| Nowadays we live in a networked society with cloud computing,online transactions and other interactions made possible by Internet technology.Unfortunately,the growing importance of IT also fosters an ever-growing wave of cybercrime,especially online fraud.Victims of online fraud can suffer significant financial and psychological distress.Despite the direct monetary costs incurred by victimization,there is ample evidence that victimization is also associated with psychological problems,including sleep deprivation,depression and even suicidal ideation.Therefore,internet fraud has become a significant public health problem that requires investigation,surveillance,and intervention.Much of this previous research has focused upon demographic factors associated with internet fraud victimization but there were many inconsistencies in the results.Studies on the psychological characteristics of gullible individuals tend to focus on a single or a few factors,without paying attention to the combined effects of different variables,and the exploration of the mechanisms and interaction of different variables is lacking.In addition,most of the previous studies only stay at the level of questionnaire surveys and theoretical discussions,and few researchers have developed predictive models of individual susceptibility to fraud that can be applied to real-life situations to assist fraud prevention efforts.This study examined the effects and mechanisms of cognitive traits,monetary motivation and personality on individuals’ susceptibility to deception by means of questionnaires and big data text analysis,and used machine learning to construct a predictive model that can be applied to predict individuals’ susceptibility to deception based on Weibo text features.In Study 1,online surveys were conducted to compare the differences between online fraud gullible and non-gullible people in terms of cognitive,personality and motivational psychological traits,examed the predictive effects of the three psychological traits on individuals’ susceptibility to gullibility.In order to explore the mechanisms of such effects,we also examed the mediating effects of cognitive traits between Big 5 personality and gullibility,monetary motivation and gullibility.Study 2 consists of two sub-studies.First,we searched for gullible and non-gullible users on the Weibo platform and analyzed their Weibo posts,calculated word frequency and scores of psychological indicators(e.g.,cognition,personality,and monetary motivation)from the collected data.By examing the differences between gullible and non-gullible users on psychological content indicators,we laid the foundation for constructing a prediction model for the susceptibility to online fraud.Secondly,using the above scores of psychological indicators as model input features,a prediction model of individual’s susceptibility to internet fraud was constructed based on machine learning algorithms such as support vector machine,random forest,and logistic regression.Results demonstrated that:(1)In terms of monetary motivation,perceived risk-reward positively predicted individual’s susceptibility to internet fraud while the effect of materialism was not significant.In terms of Big Five personality,neuroticism also positively predicted susceptibility while extraversion and consciousness negatively predicted individual’s susceptibility to internet fraud.In terms of cognitive characteristics,both cognitive flexibility and critical thinking negatively predicted individual’s susceptibility.(2)Cognitive characteristics played a mediating role in the effects of money motivation and Big Five personality on susceptibility to online fraud.Critical thinking,rather than cognitive flexibility mediated the effect of perceived risk-reward,extraversion and neuroticism on individual susceptibility to online fraud.(3)A prediction model of individual susceptibility to online fraud was constructed based on the personality,monetary motivation,and cognitive-related indicators obtained from the Weibo posts,and the F1 value of the optimal model reached 0.82,in which the most effective feature was the money word.After adding other LIWC text features as model input,the F1 value of the optimal model reached 0.85,providing a new channel for identifying vulnerable people. |
