arXiv subjects

Fariba Karimi

Publications and source records attributed to Fariba Karimi.

At least 19 recordsLinked to original sources

arXiv2026

Systematic comparison of gender inequality in scientific rankings across disciplines

Participation of Women in academia has grown over recent decades. Yet, it is unclear how this growth translates into representation at the top of academic rankings (measured by scientific productivity and citations). Here, we investigate gender gaps in productivity, citations, and coauthorship networks across 18 fields, using 67.7 million papers published between 1975 and 2020 in the Semantic Scholar Open Research Corpus, with the authors binary gender inferred from names using Genderize and Namsor. We find that women remain consistently underrepresented in top-ranked positions across all fields, even in disciplines where their overall participation is relatively high. We observe that rankings are generally becoming increasingly rigid over time, with fewer researchers entering or leaving top-ranked positions from one year to the next in most fields, although we do not test whether this rigidity contributes to the disparities we document. Across fields, highly productive men receive more citations than the closest available women based on the observed productivity and career stages. However, when top-ranked women are compared with their closest male counterparts in terms of career stage and publication profile (accounting for venue prestige and authorship position), gaps narrow in some fields and in a few others disappear or reverse, showing that, among highly productive researchers, differences in research output alone do not fully account for the citation gaps.

cs.SI↗

arXiv2026

From Network Inequality to Network Fairness: A Perspective on Responsible Decision-Making

Social networks shape how individuals make decisions and how opportunities are distributed. However, the mechanisms that generate these networks often reflect pre-existing inequalities, and technologies that rely on network-derived signals risk further amplifying such disparities. Algorithmic fairness research largely treats networks as a fixed background, grounding analysis almost exclusively in distributive justice and overlooking how network structures systematically bias decision-making. In this Perspective, we identify ten network effects and trace how they create structural biases in the relationship between what we intend to measure and what we observe. Using academic hiring as an example, we show that network biases are not inherently harmful or beneficial. Determining their legitimacy requires examining the entire decision-making process through the lenses of both distributive and procedural justice while engaging all affected stakeholders. We therefore call for a holistic, networked approach to fairness that moves beyond static group categories and recognizes the dynamic, relational, and structural nature of inequality.

cs.SI↗

arXiv2026

Mapping urban segregation through co-residence network reconstruction

Urban segregation poses a critical challenge for cities, exacerbating inequalities, social tensions, fears, and polarisation. It emerges from the interplay of socio-economic disparities, housing constraints, and residential preferences, and can disproportionately affect migrant communities. Here, we study residential segregation in Vienna using a city-wide administrative snapshot of registered residents, covering the full foreign population and Austrian nationals at the district level. We introduce a network-based approach by constructing a statistically validated co-residence network in which nodes represent nationalities and links capture whether pairs of groups live in the same districts more or less often than expected under a population-size-preserving null model. Applying community detection to this network reveals two major clusters of nationalities with distinct co-residence patterns. These clusters are systematically associated with district-level income disparities and diversity, while also reflecting the geographical proximity of countries of origin, with nationalities from nearby regions tending to share similar residential patterns within Vienna. Our results show how network methods can provide an intuitive and interpretable map of urban residential sorting, complementing traditional segregation indices and highlighting the multiple dimensions underlying migrant integration in diverse cities.

physics.soc-ph↗

arXiv2026

How large should academic departments be?

Academic departments are the primary unit of scholarship and education at universities, and they vary vastly in their sizes. However, the consequences and natural dynamics of department size are poorly understood. Small departments face disproportionate teaching and administrative overhead per faculty member, while large ones face coordination costs and thematic incoherence. Here, we characterize and model the dynamics of academic department sizes using $14,000$ U.S.-based departments in eight academic domains. Across all domains, similar broad-tailed distributions reveal a common size range from 4 to 23 faculty members, widening across domains at its upper border. Annual size-dependent closure risks and growth rates indicate that stability is greatest in this size range: below it, small departments either close or grow quickly; within it, closure risk is low and sizes stabilize; above it, large departments can persist, with marginal attrition and minimal closure risk. An analytically tractable model of size-dependent coagulation and fragmentation, informed only by the aggregated size distribution, reproduces department dynamics across the full size range. Rescaling each domain by its most stable size reveals a common regression toward the stable range across most domains. Our results establish academic departments as organizations with natural size dynamics defined by a grow-or-close pattern for the smallest departments, and a weak pressure against unlimited growth for the largest.

physics.soc-ph↗

arXiv2026

Higher-order modeling of face-to-face interactions

The most fundamental social interactions among humans occur face-to-face. Their features have been extensively studied in recent years, owing to the availability of high-resolution data on individuals' proximity. Mathematical models based on mobile agents have been crucial to understanding the spatio-temporal organization of face-to-face interactions. However, these models focus on dyadic relationships only, failing to characterize interactions in larger groups of individuals. Here, we propose a model in which agents interact with each other by forming groups of different sizes. Each group has a degree of social attractiveness, based on which neighboring agents decide whether to join. Our framework reproduces different properties of groups in face-to-face interactions, including their distribution, the correlation in their number, and their persistence in time, which dyadic models cannot replicate. Furthermore, it captures homophilic patterns at the level of higher-order interactions, going beyond standard pairwise approaches. Our work provides further evidence that higher-order interactions are key to describe human face-to-face contacts, paving the way for further investigation of how group dynamics at a microscopic scale affects social phenomena at a macroscopic scale.

physics.soc-ph↗

arXiv2026

Contact, conflict, or opportunity? Out-group exposure creates tie opportunity, not tolerance

Three theories offer competing predictions about how people respond to growing diversity in their social environment. Contact theory suggests more exposure to out-groups reduces prejudice; conflict theory predicts a stronger in-group preference; structural opportunity theory argues that shifts in behaviour only reflect changes in the opportunity structure rather than in underlying preference. We test these predictions using friendship and rejection nominations from nearly 5,000 students in 228 classrooms, across gender, ethnicity, and socio-economic status. We estimate individual preference using a multilevel model based on the Wallenius hypergeometric distribution, which accounts for the finite, asymmetric pool of potential ties. Results show that for ethnicity and socio-economic status, preferences are largely unaffected by classroom composition. For gender, however, same-gender preference strengthens as the out-group increases, supporting conflict theory. This means greater diversity does not necessarily change the intrinsic preference of students toward out-group peers, but creates more opportunities for cross-group interactions.

physics.soc-ph↗

arXiv2026

Academic collaborations and movements towards successful careers in physics

Collaboration networks evolve throughout academic careers, yet few studies systematically examine how these network dynamics relate to long-term career success and mobility. Analysing 35,708 physicists' careers spanning at least 15 years, we use time series clustering to identify ten distinct evolution patterns of network size and clustering coefficient across career years 5 to 15. We report three key results. First, authors who begin with loosely connected networks and progressively tighten their networks while expanding network size during mid-career achieve the highest PI attainment rates, publication output, and citation impact. Second, despite different starting points, network evolution patterns associated with better outcomes converge toward moderate clustering by career year 15, suggesting an optimal balance between core team cohesion and diverse external connections. Third, mobility is positively associated with these successful network evolution patterns and remains positively associated with scientific outcomes even after controlling for network evolution patterns.

physics.soc-ph↗

arXiv2025

Stronger together? The homophily trap in networks

While homophily -- the tendency to link with similar others -- may nurture a sense of belonging and shared values, it can also hinder diversity and widen inequalities. Here, we unravel this trade-off analytically, revealing homophily traps for minority groups: scenarios where increased homophilic interaction among minorities negatively affects their structural opportunities within a network. We demonstrate that homophily traps arise when minority size falls below 25% of a network, at which point homophily comes at the expense of lower structural visibility for the minority group. Our work reveals that social groups require a critical size to benefit from homophily without incurring structural costs, providing insights into core processes underlying the emergence of group inequality in networks.

cs.SI↗

arXiv2025

A Deep Latent Factor Graph Clustering with Fairness-Utility Trade-off Perspective

Fair graph clustering seeks partitions that respect network structure while maintaining proportional representation across sensitive groups, with applications spanning community detection, team formation, resource allocation, and social network analysis. Many existing approaches enforce rigid constraints or rely on multi-stage pipelines (e.g., spectral embedding followed by $k$-means), limiting trade-off control, interpretability, and scalability. We introduce \emph{DFNMF}, an end-to-end deep nonnegative tri-factorization tailored to graphs that directly optimizes cluster assignments with a soft statistical-parity regularizer. A single parameter $λ$ tunes the fairness--utility balance, while nonnegativity yields parts-based factors and transparent soft memberships. The optimization uses sparse-friendly alternating updates and scales near-linearly with the number of edges. Across synthetic and real networks, DFNMF achieves substantially higher group balance at comparable modularity, often dominating state-of-the-art baselines on the Pareto front. The code is available at https://github.com/SiamakGhodsi/DFNMF.git.

cs.LG↗

arXiv2025

Network Inequality through Preferential Attachment, Triadic Closure, and Homophily

Inequalities in social networks arise from linking mechanisms, such as preferential attachment (connecting to popular nodes), homophily (connecting to similar others), and triadic closure (connecting through mutual contacts). While preferential attachment mainly drives degree inequality and homophily drives segregation, their three-way interaction remains understudied. This gap limits our understanding of how network inequalities emerge. Here, we introduce PATCH, a network growth model combining the three mechanisms to understand how they create disparities among two groups in synthetic networks. Extensive simulations confirm that homophily and preferential attachment increase segregation and degree inequalities, while triadic closure has countervailing effects: conditional on the other mechanisms, it amplifies population-wide degree inequality while reducing segregation and between-group degree disparities. We demonstrate PATCH's explanatory potential on fifty years of Physics and Computer Science collaboration and citation networks exhibiting persistent gender disparities. PATCH accounts for these gender disparities with the joint presence of preferential attachment, moderate gender homophily, and varying levels of triadic closure. By connecting mechanisms to observed inequalities, PATCH shows how their interplay sustains group disparities and provides a framework for designing interventions that promote more equitable social networks.

physics.soc-ph↗

arXiv2025

Whose Name Comes Up? Auditing LLM-Based Scholar Recommendations

This paper evaluates the performance of six open-weight LLMs (llama3-8b, llama3.1-8b, gemma2-9b, mixtral-8x7b, llama3-70b, llama3.1-70b) in recommending experts in physics across five tasks: top-k experts by field, influential scientists by discipline, epoch, seniority, and scholar counterparts. The evaluation examines consistency, factuality, and biases related to gender, ethnicity, academic popularity, and scholar similarity. Using ground-truth data from the American Physical Society and OpenAlex, we establish scholarly benchmarks by comparing model outputs to real-world academic records. Our analysis reveals inconsistencies and biases across all models. mixtral-8x7b produces the most stable outputs, while llama3.1-70b shows the highest variability. Many models exhibit duplication, and some, particularly gemma2-9b and llama3.1-8b, struggle with formatting errors. LLMs generally recommend real scientists, but accuracy drops in field-, epoch-, and seniority-specific queries, consistently favoring senior scholars. Representation biases persist, replicating gender imbalances (reflecting male predominance), under-representing Asian scientists, and over-representing White scholars. Despite some diversity in institutional and collaboration networks, models favor highly cited and productive scholars, reinforcing the rich-getricher effect while offering limited geographical representation. These findings highlight the need to improve LLMs for more reliable and equitable scholarly recommendations.

cs.CY↗

arXiv2025

hyperFA*IR: A hypergeometric approach to fair rankings with finite candidate pool

Ranking algorithms play a pivotal role in decision-making processes across diverse domains, from search engines to job applications. When rankings directly impact individuals, ensuring fairness becomes essential, particularly for groups that are marginalised or misrepresented in the data. Most of the existing group fairness frameworks often rely on ensuring proportional representation of protected groups. However, these approaches face limitations in accounting for the stochastic nature of ranking processes or the finite size of candidate pools. To this end, we present hyperFA*IR, a framework for assessing and enforcing fairness in rankings drawn from a finite set of candidates. It relies on a generative process based on the hypergeometric distribution, which models real-world scenarios by sampling without replacement from fixed group sizes. This approach improves fairness assessment when top-$k$ selections are large relative to the pool or when protected groups are small. We compare our approach to the widely used binomial model, which treats each draw as independent with fixed probability, and demonstrate$-$both analytically and empirically$-$that our method more accurately reproduces the statistical properties of sampling from a finite population. To operationalise this framework, we propose a Monte Carlo-based algorithm that efficiently detects unfair rankings by avoiding computationally expensive parameter tuning. Finally, we adapt our generative approach to define affirmative action policies by introducing weights into the sampling process.

cs.CY↗

arXiv2025

Gender differences in collaboration and career progression in physics

We examine gender differences in collaboration networks and academic career progression in physics. We use the likelihood and time to become a principal investigator (PI) and the length of an author's career to measure career progression. Utilising logistic regression and accelerated failure time models, we examine whether the effect of collaboration behaviour varies by gender. We find that, controlling for the number of publications, the relationship between collaborative behaviour and career progression is almost the same for men and women. Specifically, we find that those who eventually reach principal investigator (PI) status, tend to have published with more unique collaborators. In contrast, publishing repeatedly with the same highly interconnected collaborators and/or larger number of co-authors per publication is characteristic of shorter career lengths and not attaining PI status. We observe that women tend to collaborate in more tightly connected and larger groups than men. Finally, we observe that women are less likely to attain the status of PI throughout their careers and have a lower survival probability compared to men, which calls for policies to close this crucial gap.

physics.soc-ph↗

arXiv2025

Gender disparities in the dissemination and acquisition of scientific knowledge

Recent research has challenged the widespread belief that gender inequities in academia would disappear simply by increasing the number of women. More complex causes might be at play, embodied in the networked structure of scientific collaborations. Here, we aim to understand the structural inequality between male and female scholars in the dissemination of scientific knowledge. We use a large-scale dataset of academic publications from the American Physical Society (APS) to build a time-varying network of collaborations from 1970 to 2020. We model knowledge dissemination as a contagion process in which scientists become informed based on the propagation of knowledge through their collaborators. We quantify the fairness of the system in terms of how women acquire and diffuse knowledge compared to men. Our results indicate that knowledge acquisition and diffusion are slower for women than expected. We find that the main determinant of women's disadvantage is the gap in the cumulative number of collaborators, highlighting how time creates structural disadvantages that contribute to marginalize women in physics. Our work sheds light on how the dynamics of scientific collaborations shape gender disparities in knowledge dissemination and calls for a deeper understanding on how to intervene to improve fairness and diversity in the scientific community.

physics.soc-ph↗

arXiv2024

Intersectional inequalities in social networks

Social networks are shaped by complex, intersecting identities that drive our connection preferences. These preferences weave networks where certain groups hold privileged positions, while others become marginalized. While previous research has examined the impact of single-dimensional identities on inequalities of social capital, social disparities accumulate nonlinearly, further harming individuals at the intersection of multiple disadvantaged groups. However, how multidimensional connection preferences affect network dynamics and in what forms they amplify or attenuate inequalities remains unclear. In this work, we systematically analyze the impact of multidimensionality on social capital inequalities through the lens of intersectionality. To this end, we operationalize several notions of intersectional inequality in networks. Using a network model, we reveal how attribute correlation (or consolidation) combined with biased multidimensional preferences lead to the emergence of counterintuitive patterns of inequality that are unobservable in one-dimensional systems. We calibrate the model with real-world high school friendship data and derive analytical closed-form expressions for the predicted inequalities, finding that the model's predictions match the observed data with remarkable accuracy. These findings hold significant implications for addressing social disparities and inform strategies for creating more equitable networks.

physics.soc-ph↗

arXiv2024

The dynamics of diversity on corporate boards

Diversity in leadership positions, including corporate boards, is an important aspect of equality. It is important because it is the key to better decision-making and innovation, and above all, it paves the way for future generations to participate and shape our society. Many studies emphasize the importance of the visibility of role models and the effect that connectivity has on the success of minorities in leadership. However, the connectivity of firms, the dynamics of the adoption of minorities into leadership positions, and the long-term effects in terms of group dynamics and visibility are not well understood. Here, we present a model that shows how these effects work together in a dynamic model that is calibrated with empirical data of firm and board networks. We show that homophily -- the appointment of minorities is influenced by the presence of minorities in a board and its neighboring entities -- is an important effect shaping the trajectory towards equality. We further show how perception biases and feedback related to the visibility of minority members influence the dynamic. We find that reaching equality can be sped up or slowed down depending on the distribution of minorities in central firms. These insights bear significant implications for policy-making geared towards fostering equality and diversity within corporate boards.

econ.TH↗

arXiv2024

Cumulative Advantage of Brokerage in Academia

Science is a collaborative endeavor in which "who collaborates with whom" profoundly influences scientists' career trajectories and success. Despite its relevance, little is known about how scholars facilitate new collaborations among their peers. In this study, we quantify brokerage in academia and study its effect on the careers of physicists worldwide. We find that early-career participation in brokerage increases later-stage involvement for all researchers, with increasing participation rates and greater career impact among more successful scientists. This cumulative advantage process suggests that brokerage contributes to the unequal distribution of success in academia. Surprisingly, this affects both women and men equally, despite women being more junior in all brokerage roles and lagging behind men's participation due to their late and slow arrival to physics. Because of its cumulative nature, promoting brokerage opportunities to early career scientists might help reduce the inequalities in academic success.

physics.soc-ph↗

arXiv2024

The hidden architecture of connections: How do multidimensional identities shape our social networks?

Our multidimensional identities determine how we interact with each other, shaping social networks through group-based connection preferences. While interactions along single dimensions have been extensively studied, the dynamics driving multidimensional connection preferences remain largely unexplored. In this work, we develop a network model of multidimensional social interactions to tackle two crucial questions: What is the structure of our latent connection preferences, and how do we integrate information from our multidimensional identities to connect with others? To answer these questions, we systematically model different latent preference structures and preference aggregation mechanisms. Then, we compare them using Bayesian model selection by fitting empirical data from high school friendship networks. We find that a simple latent preference model consistently outperforms more complex alternatives. The calibrated model provides robust measures of latent connection preferences in real-world networks, bringing insights into how one- and multidimensional groups interact. Finally, we develop natural operationalizations of dimension salience, revealing which aspects of identity are most relevant for individuals when forming connections.

physics.soc-ph↗