustworthiness of an interaction partner may affect adolescents' expectations about the partner. However, the expectations based on prior knowledge can turn out to be incorrect, causing the need to respond adaptively during the interaction. In the current fMRI study, we investigated the effect of incorrect prior knowledge on adolescent trust behavior and on the neural processes of trust. Thirty-three adolescents (Mage = 17.2 years, SDage = 0.5 years) played two trust games with partners whose behavior was preprogrammed using an algorithm that modeled trustworthy behavior. Prior to the start of both games, participants received information suggesting that the partner in one game was untrustworthy (raising incorrect expectations) and the partner in the other game trustworthy (raising correct expectations). Results indicated that participants adapted their trust behavior following incorrect prior expectations. No evidence for a change in trust behavior was shown when prior expectations were correct. fMRI analyses revealed that when receiving the partner's response, activity in the dorsolateral prefrontal cortex and in the superior parietal gyrus were increased when participants had incorrect expectations about the partner compared to when participants had correct expectations. When making trust decisions, no significant differences in neural activity were found when comparing the two games. This study provides insight into how adolescent trust behavior and neural mechanisms are affected by expectations and provides an increased understanding of the factors that influence adolescent social interactions.

Can people detect the trustworthiness of strangers based on their facial appearance?

Although cooperation can lead to mutually beneficial outcomes, cooperative actions only pay off for the individual if others can be trusted to cooperate as well. Identifying trustworthy interaction partners is therefore a central challenge in human social life. How do people navigate this challenge? Prior work suggests that people rely on facial appearance to judge the trustworthiness of strangers. However, the question of whether these judgments are actually accurate remains debated. The present research examines accuracy in trustworthiness detection from faces and three moderators proposed by previous research. We investigate whether people show above-chance accuracy (a) when they make trust decisions and when they provide explicit trustworthiness ratings, (b) when judging male and female counterparts, and (c) when rating cropped images (with non-facial features removed) and uncropped images. Two studies showed that incentivized trust decisions (Study 1, n = 131 university students) and incentivized trustworthiness predictions (Study 2, n = 266 university students) were unrelated to the actual trustworthiness of counterparts. Accuracy was not moderated by stimulus type (cropped vs. uncropped faces) or counterparts' gender. Overall, these findings suggest that people are unable to detect the trustworthiness of strangers based on their facial appearance, when this is the only information available to them.

What is it like to trust a rock? A functionalist perspective on trust and trustworthiness in artificial intelligence

The trustworthiness (or otherwise) of AI has been much in discussion of late, not least because of the recent publication of the EU Guidelines for Trustworthy AI. Discussions range from how we might make people trust AI to AI being not possible to trust, with many points inbetween. In this article, we question whether or not these discussions somewhat miss the point, which is that people are going ahead and basically doing their own thing anyway, and that we should probably help them. Acknowledging that trust is a heuristic that is widely used by humans in a range of situations, we lean on the literature concerning how humans make trust decisions, to arrive at a general model of how people might consider trust in AI (and other artefacts) for specific purposes in a human world. We then use a series of thought experiments and observations of trust and trustworthiness, to illustrate the use of the model in taking a functionalist perspective on trust decisions, including with machines. Our hope is that this forms a useful basis upon which to develop intelligent systems in a way that considers how and when people may trust them, and in doing so empowers people to make better trust decisions about AI.

Calculating Trust Using Multiple Heterogeneous Social Networks

In today’s Internet, a web user becomes members of multiple social networks due to different types of services provided by each of these networks. This creates an opportunity to make trust decisions that go beyond individual social networks, since these networks provide single perspective of trust. To make trust inference over multiple social networks, these networks need to be consolidated. It is nontrivial as these networks are of heterogeneous nature due to different naming conventions used in these networks. Furthermore, trust metrics extracted from these networks are also varied in nature due to different trust evaluation algorithms used in each of these networks. Heterogeneity of these social networks can be overcome by using semantic technologies as it allows us to represent knowledge using ontologies. Trust data can be consolidated by using such data fusion techniques which not only provide but also preserve trust data integrity from each of the individual social network profiles. The proposed semantic framework is evaluated using two sets of experiments. Through simulations in this work, we analysed various techniques for data fusion. For identifying suitable technique that preserves the integrity of trust consolidated from each of the individual networks, analysis revealed that Weighted Ordered Weighted Averaging parameter best aggregated trust data, and, unlike other techniques, it preserved the integrity of trust from each individual network for varying participant overlap and tie overlap (p≤0.05). Similarly, for experimental analysis, we used findings of the simulation study about the best trust aggregation technique and applied the proposed framework on real-life trust data between participants, which we extracted from pairs of professional social networks. Analysis partially proved our hypothesis about generating better trust values from consolidated multiple heterogeneous networks. We witnessed an improvement in overall results for all the participants who were part of multiple social networks (p≤0.05), while disproving the claim for those existing in nonoverlapping regions of the social networks.

Citation behavior of Pondicherry University faculty in digital environment: a survey

PurposeThe purpose of this paper is to understand the citation behavior of Pondicherry University faculty with a focus on their recent publication experience, indexed in Scopus. The paper endeavors to gain an insight into their pattern of citation of scientific papers and attempts to understand the underlying motive in doing so especially in an era where information resources have transitioned from print to digital.Design/methodology/approachThe study adopted a judgmental sample consisting of 100 selected faculty members whose recent papers were indexed by Scopus database. A pre-defined questionnaire consisting of demographic profile and 23 items on citation trust and authority statements of citing sources was self-administered to them.FindingsThe findings of the study indicate that the citation behavior of faculty in Pondicherry University is complex and multi-faceted. Although majority of the faculty depicted an inclination toward normative citation behavior (concerned with the intellectual content of the work), there are others who displayed social constructivist (concerned with the social aspects of the work) citation behavior as well. However, in spite of the differences observed in the citation behavior of the faculty, it was observed that they remained traditional while making trust decisions even in the digital era. Finally, findings suggest no statistical significant difference when it comes to variables, such as gender and discipline in the citation behavior of the faculty.Originality/valueGiven the ever-increasing importance of citation in judging the quality of research journals, in ranking institutions and in determining the efficacy of faculty in India, author-based approach of understanding citer motivation definitely carry research value.

Trust evaluation based on evidence theory in online social networks

Trust is an important criterion for access control in the field of online social networks privacy preservation. In the present methods, the subjectivity and individualization of the trust is ignored and a fixed model is built for all the users. In fact, different users probably take different trust features into their considerations when making trust decisions. Besides, in the present schemes, only users’ static features are mapped into trust values, without the risk of privacy leakage. In this article, the features that each user cares about when making trust decisions are mined by machine learning to be User-Will. The privacy leakage risk of the evaluated user is estimated through information flow predicting. Then the User-Will and the privacy leakage risk are all mapped into trust evidence to be combined by an improved evidence combination rule of the evidence theory. In the end, several typical methods and the proposed scheme are implemented to compare the performance on dataset Epinions. Our scheme is verified to be more advanced than the others by comparing the F-Score and the Mean Error of the trust evaluation results.

Split-Second Trustworthiness Detection From Faces in an Economic Game.

Economic interactions often imply to gauge the trustworthiness of others. Recent studies showed that when making trust decisions in economic games, people have some accuracy in detecting trustworthiness from the facial features of unknown partners. Here we provide evidence that this face-based trustworthiness detection is a fast and intuitive process by testing its performance at split-second levels of exposure. Participants played a Trust game, in which they made decisions whether to trust another player based on their picture. In two studies, we manipulated the exposure time of the picture. We observed that trustworthiness detection remained better than chance for exposure times as short as 100ms, although it disappeared with an exposure time of 33ms. We discuss implications for ongoing debates on the use of facial inferences for social and economic decisions.

A multi-criteria decision-making formulation of trust using fuzzy analytic hierarchy process

Trust models have widespread applications in different areas. In this paper, we formulate trust as a multi-criteria decision-making (MCDM) problem. By applying this new formulation, one can treat the process of trust establishment as a decision-making process based on some set of trustworthiness criteria. By looking at trust as a whole, we may miss some important factors. Only by dividing this concept into its building blocks and weighting those building blocks based on the proposed perception of importance in a context and toward a specific goal, we can achieve a better decision. As a structural way for making trust decisions, the proposed method uses fuzzy analytic hierarchy process (AHP). Fuzzy AHP provides both the advantages of fuzziness, which is compatible with the uncertain, vague, and subjective nature of trust and AHP as a systematic approach to make decisions in complex MCDM problems. In order to aggregate the evaluations of each recommender about every trustee candidate, the Sugeno fuzzy integral is used to build the final trust score. At last, the method calculates the final recommendation score based on the trustor's subjective belief. We present an illustrative example plus an empirical experiment in the context of Web service selection to demonstrate the application of the proposed formulation of trust and the evaluation of trust scores.

Perceived trustworthiness of faces drives trust behaviour in children.

Facial appearances can powerfully influence adults' trust behaviour, despite limited evidence that these cues constitute honest signals of trustworthiness. It is not clear, however, whether the same is also true for children. The current study investigated whether, like adults, 5-year-olds and 10-year-olds are more likely to place their trust in partners that look trustworthy than those that look untrustworthy. A second, closely related question was whether children also explicitly value the information from face cues when making trust decisions. We investigated these questions using Token Quest: an economic trust game that gave participants the opportunity to make investments with a series of partners who might (or might not) repay their trust with large returns. These interactions occurred under different conditions, including one in which participants were shown the face of each partner and another in which they could 'purchase' access to faces with a portion of their investment capital. Results indicated that, like adults, 10-year-old children selectively placed their trust in those partners they perceived as looking trustworthy and many were willing to 'pay' to purchase access to these face cues during the trust game. We observed a similar profile of trust behaviour in 5-year-olds, with no significant group difference in the impact of face cues on behaviour across the three age groups. Together, these findings indicate that the influence of face cues on trust behaviour emerges early, and highlight a capacity for sophisticated social cognition in young children.

Behind HumanBoost: Analysis of Users’ Trust Decision Patterns for Identifying Fraudulent Websites

This paper analyzes users’ trust decision patterns for detecting phishing sites. Our previous work proposed HumanBoost [1] which improves the accuracy of detecting phishing sites by using users’ Past Trust Decisions (PTDs). Web users are generally required to make trust decisions whenever their personal information is requested by a website. Human-Boostassumed that a database of Web user’s PTD would be transformed into a binary vector, representing phishing or not-phishing, and the binary vector can be used for detecting phishing sites, similar to the existing heuristics. Here, this paper explores the types of the users whose PTDs are useful by running a subject experiment, where 309 participants- browsed 40 websites, judged whether the site appeared to be a phishing site, and described the criterion while assessing the credibility of the site. Based on the result of the experiment, this paper classifies the participants into eight groups by clustering approach and evaluates the detection accuracy for each group. It then clarifies the types of the users who can make suitable trust decisions for HumanBoost.

HumanBoost: Utilization of Users’ Past Trust Decision for Identifying Fraudulent Websites

This paper presents HumanBoost, an approach that aims at improving the accuracy of detecting so-called phishing sites by utilizing users’ past trust decisions (PTDs). Web users are generally required to make trust decisions whenever their personal information is requested by a website. We assume that a database of user PTDs would be transformed into a binary vector, representing phishing or not-phishing, and the binary vector can be used for detecting phishing sites, similar to the existing heuristics. For our pilot study, in November 2007, we invited 10 participants and performed a subject experiment. The participants browsed 14 simulated phishing sites and six legitimate sites, and judged whether or not the site appeared to be a phishing site. We utilize participants’ trust decisions as a new heuristic and we let AdaBoost incorporate it into eight existing heuristics. The results show that the average error rate for HumanBoost was 13.4%, whereas for participants it was 19.0% and for AdaBoost 20.0%. We also conducted two follow-up studies in March 2010 and July 2010, observed that the average error rate for HumanBoost was below the others. We therefore conclude that PTDs are available as new heuristics, and HumanBoost has the potential to improve detection accuracy for Web user.

Social Norms and Interpersonal Trust

The idea that people follow trust norms when making trust decisions is developed in an evolutionary model of adaptive play by boundedly rational agents. Because it neither implies nor is it implied by cooperation, trust is not modelled as cooperation in a Prisoners’ Dilemma but as a coordination game with Stag hunt payoffs. The game’s two pure strategy Nash equilibria correspond to high and low trust norms, the former being Pareto optimal and the latter risk-dominant. The mixed strategy Nash equilibrium corresponds to medium trust, but is not a norm in the sense that it is not stable. Multiple equilibria explain the local conformity in trust decisions within countries, and the global diversity across countries. The model suggests that most countries are low trust because the low trust norm is stochastically stable; and that a collapse of trust in medium trust countries is prevented by exogenous forces, probably formal institutions.