Top image: Part of Dead Sea Scroll 28a. 

For more than seven decades, the Dead Sea Scrolls have offered a window into the spiritual and political life of ancient Judaism, as well as early Christianity. Yet despite their immense importance, the dating of individual scrolls has remained an inexact science. Now, a groundbreaking study by an international team of scholars led by the University of Groningen has deployed artificial intelligence and radiocarbon dating to bring new precision to this ancient puzzle.

According to the University of Groningen report, the newly developed AI model, called Enoch, has enabled researchers to re-date many Dead Sea Scrolls, some of which now appear significantly older than previously thought. This technological leap not only provides firmer chronological ground for hundreds of manuscripts but also marks the first time that two biblical scroll fragments can be confidently linked to the era in which their authors are thought to have lived.

A Long-Standing Chronological Mystery

Since their discovery in the caves of Qumran between 1947 and 1956, the Dead Sea Scrolls have fundamentally reshaped modern understanding of the Bible and ancient Jewish practices. They include some of the oldest known copies of biblical texts, apocryphal writings, legal codes, and sectarian documents. Dating these manuscripts accurately is essential to understanding their context, authorship, and influence.

Traditionally, dating has relied heavily on paleography, the analysis of handwriting styles. However, this method is inherently subjective and lacks a firm empirical basis, especially because there is a lack of securely dated reference manuscripts from the period. Between the fifth century BC and the first century AD, a crucial paleographic gap has prevented scholars from establishing precise dates for most of the over 1,000 scrolls.

Fragment of the Damascus Document Scroll

Closing the Gap with AI and Radiocarbon Dating

This challenge has now been met with the innovative Enoch model, developed as part of the ERC project The Hands that Wrote the Bible. Enoch combines carbon-14 dating from 24 Dead Sea Scroll samples with AI-driven handwriting analysis to bridge the chronological gap. This approach is based on a previously developed deep neural network, BiNet, which detects micro-level ink-trace patterns in digitized manuscripts.

Using Bayesian ridge regression, the model correlates specific handwriting features with known radiocarbon dates, effectively "learning" how script styles evolved over time. The result is an AI tool that can predict the date of a manuscript with an uncertainty of only ±30 years, often more precise than carbon dating alone in the 300-50 BC range.

For this study, writing styles in digitized manuscripts were analyzed using BiNet, a previously developed deep neural network for detection of handwritten ink-trace patterns.

Revolutionizing Manuscript Dating

Enoch was tested on 135 scroll fragments. Paleographers then assessed the date predictions, which revealed new insights into the chronology of manuscript production. For instance, many scrolls once believed to date from the late Hasmonaean period (c. 150-50 BC) are now considered to be older. Similarly, the Herodian script, long thought to emerge in the mid-first century BC, appears to have been in use as early as the late second century BC.

This overlapping period of script usage alters the previously accepted narrative of script evolution in ancient Judea. It suggests a more complex landscape of literacy and scribal practice during a politically volatile period, including the Hasmonaean uprising and the rise of Roman influence.

Manuscripts Dated to Their Authors' Era

Among the study's most significant revelations is the precise dating of two biblical manuscripts to the time periods traditionally assigned to their authors. Fragment 4QDanielc (4Q114) aligns with the early 160s BC, the presumed era of the final redaction of the Book of Daniel. Likewise, 4QQoheleta (4Q109) has been dated to the third century BC, matching the scholarly estimate for the composition of Ecclesiastes (Qohelet).

While the traditional attribution of Ecclesiastes to King Solomon dates it to the tenth century BC, most scholars now accept a later Hellenistic origin. The Enoch model thus offers the first empirical confirmation of these academic theories and opens new avenues for exploring the development of biblical literature.

Maruf Dhali and Mladen Popovi?.

A Tool for Future Research

Professor Mladen Popović, director of the Qumran Institute, and Dr. Maruf Dhali, assistant professor of AI at the University of Groningen, spearheaded the research. Their interdisciplinary approach merges the physical sciences with digital analysis to bring a level of objectivity previously unattainable in manuscript studies.

"Enoch is the first complete machine-learning-based model that uses raw image inputs to produce probabilistic date predictions for ancient manuscripts," says Dr. Dhali. Importantly, the model is also transparent and explainable, making it a valuable tool for other manuscript traditions beyond the Dead Sea Scrolls.

With the ability to refine, support, or challenge paleographic assessments, Enoch is poised to become a cornerstone of ancient textual scholarship. Its implications go beyond dating: by anchoring texts more precisely in time, researchers can better understand the sociopolitical and religious milieu that shaped them.