Daejeon, July 27–31
The history of philosophy has often been narrated as an intellectual network centered on “Great Men” (Collins 1998). From the perspectives of philosophy and feminist philosophy, it has been long argued that core philosophical values—such as reason, objectivity, and universality—have historically been constructed around implicitly male subjects (Le Dœuff 1980; Lloyd 1993; Waithe 2020). Female philosophers have rarely been included in the canon, and even when they are, they have tended to be topologically marginalized in terms of their positions within philosophical networks. This center–periphery structure, through which the visibility and legitimacy of knowledge are produced, is consistent with feminist critiques that foreground the relationship between data, power, and knowledge production (D’Ignazio and Klein 2020). These theoretical critiques suggest the need for empirical investigation into how such gendered structures of knowledge are reproduced in contemporary knowledge infrastructures, including online encyclopedias. Recent quantitative studies, through large-scale analyses of link structures and Wikidata connections, have demonstrated that articles about females tend to exhibit lower in-degree and centrality in hyperlink networks, consequently occupying peripheral positions (Graells-Garrido et al. 2015; Wagner et al. 2015). However, unlike general biographical networks, philosophy's intellectual system is strongly shaped by a privileged core: the historically constituted "canon." In this system, what matters is not overall network centrality, but rather connectivity to the privileged core—access to authenticity. Given this distinctive structure of philosophical knowledge, it is necessary to examine the cost of accessing the canon. This study therefore analyzes gender bias in access to the canon by simulating access through graph structures representing connectivity, using Steiner trees and shortest paths.
To ensure cross-lingual comparability while reflecting a general understanding of the field, we constructed a cross-lingual graph using Wikipedia and Wikidata Suda et al. 2025). We defined the network scope by extracting the first-degree neighborhood (1-hop) of 130 canonical 20th-century philosophers, selected from an authoritative biographical dictionary in Japan. To capture the internal structure of this community, we retrieved all edges between them, thereby including lateral connections between these immediate associates. To ensure a fair comparison, we retained only entities present in all four target languages (EN, FR, DE, JA). We specifically filtered for nodes bearing occupation labels related to philosophy and adjacent intellectual fields (e.g., philosopher, historian of ideas, feminist, and political theorist). This process yielded approximately 500 source nodes (495 nodes), limiting the scope to 20th-century figures to minimize temporal disparities.
This section describes our graph-based method for analyzing structural gender differences. We model Wikipedia's link network as a graph and compute two types of substructures: minimum directed Steiner trees and shortest paths. The former represents a minimal set of links for a philosopher to reach all important reference pages, which implies a minimal knowledge set. The latter captures straightforward connections between pages. Using these measures, we investigate whether gender differences exist in philosophers' structural connectivity across different languages.
Our analysis reveals that while female and male philosophers engage with the canon with equal efficiency, they do so through fundamentally segregated pathways. Contrary to the hypothesis that female philosophers incur higher access costs, our Steiner Tree analysis reveals that connectivity costs are nearly identical between genders across all datasets (Figure 1-(a)). These findings suggest that females do not suffer from lower network efficiency; rather, they operate within a structurally distinct topology, achieving canonical access through equally efficient but separate subnetworks. However, while costs are equal, the path composition is drastically different. Shortest paths originating from female philosophers contain other female entities at a rate 3.4 to 4.4 times higher than paths from males (Figure 1-(b)), and permutation tests confirm that this homophily is statistically significant (p < 0.01) across all four languages. This demonstrates that female philosophers rely on a female genealogy to bridge the topological gap to the canon, effectively creating a segregated access route. Furthermore, ranking analysis identifies the specific nodes driving this segregation (Table 1): in the German dataset, "Feminism" ranks as the 5th most critical hub for female paths but remains unranked for males, and similarly, "Simone de Beauvoir" appears as a top-tier hub for females (e.g., Rank 3 in English) while being peripheral for males (Rank 30). This confirms that female engagement is structurally routed through specific gendered contexts absent from the male mainstream.
Figure 1 Comparison of connectivity costs and gendered path composition across four language editions.
Table 1 Centrality ranking comparison of gender-specific gatekeeper nodes across four languages.
Note: "Unranked" indicates the node did not appear in the top 100 for male philosophers.
By focusing on connectivity to the canon, this study elucidates the structural dimension of gender bias. Our findings suggest that gender bias in the history of philosophy manifests not as a disparity in access efficiency—since connectivity costs were found to be identical—but as a structural segregation of pathways. We show that female philosophers achieve canonical integration through distinct, gender-specific subnetworks rather than shared mainstreams. This approach provides a quantitative framework for understanding how historical exclusion functions through the routing of knowledge.