Search

Recent advances in artificial intelligence (AI) and machine learning (ML) applications have elevated accomplishments in various scientific fields, primarily those that benefit the economy and society. Contemporary threats, such as armed conflicts, natural and man-made disasters, and illegal immigration, often require fast and innovative but reliable identification aids, in which forensic anthropology has a significant role. However, forensic anthropology has not yet exploited new scientific advances but instead relies on traditionally used methods. The rare studies that employed AI and ML in developing standards for sex and age estimation did not go beyond the conceptual solutions and were not applied to real cases. In this study, on the example of Croatian populations' cranial dimensions, we demonstrated the methodology of developing sex classification models using ML in conjunction with field knowledge, resulting in sex estimation accuracy of more than 95%. To illustrate the necessity of applying scientific results, we developed a web app, CroCrania (https://crocrania.onrender.com), that can be used for sex estimation and method validation. [ABSTRACT FROM AUTHOR]

This study investigated the stroke and slope characteristics in left‐handed and right‐handed handwriting. Stroke (letters t, f, đ, and H) and slope (letters t, f, l, d, and g) directions were analyzed on in‐house samples (n = 64), revealing statistically significant differences (p ≤ 0.05) between the groups. Right‐handers predominantly exhibited left‐to‐right strokes (98%–100%), while left‐handers showed greater variability. Although statistically significant for most letters analyzed, slope direction did not demonstrate consistent patterns. A logistic regression model was developed and validated on the same sample to classify handedness based on the averaged strokes of the letter "t." The model was further tested on samples (n = 252) from a publicly available handwriting database. If the model classified the sample as produced by left hand, it was correct in 100% of cases. In contrast, when the model classified writing as right‐handed, it was correct in 73%–97% of cases, depending on the validation sample. The model classified writing as of left‐handed origin if more than 36% of the letters "t" had a stroke from right to left, while otherwise, writing was classified as of right‐handed origin. The developed method showed great potential for classifying the handedness of the author of disputed handwriting, thus eliminating individuals as text authors or narrowing down the pool of potential authors. [ABSTRACT FROM AUTHOR]

Reconstructing the face from the skull is important not only for forensic identification but also as a tool that can provide insight into the appearance of individuals from past populations. It requires a multidisciplinary approach that combines anthropological knowledge, advanced imaging methods, and artistic skills. In the present study, we demonstrate this process on the skull of an early medieval warrior from Croatia. The skeletal remains were prepared and scanned using multi-slice computed tomography (MSCT) and examined using standard anthropological and radiological methods. The analysis revealed that the remains belonged to a 35–45-year-old male individual who had suffered severe cranial trauma, probably causing his death. From MSCT images, we reconstructed a three-dimensional (3D) model of the skull, on which we digitally positioned cylinders demarking the soft tissue thickness and created the face with a realistic texture. A 3D model of the face was then optimized, printed, and used to produce a clay model. Sculpturing techniques added skin textures and facial features with scars and trauma manifestations. Finally, after constructing a plaster model, the model was painted and refined by adding fine details like eyes and hair, and it was prepared for presentation in the form of a sculpture. [ABSTRACT FROM AUTHOR]

ChatGPT-3.5, an AI language model capable of text generation, translation, summarization, and question-answering, has recently been released for public use. Studies have shown it can generate abstracts, research papers, and dissertations, and create quality essays on different topics. This led to ethical issues in using ChatGPT in academic writing, AI authorship, and evaluating students' essays. However, it is still unknown how ChatGPT performs in students' environments as a writing assistant tool and if it enhances students' essay-writing performance. In the present study, we examined students' essay-writing performances with or without ChatGPT as an essay-writing assistance tool. The average essay grade was C for both control (traditional essay-writing, n = 9) and experimental (ChatGPT-assisted essay-writing, n = 9) groups. None of the predictors affected essay scores: group, writing duration, study module, and GPA. The text unauthenticity was slightly higher in the experimental group, but the similarity among essays was generally low in the overall sample. In the experimental group, the AI classifier recognized more potential AI-generated texts. Our results demonstrate that the ChatGPT group did not perform better in either of the indicators; the students did not deliver higher quality content, did not write faster, nor had a higher degree of authentic text. We anticipate that these results can relieve some concerns about this tool's usage in academic writing. ChatGPT-assisted writing could depend on the previous knowledge and skills of the user, which might, in certain instances, lead to confusion in inexperienced users and result in poorer essay writing performance. [ABSTRACT FROM AUTHOR]

The aim of this study was to test the sexual dimorphism of orbital measurements in the Croatian population using multi-slice computed tomography (MSCT) images. We have retrospectively taken 414 head CT scans of adults from Croatian clinical hospitals in Split and Zagreb (214 males and 200 females) with slice thickness < 1 mm and no pathological or traumatic changes that could affect the measurements. DICOM files were imported into Stratovan Checkpoint Software and viewed in 2D and 3D using semi-transparent 3D volume rendering. Eight standard measurements were calculated based on twelve orbital landmarks (six paired). Principal component analysis (PCA) was used to explore sexual and regional differences, and linear discriminant analysis was used to develop sex classification models. The PCA showed separation based on sex and region, and additional analysis demonstrated that females and males in Split and Zagreb differed in four orbital measurements (P ≤ 0.001). Only those measurements that did not show regional differences were further analyzed, and all showed statistically significant sexual dimorphism. The accuracy of univariate functions for sex estimation ranged from 53.43 to 71.88%, and for multivariate function, the accuracy was 73.45%. The orbital measurements of the Croatian population showed restricted forensic significance for sex classification. On the other hand, we have shown that they can have a potential for exploring the inter- and intra-population differences. [ABSTRACT FROM AUTHOR]

This study examined if the cranial measurements from Data Collection Procedures for Forensic Skeletal Material 2.0 are repeatable when measured in dry bones and MSCT images and if the virtual measurements correspond to the physical ones. The sample included 33 dry crania imaged by MSCT. Two observers measured dry bones, two placed landmarks on 2D and 3D MSCT reconstructions, and one conducted measurements/landmarking on both media. One of the observers for each media repeated the measurements. Technical and relative technical error of measurement (TEM and rTEM) and percentage differences were calculated to examine the repeatability of measurements and compare measuring modalities. Intraobserver rTEM was above 1.5% for six bone measurements: FOB, ZOB, OBB, NLH, DKB, MDH (1.51%–4.87%) and for seven MSCT measurements: OBH, FOB, OBB, MDH, NLB, ZOB, DKB (1.57%–5.55%). The interobserver rTEM was above the acceptable level (>2%) for 11 measurements: PAC, NLH, OBB, EKB, MAL, FOB, NLB, OBH, ZOB, DKB, and MDH (2.01%–9.34%). The percentage differences were not systematically larger for measurements taken by the same user on both modalities than those obtained by different users on the same modality. When physical and MSCT measurements were tested on sex classification standards, the proportion of crania classified as male or female did not significantly differ (p > 0.05). The study showed that physical and virtual cranial measurements could be interchangeable for developing or applying sex estimation standards. However, clarifications and adaptations are necessary for measurements of mastoid, nasal, and orbital regions that did not meet the standard criteria. [ABSTRACT FROM AUTHOR]

The introduction of interdisciplinarity and industry-academia collaborations (IAC) into higher education institutions (HEIs) and curricula as tools for promoting sustainable development has been debated both in academic and non-academic contexts. While overall rising trends in the acceptance of interdisciplinarity and IAC exist, research has stressed difficulty in implementation and practices. We conducted eight focus groups at six European Universities (members of the SEA-EU alliance) and analysed the transcripts using Braun and Clarke's reflexive thematic approach to qualitative analysis in order to develop themes on barriers and facilitators to both conducting interdisciplinarity and IAC, as well as the inclusion of university students in interdisciplinary research. We observed that the main barriers to IR and IAC and the inclusion of students in such activities include traditional HEI structures focused on single-discipline approaches, a lack of joint platforms for IR and IAC, and academic differences (publication outcome differences, academic background). Likewise, a lack of funding (especially for early career researchers), employability (for students willing to do a research career), and a lack of validation by HEIs for researchers conducting IR and IAC are major barriers. To IDR- and IAC-related activities, a top-down approach is needed to restructure HEIs and make them more accommodating to both students and staff willing to conduct IR and IAC activities, thus refocusing them towards sustainability. [ABSTRACT FROM AUTHOR]

Simple Summary: This study adopts a dynamic methodology to explore challenges to the practical application of the adjusted binary classification (ABC) approach, which are related to the unmodifiable characteristics of data used in its development, such as intrasexual variation (sexual dimorphism) of variables and methodological factors such as the selected classification algorithm and sample size. The adequacy of a training dataset's size was judged relative to the classification performance in an independent test set. Finding an optimal classifier was also addressed in this study, wherein the results demonstrate that both statistical modeling and machine learning techniques perform almost equally in the univariate models; however, differences are evident in the multivariate model due to the different number of variables included via the feature selection process, as well as the effect of inadequate training sample size relative to the test set. This approach is particularly useful when quick classification/prediction is required for making real-time forensic decisions. The adjusted binary classification (ABC) approach was proposed to assure that the binary classification model reaches a particular accuracy level. The present study evaluated the ABC for osteometric sex classification using multiple machine learning (ML) techniques: linear discriminant analysis (LDA), boosted generalized linear model (GLMB), support vector machine (SVM), and logistic regression (LR). We used 13 femoral measurements of 300 individuals from a modern Turkish population sample and split data into two sets: training (n = 240) and testing (n = 60). Then, the five best-performing measurements were selected for training univariate models, while pools of these variables were used for the multivariable models. ML classifier type did not affect the performance of unadjusted models. The accuracy of univariate models was 82–87%, while that of multivariate models was 89–90%. After applying ABC to the crossvalidation set, the accuracy and the positive and negative predictive values for uni- and multivariate models were ≥95%. Sex could be estimated for 28–75% of individuals using univariate models but with an obvious sexing bias, likely caused by different degrees of sexual dimorphism and between-group overlap. However, using multivariate models, we minimized the bias and properly classified 81–87% of individuals. A similar performance was also noted in the testing sample (except for FEB), with accuracies of 96–100%, and a proportion of classified individuals between 30% and 82% in univariate models, and between 90% and 91% in multivariate models. When considering different training sample sizes, we demonstrated that LR was the most sensitive with limited sample sizes (n < 150), while GLMB was the most stable classifier. [ABSTRACT FROM AUTHOR]

Plagiarism is one of the most severe academic integrity issues. This study examined students' knowledge of and attitudes towards plagiarism, tested their ability to recognize plagiarism, and explored the association of study levels and attendance in courses dealing with referencing rules and plagiarism with students' attitudes and knowledge. A cross-sectional online survey was conducted at the University of Split, comprising the students of all schools and study levels (n = 388). Overall, results indicate the students were not very familiar with referencing rules and did not perform well on either theoretical questions or practical examples. However, they demonstrated positive attitudes towards plagiarism avoidance as well as towards compliance with academic integrity with respect to the correct use of research publications. Students' self-reported attendance in courses dealing with referencing rules and plagiarism avoidance was not associated with their knowledge of and attitudes toward plagiarism. These findings are important for a general understanding of students' attitudes, and the relation of practical and theoretical knowledge of plagiarism. Furthermore, the academic community addresses plagiarism not only as an ethical and regulatory violation but also as a direct consequence of a lack of knowledge, and of academic illiteracy. Study programs should be adjusted and long-term policies developed at all academic levels to promote a positive climate among students towards responsible academic writing. [ABSTRACT FROM AUTHOR]

In a Sidonian sarcophagus, from the Late Antique/early Christian period, skeletal remains of two persons were found. One of them, male, 30-50 years old, was found almost completely ankylosed, with highly osteoporotic bones and prominent erosion of joint surfaces. We diagnosed rheumatoid arthritis based on the eroded odontoid process, mandibular condyles, distal humerus, proximal and distal ulna, as well ankylosed hand and foot bones. Despite the fact that ankyloses of vertebrae and sacroiliac joint could point towards ankylosing spondylitis, the lack of typical vertebral ankyloses and new bone formation led to exclusion. In a practical sense, due to the advanced stage of the disease, the man was fixed in the supine position, on the left, with his head turned to the right. Apparently, he could not move and had problems with chewing and breathing. But, the high standard of provided healthcare probably enabled him to survive in advanced stages of the disease. This case shed light on the antiquity of the disease, its medical, and social context and provided the example of most extreme osteological changes reported in the paleopathological and medical literature. [ABSTRACT FROM AUTHOR]

The determination of sex is one of the first steps in anthropological analysis. When skeletal remains are fragmented, the most useful approach is application of osteometric methods. The methods are population specific, and therefore require development of discriminant functions for each population group. The aim of this study was to test sexual dimorphism of femoral measurements and to calculate discriminant functions applicable for sex determination on fragmented skeletal remains on the late antique sample from the Eastern Adriatic coast (2nd-6th century AD). 214 randomly chosen skeletons from the excavation site Solin-Smiljanovac were analyzed. Sex and age were assessed using standard anthropological methods, and skeletons were examined for pathologic and traumatic changes. In the next step, we selected 27 female and 48 male skeletons free of peri- or post-mortem changes that could affect measurements Eight standard femoral measurements were taken. Sexual dimorphism was initially compared using independent sample t test, after which discriminant functions were computed. All femoral measurements showed statistically significant sexual dimorphism (p<0.001). Ten discriminant functions for every part of femur were calculated and obtained classification accuracy of 73.1-91.8%. This study reached relatively high classification, which will improve further analysis of the skeletal remains from the Salona necropolis. Due to similar population structure in the Roman period across the Adriatic coast, the discriminant functions could be applicable for all populations from the same period and area. This study also raised a few methodological questions showing that when creating discriminant functions we should consider not only the accuracy, but also the applicability based on the experience from the anthropology laboratory that considers the state of preservation and frequent pathology. [ABSTRACT FROM AUTHOR]

Sister Marija Krucifiksa Kozulić (1852–1922) was a Croatian nun who is in consideration for beatification by the Vatican, which is facilitated by the identification of her 20th-century remains. Sister Marija was buried in a tomb in Rijeka, Croatia, along with other nuns including her biological sister, Tereza Kozulić (1861–1933). When the remains were exhumed in 2011, they were found in a deteriorated state and commingled with several other sets of remains. Thus, mitochondrial genome sequencing of the long bones was performed to sort the remains by mitochondrial haplotype. Two similar but unique haplotypes belonging to haplogroup H1bu were identified, and samples from these bones were subjected to autosomal short tandem repeat (STR) and single nucleotide polymorphism (SNP) sequencing. Although only partial profiles were obtained, the data were sufficient for kinship analysis with the profile of a paternal niece of Sister Marija (Fides Kozulić). The data indicate that it is 574,195-fold more likely that the two sets of skeletal remains represent 2nd-degree relatives of Fides than sisters who are unrelated to Fides. Although it is impossible to discern which set of remains belongs to Marija and which belongs to Tereza, forensic genomics methods have enabled identification of the sisters. [ABSTRACT FROM AUTHOR]