The Conflicting Incentives Risk Analysis (CIRA) method makes predictions about human decisions to characterize risks within the domain of information security. Since traditional behavior prediction approaches utilizing personal features achieve low prediction accuracies in general, there is a need for improving predictive capabilities. Therefore, the primary objective of this study is to propose and test a psychological approach for behavior prediction, which utilizes features of situations to achieve improved predictive accuracy. An online questionnaire was used for collecting behavioral and trait data to enable a comparison of approaches. Results show that the proposed behavior prediction approach outperforms the traditional approach across a range of decisions. Additionally, interrater reliabilities are analyzed to estimate the extent of objectivity in situation evaluations, providing an indication about the potential performance of the approach when a risk analyst needs to rely on unobtrusive assessment of actiondesirability.