We present the development and validation of a self-report instrument on Cognitions and Emotions about Child Sexual Abuse (CECSA). Three subscales, consisting of 23 items in total, were developed in a sample of 801 humanities students by means of exploratory factor analysis and Ant Colony Optimization, an automated item selection strategy used to simultaneously optimize model fit, reliability, and predictive validity. The “Naïve Confidence” subscale reflects overestimating one's ability to recognize abused children and overestimating the accuracy of children’s abuse reports, the "Emotional Reactivity" subscale measures the intensity of one's emotional reactions towards the topic of child sexual abuse (CSA), and the "Justice System Distrust" subscale covers distrusting the justice system’s ability to prosecute CSA cases. The CECSA showed adequate model fit and good internal consistencies. Bivariate correlations with other self-report measures demonstrated convergent validity. Importantly, all three subscales predicted biased evaluations towards the abuse hypothesis in scenarios of children displaying unspecific behavioral problems. This indicates predictive validity of the CECSA as an instrument measuring vulnerability for interviewer bias. The CECSA can be used to assess individual training needs of professionals who conduct interviews or conversations with children about abuse suspicions and may help to develop and evaluate interviewer trainings.