Researchers from universities in Israel and around the world used AI-based statistical models to identify three distinct writing styles in the first nine books of the Bible – and assign controversial chapter authors to these groups.
Artificial intelligence is transforming many industries—from medicine to film. Now, researchers are using it to analyze one of the oldest and most revered texts: the Bible.
An international research team, led by Dr. Shira Feigenbaum-Golovin, a research professor of mathematics at Duke University, combined tools from the field of artificial intelligence, statistical models, and linguistic analysis to tackle a central question in biblical studies: who wrote the Bible?
The study was published in the journal PLOS One.
By analyzing subtle differences in word usage, the team was able to identify three distinct writing styles (or scribal traditions) throughout the first nine books of the Bible, known as the Enneateuch.
The same model was later used to attempt to associate additional chapters with putative authors, while the model also presented the linguistic explanations for its conclusions.
That's how it all started.
In 2010, Feigenbaum-Golovin began collaborating with Prof. Israel Finkelstein, Professor Emeritus at Tel Aviv University and Head of the School of Archaeology and Maritime Civilizations at the University of Haifa, in an attempt to associate inscriptions on pottery fragments dating back some 2,600 years to specific authors using morphological-statistical analysis.
Their findings were first reported in the New York Times.
According to her, "We were able to show that these inscriptions can serve as clues to the dating of biblical writings, and this led us to establish the current research team."
Multidisciplinary collaboration
The research was conducted in two stages. The team included scientists from various fields: Alon Kipnis (Reichmann University), Axel Biller (Protestant Faculty in Paris), Eli Piasecki (Tel Aviv University), Thomas Romer (Collège de France) and Feigenbaum-Golovin herself.
They analyzed language patterns in three major sections of the Bible: Deuteronomy, Deuteronomistic History (Joshua through Kings), and the priestly writings in the Torah.
The results supported an existing research consensus: Deuteronomy and the books of history are closer to each other than to the priestly writings.
Prof. Romer explained: "We found that each group has a different style – even with simple words like 'no', 'asher' or 'king'. Our method accurately identifies these differences."
Model testing
To test the model's effectiveness, 50 chapters previously assigned to known writing groups were analyzed, and the model was able to correctly assign them using a quantitative formula.
Next, chapters whose assignment was disputed were examined, and here too the model was able to assign them to the most likely group – and even explain the choice based on the frequency of words or roots.
Alon Kipnis emphasized: "One of the main advantages of the method is the ability to explain the result – to point out the words that led to the association of a particular chapter."
Language barriers and unique techniques
The main challenge was to find biblical texts that had not undergone extensive editing and that retained their original form. Such texts are often short, making standard computer-based learning methods inapplicable.
Therefore, instead of models that require huge amounts of data, the researchers used a direct, comparative approach that examined sentence patterns and word frequency.
A surprising find
One notable example was the story of the Ark of the Covenant in the books of Samuel. Although sometimes considered part of the same plot, the model showed that 1 Samuel does not fit into either group, while 2 Samuel fits into the historical style.
Feigenbaum-Golovin concluded: "This method is also suitable for other historical documents – for example, you can check whether a document attributed to Abraham Lincoln is authentic or forged."
Finkelstein added: "This study presents a new paradigm for analyzing ancient texts."
The team is now examining the application of the method to the Judean Desert Scrolls. Feigenbaum-Golovin emphasized: "This is a unique collaboration between science and the humanities. A surprising symbiosis that expands boundaries."
The article is based on a statement from Duke University.
More of the topic in Hayadan:
