Glossary

Key terms used across this project, covering early church history and LLM evaluation.

Androcentric: Centred on or privileging the male perspective. In historical scholarship, androcentric sources describe social and religious life primarily through men’s roles and experiences, rendering women’s contributions invisible or marginal.
Apocrypha: Texts related to early Christianity that were not included in the canonical New Testament. Some apocryphal texts — such as the Acts of Paul and Thecla — are significant sources for the study of women in early Christianity.
Cappadocian Fathers: Three fourth-century theologians from the Cappadocia region (modern Turkey): Basil of Caesarea, Gregory of Nyssa, and Gregory of Nazianzus. Their work on Trinitarian theology was foundational for the Council of Constantinople (381 CE). Macrina the Younger is considered by some scholars to have been equally formative.
Deaconess: A woman who held the office of deacon (diakonos, also diakonissa) in the early church. The precise nature and extent of deaconesses’ roles is debated: in some traditions they assisted with baptism of women, visited sick women, and maintained order in the women’s section of worship. In some documented cases (e.g., Olympias) they held institutional authority over monastic communities.
Diakonos: Greek term meaning “servant” or “minister,” used in the New Testament for both men and women (e.g., Phoebe in Romans 16:1). The translation of diakonos as “deacon” when referring to men but “helper” or “servant” when referring to women has been a significant site of contested scholarship.
Gender Bias (in LLMs): The tendency of large language models to reproduce stereotyped, unequal, or historically distorted representations of gender. In the context of this project, gender bias includes: describing women’s roles as secondary or supportive when the historical record indicates leadership; failing to mention female figures; and uncritically reproducing androcentric framings from primary sources.
Hagiography: A genre of biography devoted to the lives of saints, typically written with devotional and rhetorical aims rather than strict historical accuracy. Many of our primary sources for women in early Christianity (e.g., Gregory of Nyssa’s Life of Macrina, the anonymous Life of Olympias) are hagiographies, which requires critical reading of their historical claims.
Hallucination (LLM): A response generated by a large language model that is fluent and confident but factually incorrect, invented, or misattributed. Hallucination is particularly prevalent in domains with sparse training data, such as the biographies of lesser-known historical figures.
Heresiologist: An author who wrote to identify, describe, and refute movements deemed heretical. Heresiological texts (e.g., those of Epiphanius of Salamis, Irenaeus of Lyon) are often the only surviving sources for Gnostic and Montanist communities, in which women’s leadership was prominent — but these sources are polemical and must be read with care.
Inter-annotator Agreement: A measure of how consistently multiple human raters apply the same criteria to the same items. High inter-annotator agreement (measured via Krippendorff’s alpha or Cohen’s kappa) is a prerequisite for treating human evaluation scores as reliable.
Krippendorff’s Alpha: A statistical measure of inter-annotator agreement that can be applied to any number of annotators and any measurement level (nominal, ordinal, interval). It is the preferred metric for this project because the rubric uses ordinal Likert scales with three or more annotators.
Montanism: A second-century Christian movement founded by Montanus in Phrygia (modern Turkey), notable for its emphasis on ecstatic prophecy and for the prominent roles of its female prophetesses Maximilla and Priscilla. Condemned as heretical by proto-orthodox authorities, but significant for the study of women’s religious authority in early Christianity.
Patristic: Relating to the Church Fathers (patres), the influential early Christian writers of roughly the second through eighth centuries CE. “Patristics” is the scholarly study of this literature. The gendered etymology of the term is itself a marker of the androcentric tradition the project interrogates.
Pointwise Scoring: An evaluation paradigm in which a single LLM response is rated on a numerical scale (e.g., 1–5) along one or more dimensions. Contrasted with pairwise comparison, in which two responses are compared directly.
Prostatis: Greek term used in Romans 16:2 to describe Phoebe. Often translated as “patron” or “helper,” but the term in its civic usage denoted a protector or leader with formal authority. The translation choice carries significant implications for our understanding of Phoebe’s role.
Reference-Guided Scoring: An evaluation paradigm in which LLM responses are assessed against reference material prepared in advance, rather than against each annotator’s independent recall. In this project the reference is a researched profile and fact-check sheet per figure, used for the Historical Accuracy and Epistemic Quality dimensions.