Comparison of risk of recidivism among sexual offenders with and without sexual preference disorders using the STATIC-99R instrument.

Purpose: The STATIC-99 instrument is one of the tools used for the assessment of the risk of recidivism, in line with the actuarial approach. One of the risk factors indicated by the scientific literature as having the greatest significance is sexual preference disorder. The aim of the study was to verify whether sexual offenders diagnosed with sexual preference disorders have a higher risk of recidivism. The study also aimed to present, for the first time in Poland, a quantitative scoring of individual risk factors in STATIC-99R and their prevalence, allowing for the verification of the theoretical validity of the STATIC-99R instrument in the analysis of the population of sexual offenders in Poland. Methods: The study material consisted of 100 court and penitentiary files of perpetrators of crimes against sexual freedom from 11 Polish penal institutions and remand centers. We used the STATIC-99R to evaluate each case. Results: The distribution of the individual STATIC-99R risk factors in the population of the Polish sexual offenders is presented. The diagnosis of sexual preference disorders had no influence on the total STATIC-99R score but was associated with its individual factors. Conclusions: It can be concluded that the theoretical validity of the STATIC-99R tool is also relevant to the Polish study population and may be used in clinical practice.

Using expert-reviewed CSAM to train CNNs and its anthropological analysis.

Machine learning methods for the identification of child sexual abuse materials (CSAM) have been previously studied, however, they have serious limitations. Firstly, the training sets used to train the appropriate machine learning algorithms were not previously annotated by a forensic expert in anthropology. Secondly, previously presented solutions have rarely used models trained using real pornographic content involving children. Thirdly, previous studies have not presented a detailed justification for the classification decisions made, which is important due to the recent guidelines of the European Commission (Artificial Intelligence Act). The aim of the study was to train convolution neural networks (CNNs) using expert-labelled CSAM images and thereby identify the elements of the body and/or the environment that are critical for classifications by the neural network. To train and evaluate machine learning models, we used 60,000 images equally divided into four classes (CSAM images, images displaying sexual activity involving adults, images of people without sexual activity, and images not containing people). We used four neural network architectures: MobileNet, ResNet152, xResNet152 and its modification ResNet-s, designed for the purpose of research. The trained models provided high accuracy of classifying CSAM images: xResNet152 (F1 = 0.93, 92,8%), xResNet-s (F1 = 0.93, 93,1%), ResNet152 (F1 = 0.90, 91,39%), MobileNet (F1 ranged from 0.85 to 0.87, accuracy ranged from 86% to 87%). The results of the conducted research suggest that using expert knowledge (in sexology and anthropology) significantly improved the accuracy of the models. In regard to further anthropological analysis, the results indicate that the breasts, face and torso are crucial areas for the classification of pornographic content with children's participation. Results suggests that the ResNet-s neural network may be a reliable tool for clinical work and to support the work of experts witnesses in the field of anthropology. The study design received a positive opinion of the Ethics Committee of the Faculty of Mathematics, Informatics and Mechanics of the University of Warsaw. The clinical material was used for research purposes with the consent of the relevant prosecutor's offices. Authors provided free version of Windows application to classify CSAM for forensic experts, policemen and prosecutors at the OSF repository (DOI: 10.17605/OSF.IO/RU7JX).

Classification of nine types of pornographic materials using the sAI 0.3 model. Continuation of the pilot study. / Klasyfikacja dziewieciu typów materialów pornograficznych za pomoca modelu sAI 0.3. Kontynuacja badan pilotazowych.

OBJECTIVES: Legal pornographic materials are a heterogenous group of audiovisual materials that depict one or more person over the age of eighteen engaging in sexual activities. The aim of this study was to train a model that could classify given types of pornographic materials. METHODS: Materials included in the training set (3,600 materials) and the validation set (900 materials) were manually classified and tagged by psychologists-sexologists. Then, a deep neural network was trained on the dataset. Six models based on different architectures of convolutional neural networks were included in the study (ResNet152, ResNet101, VGG19, VGG16, Squeezenet 1.1, Squeezenet 1.0). Each model was trained on the same group of photographs, and fast.ai library was used for the training process. RESULTS: The final model allows for the classification of more types of pornographic materials with greater efficiency than the pilot model, and thanks to the manual labelling of individual photographs, the limitations of the classification are known. CONCLUSIONS: The possible applications of the model in clinical sexology and psychiatry are discussed. The application of deep neural networks in sexology seems to be particularly promising for at least two reasons. Firstly, a tool for automated detection of pornographic materials involving minors can be developed and used during criminal proceeding. Secondly, after retraining the presented model on photographs of men and women not engaging in sexual activity the model could be used to filter content that is inappropriate for minors.

The use of machine learning to support the therapeutic process - strengths and weaknesses.

Purpose: Artificial neural networks, "artificial intelligence" or machine learning now dominate a number of areas, making many activities automatic and thus affecting the safety and comfort of life. Neural networks might provide intelligent decisions with limited human assistance. Medicine also uses artificial intelligence, also in models designed to support the therapeutic process. The aim of this article is to define the main directions of development of machine learning applications in supporting the therapeutic processes. Views: Currently, the literature distinguishes at least a few applications of new technologies of varying degrees of advancement, with machine learning at the forefront [6]. It seems that the researchers are most interested in personalizing notifications of therapeutic applications, modifying therapeutic programs in a manner adapted to the patient's problems, and conducting "intelligent" conversations with them. Conclusions: There are dangers in using machine learning methods to support the therapeutic process. Particular attention should be paid to ensuring the full privacy of the implemented applications; moreover, selling user data of this type to third parties, such as those that sell certain medications or dietary supplements, would be ethically questionable. There are no legal regulations (or a system of recommendations of relevant scientific societies) that would limit proven applications to support the therapeutic process of a given disorder in the future, and which were created solely for the financial purpose of authors who did not conduct substantive consultations.

Description of the neural network based on AB/DL pictures. Possible implications for forensic sexology.

Purpose: Neural networks might be an appropriate solution for the categorization of child sexual abuse materials (CSAM) in forensic cases. The aim of this study was to present a neural network model that may be able to categorize objects and behaviors, which are visible in CSAM, using pictures visually similar to CSAM (AB/DL), involving persons who have paraphilic preferences for watching adult women or men dressed like children or involved in activities typical for children, such as playing. Methods: The dataset consisted of 2251 photos divided into five classes. 1914 photos were randomly used for the training of the neural network, while 337 photos were used for its later validation. The Fast.ai and PyTorch libraries were used for the training of the neural network using the ResNet152 model. We used five classes, two of which were imported from the sexACT dataset, and three of which that were collected for this study. Results: The model was able to classify selected classes with a relatively high accuracy (95%); on the other hand, further improvement of the network is needed, considering the fact that the final validation loss was moderate (0.17). Conclusions: The model presented might be effective in the classification of several objects and behaviors presented in a number of pornography categories ("paraphilic infantilism", "sexual activity", "nude women", "dressed women", "sexual activity - spanking"). As the results are promising, further research on real CSAM is justified.

Polish adaptation of emotional Stroop test in assessment of pedophilia - a pilot study. / Polska adaptacja zadania Stroopa w warunku emocjonalnym do oceny preferencji pedofilnych ­ badania pilotazowe.

OBJECTIVES: A pilot study was conducted in order to construct a Polish adaptation of emotional Stroop test in assessment of pedophilia. METHODS: The study consisted of two stages. The first stage involved creating test material by ranking words in adequate lists by competent experts. The second stage consisted of empirical verification of the principle of emotional Stroop test in a non-clinical population. RESULTS: Based on the assessment of five competent experts, words were ordered from the most to the least sexually arousing (Kendall's W from 0.368 to 0.693). Six ranked lists were obtained, and the competent experts were subsequently asked to assess whether these lists were suitable for the study (Lawshe's Content Validity Ratio from 0.6 to 1.0). Two categories of words were merged. Five ranked lists were obtained, and the competent experts were subsequently asked again to assess whether these lists were suitable for the study (Lawshe's Content Validity Ratio 1.0). The created lists of words were approved by allcompetent experts. Based on the experimental study conducted on a non-clinical population, it was shown that, in accordance with the principle of the test, the mean response time for sexually related words was longer that for neutral words. The mean response time for children-related words did not differ significantly from response time for neutral words. CONCLUSIONS: Based on the study with competent experts and conducted experiments, an initial Polish adaptation of the emotional Stroop test for diagnosis of pedophilia has been created. Further studies with persons with pedophilia are needed to implement the test in clinical setting.