Chetana Sabnis

How do states decide which relationships are "family"? Political science scholarship mostly treats the family as either pre-political or as a site of politics, leaving little discussion of how states constitute family. Addressing this gap, this paper develops the Intimacy Contract, a theory arguing that states selectively recognize relationships and extend entitlements (e.g., inheritance, child custody) to those that resemble the state’s ideological vision of family while dismissing those that do not comport with this vision, regardless of comparable interdependence. I illustrate this theory by analyzing over 6,000 cases involving relationships that lack default legal recognition and whose legitimacy is socially contested: women in extramarital relationships seeking spousal entitlements in India, and non-biological caregivers seeking parental rights in the U.S. I show that courts recognize relationships when they resemble the state’s ideal family. These findings demonstrate that states constitute family, expanding our understanding of state power.

Many social scientists see the institutionalization of monogamous marriage as promising an egalitarian society. This essay challenges this view by arguing that monogamy is an exclusionary institution, extending and withholding recognition and entitlements on an ideological basis. I demonstrate this argument by analyzing the selective recognition of extramarital relationships across 2,721 Indian appellate court cases from 2000 to 2024. When these relationships resemble the state’s ideological vision of monogamous marriage—a conjugal life, a traditional gender dynamic, and shared religiosity—courts are almost 35 percentage points more likely to recognize them and extend entitlements (e.g., spousal support, protection against violence) than when these features are absent. This selective recognition has profound consequences for women experiencing abuse or abandonment in extramarital relationships in India today. Those whose relationships fall outside the mold of ideal monogamous marriage can be left with little or no access to state protection, highlighting the exclusionary logic underlying the system of monogamy that, in theory, promises equality.

This descriptive inference paper examines the ideological conditions under which individuals are recognized as "parents" in the United States. Existing scholarship argues that parental rights are adjudicated through a pragmatic approach with courts attending to particular caregiving practices rather than abstract ideals of family. Others caution that this approach disadvantages families whose caregiving practices are shaped by structural inequalities. I show that while both perspectives correctly identify caregiving as central, neither captures what judges recognize as constituting caregiving. Drawing on more than 4,000 U.S. state appellate court decisions from the last twenty-five years, I find that individuals who invest substantial time, effort, and material resources in a child are more than 50 percentage points more likely to be recognized as parents even when doing so displaces a biological parent. This standard of caregiving is applied similarly to men and women and to kin and non-kin, suggesting that U.S. courts are governed by an ideology of intensive caregiving that predominantly focuses on the extent to which an individual centers the child.

This article examines the use of large language models (LLMs) to collect and classify information in court data. While computational approaches have traditionally relied on pattern matching techniques, these methods struggle to obtain detailed information from court data. Drawing on projects involving more than 2,700 Indian court cases and 4,800 U.S. state cases, I argue that LLMs offer three advantages over traditional approaches: they can identify concepts from natural language descriptions, synthesize information across complex documents, and lower the technical barriers to large-scale text analysis. I develop a framework to use LLMs effectively so that the model draws on relevant information and produces reliable outputs. In a hand-coding validation of 600 cases, the framework produced classifications that mostly aligned with human coding. The article concludes by discussing the ethical considerations, including environmental costs, research labor implications, and risks of concentrating power within a small number of firms.