Large language models (LLMs) increasingly mediate human communication, decision support, content creation, and information retrieval. Despite impressive fluency, these systems frequently produce biased or stereotypical content, especially when prompted with socially sensitive language. A growing body of research has demonstrated that such biases disproportionately affect low-resource languages, where training data is limited and culturally unrepresentative. This paper presents a comprehensive study of inference-time bias mitigation, a strategy that avoids retraining or fine-tuning and instead operates directly on model outputs. Building on preference-ranking models (PRMs), we introduce a unified evaluation framework comparing three methods: (1) baseline single-word generation, (2) PRM-Select best-of-N sampling, and (3) PRM-Sequential refinement guided by PRM critiques. We evaluate these techniques across 200 English prompts and their Urdu counterparts, designed to reflect socio-cultural contexts relevant to gender, ethnicity, religion, nationality, disability, profession, age, and socioeconomic categories. Using GPT-3.5 as a candidate generator and GPT-4o-mini as a PRM-based bias and utility scorer, we provide an extensive quantitative analysis of bias reduction, utility preservation, and cross-lingual disparities. Our findings show: (a) substantial gains over the baseline for both languages; (b) consistently lower fairness scores for Urdu across all methods, highlighting structural inequities in multilingual LLM training; and (c) distinct improvement trajectories between PRM-Select and PRM-Sequential. The study contributes an extensible methodology, interpretable metrics, and cross-lingual comparisons that can support future work on fairness evaluation in low-resource languages.
High memory demands of generative language models have drawn attention to quantization, which reduces computational cost, memory usage, and latency by mapping model weights to lower-precision integers. Approaches such as GPTQ effectively minimize input-weight product errors during quantization; however, recent empirical studies show that they can increase biased outputs and degrade performance on fairness benchmarks, and it remains unclear which specific weights cause this issue. In this work, we draw new links between quantization and model fairness by adding explicit group-fairness constraints to the quantization objective and introduce Fair-GPTQ, the first quantization method explicitly designed to reduce unfairness in large language models. The added constraints guide the learning of the rounding operation toward less-biased text generation for protected groups. Specifically, we focus on stereotype generation involving occupational bias and discriminatory language spanning gender, race, and religion. Fair-GPTQ has minimal impact on performance, preserving at least 90% of baseline accuracy on zero-shot benchmarks, reduces unfairness relative to a half-precision model, and retains the memory and speed benefits of 4-bit quantization. We also compare the performance of Fair-GPTQ with existing debiasing methods and find that it achieves performance on par with the iterative null-space projection debiasing approach on racial-stereotype benchmarks. Overall, the results validate our theoretical solution to the quantization problem with a group-bias term, highlight its applicability for reducing group bias at quantization time in generative models, and demonstrate that our approach can further be used to analyze channel- and weight-level contributions to fairness during quantization.
Sharmishtha Raghuvanshi, Supriya Sanjay Nikam, Manisha Karne
et al.
This research examines the persistent challenge of health inequalities in India, departing from the conventional focus on aggregate improvements in mortality rates. While India has achieved progress in overall health indicators since independence, the distribution of health outcomes remains uneven, a fact starkly highlighted by the COVID-19 pandemic. This study investigates the socio-economic determinants of health disparities using the National Family and Health Survey (NFHS)-5 data from 2019-20, focusing on both national and state-level analyses, specifically for Maharashtra. Employing a health economics framework, the analysis delves into individual-level data, population shares, self-reported morbidity prevalence, and treatment patterns across diverse socio-economic groups. Regression analyses, stratified by gender, are conducted to quantify the impact of socio-economic factors on reported morbidity. Furthermore, a Fairlie decomposition, an extension of the Oaxaca decomposition, is utilised to dissect the gender gap in morbidity, assessing the extent to which observed differences are attributable to explanatory variables. The findings reveal a significant burden of self-reported morbidity, with approximately one in nine individuals in India and one in eight in Maharashtra reporting morbidity. Notably, women exhibit nearly double the morbidity rate compared to men. The decomposition analysis identifies key drivers of gender disparities. In India, marital status exacerbates these differences, while insurance coverage, caste, urban residence, and wealth mitigate them. In Maharashtra, urban residence and marital status widen the gap, whereas religion, caste, and insurance coverage narrow it. This research underscores the importance of targeted policy interventions to address the complex interplay of socio-economic factors driving health inequalities in India.
As large language models (LLMs) and LLM-based agents increasingly interact with humans in decision-making contexts, understanding the trust dynamics between humans and AI agents becomes a central concern. While considerable literature studies how humans trust AI agents, it is much less understood how LLM-based agents develop effective trust in humans. LLM-based agents likely rely on some sort of implicit effective trust in trust-related contexts (e.g., evaluating individual loan applications) to assist and affect decision making. Using established behavioral theories, we develop an approach that studies whether LLMs trust depends on the three major trustworthiness dimensions: competence, benevolence and integrity of the human subject. We also study how demographic variables affect effective trust. Across 43,200 simulated experiments, for five popular language models, across five different scenarios we find that LLM trust development shows an overall similarity to human trust development. We find that in most, but not all cases, LLM trust is strongly predicted by trustworthiness, and in some cases also biased by age, religion and gender, especially in financial scenarios. This is particularly true for scenarios common in the literature and for newer models. While the overall patterns align with human-like mechanisms of effective trust formation, different models exhibit variation in how they estimate trust; in some cases, trustworthiness and demographic factors are weak predictors of effective trust. These findings call for a better understanding of AI-to-human trust dynamics and monitoring of biases and trust development patterns to prevent unintended and potentially harmful outcomes in trust-sensitive applications of AI.
This article aims to formulate a philosophical problem that is grounded upon the Pan-Abrahamic nature of early Islam, focusing on the implications that this has for understanding the identity of the contemporary Islamic community. This philosophical problem—termed the Pan-Abrahamic Problem—is structured around the examination of Prophet Muhammad’s leadership and the inclusivity of the early Islamic community, as proposed by Fred Donner in the form of the Pan-Abrahamic Thesis. The formulation of this philosophical problem is presented through the lens of the philosophical criteria of continuity and connectedness of aims (doctrine) and organisation, as proposed by Richard Swinburne. This philosophical problem will, thus, offer a challenge against traditional exclusivist narratives within Islam, ultimately aiming to emphasise the inclusive and pluralistic foundation of the religion and the significance of this for the contemporary Islamic identity.
The ability for individuals to constructively engage with one another across lines of difference is a critical feature of a healthy pluralistic society. This is also true in online discussion spaces like social media platforms. To date, much social media research has focused on preventing ills -- like political polarization and the spread of misinformation. While this is important, enhancing the quality of online public discourse requires not just reducing ills but also promoting foundational human virtues. In this study, we focus on one particular virtue: ``intellectual humility'' (IH), or acknowledging the potential limitations in one's own beliefs. Specifically, we explore the development of computational methods for measuring IH at scale. We manually curate and validate an IH codebook on 350 posts about religion drawn from subreddits and use them to develop LLM-based models for automating this measurement. Our best model achieves a Macro-F1 score of 0.64 across labels (and 0.70 when predicting IH/IA/Neutral at the coarse level), higher than an expected naive baseline of 0.51 (0.32 for IH/IA/Neutral) but lower than a human annotator-informed upper bound of 0.85 (0.83 for IH/IA/Neutral). Our results both highlight the challenging nature of detecting IH online -- opening the door to new directions in NLP research -- and also lay a foundation for computational social science researchers interested in analyzing and fostering more IH in online public discourse.
Chandreyee Roy, Hang-Hyun Jo, János Kertész
et al.
Members of a society can be characterized by a large number of features, such as gender, age, ethnicity, religion, social status, and shared activities. One of the main tie-forming factors between individuals in human societies is homophily, the tendency of being attracted to similar others. Homophily has been mainly studied with focus on one of the features and little is known about the roles of similarities of different origins in the formation of communities. To close this gap, we analyze three datasets from Information and Communications Technology (ICT) services, namely, two online social networks and a network deduced from mobile phone calls, in all of which metadata about individual features are available. We identify communities within egocentric networks and surprisingly find that the larger the community is, the more overlap is found between features of its members and the ego. We interpret this finding in terms of the effort needed to manage the communities; the larger diversity requires more effort such that to maintain a large diverse group may exceed the capacity of the members. As the ego reaches out to her alters on an ICT service, we observe that the first alter in each community tends to have a higher feature overlap with the ego than the rest. Moreover the feature overlap of the ego with all her alters displays a non-monotonic behaviors as a function of the ego's degree. We propose a simple mechanism of how people add links in their egocentric networks of alters that reproduces all the empirical observations and shows the reason behind non-monotonic tendency of the egocentric feature overlap as a function of the ego's degree.
Objectives: This research investigates the most important efforts made by the Abbasid Caliphs against the people of innovation and deviation. They had successful efforts throughout the Abbasid era to eliminate the deviants and prevent them from carrying out their activities. The current research also aimed to investigate how the deviants had worked to undermine the Abbasid authority and the Islamic religion, and how they exaggerated in their words about religion and began to spread their deviant ideas and beliefs in an attempt to restore their glory and ancient authority. It also aimed to shed light on how the Abbasid caliphs face their danger, limit their influence, and eliminate some of their leaders. Methodology: To achieve the objectives of this study, the researchers adopted the historical descriptive approach when presenting the position of the Abbasid Caliphate on intellectual deviation, represented by the position of the Abbasid Caliphs against the intellectual deviation that appeared clearly in the Abbasid era by referring to specialized historical sources and references, dealing accurately while employing those sources to reach the most accurate information and clarify the position of the Abbasid Caliphate on those deviations.Results: The study showed many important results that were obtained during the research period, perhaps the most prominent of which are the following:1- The Abbasid era witnessed dangerous deviant intellectual movements carried out by the enemies of Islam to undermine Arabism and Islamic culture and strike the authority of the Arabs, by entering the ranks of Islamic sects and inclining them towards intellectual deviation.2- The Abbasid Caliphs worked to make efforts to eliminate them and work to cut off their continuous movements throughout the Abbasid era.3- Ending the issue of saying that the Qur’an was created, through the successful4- The Abbasid sultans had similar efforts to the efforts of the caliphs to eliminate heresies and deviations by the Ghaznavid and Seljuk sultans and Sultan Salahuddin Al-Ayyubi.Conclusion: The study recommends the necessity of emphasizing on researchers to study intellectual deviations in depth and to know the reasons that led to their emergence, who is behind them and the goals they seek to achieve. It also recommend to work on refuting the claims and falsehoods of those movements and rejecting their ideas that are foreign to Islamic society.
History of scholarship and learning. The humanities
In the Jeju shamanic religion, chickens have been sacrificed for <i>aekmagi,</i> a ritual to prevent <i>aek</i>, a looming misfortune that may cause death. Whereas ordinary participants are thought to be at risk of harm when possessing or eating chickens or other offerings made to prevent <i>aek</i>, the <i>simbang</i>, Jeju shamans, are thought to be immune to it. <i>Simbang</i> are believed to be permanently on the threshold between the human and the divine realms. They help remove <i>aek</i> but are not harmed by it, because it only harms humans in the human realm, not the person on the boundary. While the other participants are temporarily placed in the liminal state during <i>aekmagi</i> and come back to the ordinary living human realm after the ritual, <i>simbang</i> remain in the perpetual liminal state. Chicken sacrifice has been omitted from <i>aekmagi</i> since around 2010 in most places in Jeju-do. Though ritual killing is no longer practiced, adherents still think that <i>aek</i> is prevented by <i>aekmagi</i>. The Jeju people believe that gods are the main agents of preventing <i>aek</i> and that they can persuade the gods to do the work without receiving chickens’ lives. In addition, due to the change in people’s view on killing animals, <i>aekmagi</i> without chicken sacrifice has become a more efficient ritual system for nourishing social sustenance by following the new social prescription.
The recent surge in the research of diffusion models has accelerated the adoption of text-to-image models in various Artificial Intelligence Generated Content (AIGC) commercial products. While these exceptional AIGC products are gaining increasing recognition and sparking enthusiasm among consumers, the questions regarding whether, when, and how these models might unintentionally reinforce existing societal stereotypes remain largely unaddressed. Motivated by recent advancements in language agents, here we introduce a novel agent architecture tailored for stereotype detection in text-to-image models. This versatile agent architecture is capable of accommodating free-form detection tasks and can autonomously invoke various tools to facilitate the entire process, from generating corresponding instructions and images, to detecting stereotypes. We build the stereotype-relevant benchmark based on multiple open-text datasets, and apply this architecture to commercial products and popular open source text-to-image models. We find that these models often display serious stereotypes when it comes to certain prompts about personal characteristics, social cultural context and crime-related aspects. In summary, these empirical findings underscore the pervasive existence of stereotypes across social dimensions, including gender, race, and religion, which not only validate the effectiveness of our proposed approach, but also emphasize the critical necessity of addressing potential ethical risks in the burgeoning realm of AIGC. As AIGC continues its rapid expansion trajectory, with new models and plugins emerging daily in staggering numbers, the challenge lies in the timely detection and mitigation of potential biases within these models.
As society transitions towards an AI-based decision-making infrastructure, an ever-increasing number of decisions once under control of humans are now delegated to automated systems. Even though such developments make various parts of society more efficient, a large body of evidence suggests that a great deal of care needs to be taken to make such automated decision-making systems fair and equitable, namely, taking into account sensitive attributes such as gender, race, and religion. In this paper, we study a specific decision-making task called outcome control in which an automated system aims to optimize an outcome variable $Y$ while being fair and equitable. The interest in such a setting ranges from interventions related to criminal justice and welfare, all the way to clinical decision-making and public health. In this paper, we first analyze through causal lenses the notion of benefit, which captures how much a specific individual would benefit from a positive decision, counterfactually speaking, when contrasted with an alternative, negative one. We introduce the notion of benefit fairness, which can be seen as the minimal fairness requirement in decision-making, and develop an algorithm for satisfying it. We then note that the benefit itself may be influenced by the protected attribute, and propose causal tools which can be used to analyze this. Finally, if some of the variations of the protected attribute in the benefit are considered as discriminatory, the notion of benefit fairness may need to be strengthened, which leads us to articulating a notion of causal benefit fairness. Using this notion, we develop a new optimization procedure capable of maximizing $Y$ while ascertaining causal fairness in the decision process.
Prepared by the ExoPAG Science Interest Group, 2 on Exoplanet Demographics, Jessie L. Christiansen
et al.
Goal 1 of the National Academies of Science, Engineering and Mathematics Exoplanet Science Strategy is "to understand the formation and evolution of planetary systems as products of the process of star formation, and characterize and explain the diversity of planetary system architectures, planetary compositions, and planetary environments produced by these processes", with the finding that "Current knowledge of the demographics and characteristics of planets and their systems is substantially incomplete." One significant roadblock to our ongoing efforts to improve our demographics analyses is the lack of comprehensive meta-data accompanying published exoplanet surveys. The Exoplanet Program Analysis Group (ExoPAG) Science Interest Group 2: Exoplanet Demographics has prepared this document to provide guidance to survey architects, authors, referees and funding agencies as to the most valuable such data products for five different exoplanet detection techniques - transit, radial velocity, direct imaging, microlensing and astrometry. We find that making these additional data easily available would greatly enhance the community's ability to perform robust, reproducible demographics analyses, and make progress on achieving the most important goals identified by the exoplanet and wider astronomical community.
Marcelo Sartori Locatelli, Pedro Calais, Matheus Prado Miranda
et al.
Politicization is a social phenomenon studied by political science characterized by the extent to which ideas and facts are given a political tone. A range of topics, such as climate change, religion and vaccines has been subject to increasing politicization in the media and social media platforms. In this work, we propose a computational method for assessing politicization in online conversations based on topic shifts, i.e., the degree to which people switch topics in online conversations. The intuition is that topic shifts from a non-political topic to politics are a direct measure of politicization -- making something political, and that the more people switch conversations to politics, the more they perceive politics as playing a vital role in their daily lives. A fundamental challenge that must be addressed when one studies politicization in social media is that, a priori, any topic may be politicized. Hence, any keyword-based method or even machine learning approaches that rely on topic labels to classify topics are expensive to run and potentially ineffective. Instead, we learn from a seed of political keywords and use Positive-Unlabeled (PU) Learning to detect political comments in reaction to non-political news articles posted on Twitter, YouTube, and TikTok during the 2022 Brazilian presidential elections. Our findings indicate that all platforms show evidence of politicization as discussion around topics adjacent to politics such as economy, crime and drugs tend to shift to politics. Even the least politicized topics had the rate in which their topics shift to politics increased in the lead up to the elections and after other political events in Brazil -- an evidence of politicization.
The medievalist caricature of sexual regress in Late-Medieval female monasteries: This article confronts the widely published medievalist caricature of sexual regress in Late-Medieval female monasteries by presenting a statistical analysis of the relatively low (measured against the Early and Central Middle Ages) frequency of sexual contact between monks and nuns, monks and monks, and nuns and nuns in 15th century England. C.H. Knudsen’s examination of the pastoral register of the bishop of Lincoln, William Alnwick, in the period from 1436 to 1449 is utilised to counter the common, yet profoundly modernist notion of the Late-Medieval ‘wayward nun’. Five idea-historical developments from the Early and Central Middle Ages are presented as a backdrop to this statistical analysis, showing that sexual encounters in monasteries in the Early to Central Middle Ages in the Latin West occurred more often than merely sporadic. Having defined medievalism as ‘post-Medieval ideological-reductionist and anachronistic reconstructions of the Middle Ages, whereby the Middle Ages is essentialised by one or more contingencies’, it becomes clear that the notion of ‘sexual regress in Late Medieval female monasteries’ with the image of the ‘wayward nun’ centralised therein, points to a form of medievalism: a single contingent aspect of Medieval female monasteries – the occurrence of sexual contact, however discreet – is used to present a fabricated totality of a complex socio-historical context. How complex this historical context indeed is, becomes apparent in Knudsen’s analysis of the bishop of Lincoln’s pastoral register during his 79 visits to 70 monasteries and interrogations of 217 nuns and 528 monks. Concluding that the ‘promiscuous monk’ was a far more general phenomenon than the ‘wayward nun’ in the Later Middle Ages, Knudsen’s analysis confirms that the Middle Ages is still as much a domain of research as it is a realm of fantasy today. The modernist fixation on the Late-Medieval ‘wayward nun’ is, for example, expressed in Heinrich Lossow’s (1843–1897) provocative painting Die Versündigung (ca.1880). It is argued that the ‘wayward nun’ in Lossow’s painting was a self-conscious attempt to escape from the impasse created by Victorian sexual repression: just as in every other 19th and early 20th century representation of sexual regress in Late-Medieval female monasteries, ‘she’ was nothing more than vulgar fiction.
Intradisciplinary and/or interdisciplinary implications: This critique of the medievalist caricature of sexual regress in Late-Medieval female monasteries overlaps with a variety of philosophical and theological disciplines, including Medieval philosophy, Medieval history, church history, patristics, philosophy of religion and sociology of religion. Whenever these proximate disciplines are impacted by niche Medieval research, it may hold implications that these disciplines could take note of.
Among the wealth of buildings consecrated to the Shinto religion, Yamaguchi has a Christian place of worship: the Sanctuary of Saint Francis Xavier. The lot chosen is the same on which the previous church stood, symbol of the city and destroyed by a fire in 1991. Consonances and relations could be found with spaces that have wished to express with force the tension towards the Absolute not only in our Western world but also in the spaces of Oriental architecture, in particular Japanese, where the striving towards that spoliation and essentialness of things that is, a bottom, a thirst for truth is perceived. Finally, our effort had as its aim to make the architecture speak a universal language of the hearth of man, rapt to the infinite desire to experience beauty and to find himself once again within it to have an authentic experience of interior joy.
In this paper, we extend Gary Becker's empirical analysis of the marriage market to same-sex couples. Becker's theory rationalizes the well-known phenomenon of homogamy among different-sex couples: individuals mate with their likes because many characteristics, such as education, consumption behaviour, desire to nurture children, religion, etc., exhibit strong complementarities in the household production function. However, because of asymmetries in the distributions of male and female characteristics, men and women may need to marry "up" or "down" according to the relative shortage of their characteristics among the populations of men and women. Yet, among same-sex couples, this limitation does not exist as partners are drawn from the same population, and thus the theory of assortative mating would boldly predict that individuals will choose a partner with nearly identical characteristics. Empirical evidence suggests a very different picture: a robust stylized fact is that the correlation of the characteristics is in fact weaker among same-sex couples. In this paper, we build an equilibrium model of same-sex marriage market which allows for straightforward identification of the gains to marriage. We estimate the model with 2008-2012 ACS data on California and show that positive assortative mating is weaker for homosexuals than for heterosexuals with respect to age and race. Our results suggest that positive assortative mating with respect to education is stronger among lesbians, and not significantly different when comparing gay men and married different-sex couples. As regards labor market outcomes, such as hourly wages and working hours, we find some indications that the process of specialization within the household mainly applies to different-sex couples.
Societal ideas and trends dictate media narratives and cinematic depictions which in turn influences people's beliefs and perceptions of the real world. Media portrayal of culture, education, government, religion, and family affect their function and evolution over time as people interpret and perceive these representations and incorporate them into their beliefs and actions. It is important to study media depictions of these social structures so that they do not propagate or reinforce negative stereotypes, or discriminate against any demographic section. In this work, we examine media representation of professions and provide computational insights into their incidence, and sentiment expressed, in entertainment media content. We create a searchable taxonomy of professional groups and titles to facilitate their retrieval from speaker-agnostic text passages like movie and television (TV) show subtitles. We leverage this taxonomy and relevant natural language processing (NLP) models to create a corpus of professional mentions in media content, spanning more than 136,000 IMDb titles over seven decades (1950-2017). We analyze the frequency and sentiment trends of different occupations, study the effect of media attributes like genre, country of production, and title type on these trends, and investigate if the incidence of professions in media subtitles correlate with their real-world employment statistics. We observe increased media mentions of STEM, arts, sports, and entertainment occupations in the analyzed subtitles, and a decreased frequency of manual labor jobs and military occupations. The sentiment expressed toward lawyers, police, and doctors is becoming negative over time, whereas astronauts, musicians, singers, and engineers are mentioned favorably. Professions that employ more people have increased media frequency, supporting our hypothesis that media acts as a mirror to society.
Mantello Peter, Manh-Tung Ho, Minh-Hoang Nguyen
et al.
Human resource management technologies have moved from biometric surveillance to emotional artificial intelligence (AI) that monitor employees' engagement and productivity, analyze video interviews and CVs of job applicants. The rise of the US$20 billion emotional AI industry will transform the future workplace. Yet, besides no international consensus on the principles or standards for such technologies, there is a lack of cross-cultural research on future job seekers' attitude toward such use of AI technologies. This study collects a cross-sectional dataset of 1,015 survey responses of international students from 48 countries and 8 regions worldwide. A majority of the respondents (52%) are concerned about being managed by AI. Following the hypothetico-deductivist philosophy of science, we use the MCMC Hamiltonian approach and conduct a detailed comparison of 10 Bayesian network models with the PSIS-LOO method. We consistently find having a higher income, being male, majoring in business, and/or self-rated familiarity with AI correlate with a more positive view of emotional AI in the workplace. There is also a stark cross-cultural and cross-regional difference. Our analysis shows people from economically less developed regions (Africa, Oceania, Central Asia) tend to exhibit less concern for AI managers. And for East Asian countries, 64% of the Japanese, 56% of the South Korean, and 42% of the Chinese professed the trusting attitude. In contrast, an overwhelming majority of 75% of the European and Northern American possesses the worrying/neutral attitude toward being managed by AI. Regarding religion, Muslim students correlate with the most concern toward emotional AI in the workplace. When religiosity is higher, the correlation becomes stronger for Muslim and Buddhist students.
Aleksandra Kielan, Aleksandra Kielan, Mariusz Jaworski
et al.
Depression is the most common psychiatric disorder in people who die by suicide. Awareness of risk factors for suicide in depression is important for clinicians. The study was aimed at establishing models of factors related to the level of depression and suicidal behavior among men from three different groups—in men with depressive disorder, in comparison to men with physical disorder and healthy men. A total of 598 men were included in the study. The following questionnaires were used in research model: test with sociodemographic variables, AUDIT Test, Fagerström Test, Generalized Self-Efficacy Scale (GSES), Inventory for Measuring Coping with Stress (Mini-COPE), Resilience Evaluation Questionnaire (KOP-26), Suicide Behaviors Questionnaire—Revised (SBQ-R) by Osman, and Gotland Male Depression Scale. In men with depression, the positive factors strongly related to the intensity of depression and suicidal behavior were as follows: vocational education, active coping, turning toward religion, social competence for resilience, and bachelor status. The factors negatively related to the intensity of depression and suicidal behavior in this group were as follows: unemployed status, student status, low satisfaction with the financial situation, having children, history of mental disorders in family, alcohol addiction, and seeking instrumental support. In the group of men with physical disorders, the following protection factors were identified: the medium or small city as a place of living, active coping, venting, and personal competence. The following risk factors were identified in this group: psychiatric treatment in the past. In the group of healthy men, the following protective factors were identified: the medium city as a place of living, positive reappraisal, planning abilities, and personal and social competence for resilience. In this group, the following risk factors were identified: vocational and higher education, student status, satisfaction with the financial situation, having more than one children, the occurrence of mental disorders in the family, the occurrence of alcohol abuse in the family, and use of psychoactive substances as a strategy of dealing with stress. The risk factors identified in this study should be included in the clinical assessment of depression and suicidal behavior risk in male patients. There are some protective factors identified, including productive coping and personal and social competencies, which can be developed and should be especially considered and strengthened in mental health promotion programs aimed at men.