AI advances in archaeology

Recently, researchers from the University of Groningen in the Netherlands have used the latest in AI technology to study the Great Isaiah Scroll (known as 1QIsaa), an ancient manuscript dating to the 2nd century BC that is considered one of the Dead Sea Scrolls.

Artificial intelligence (AI) is currently revolutionising many areas of society, and archaeology is no exception. While humans are quite good at recognising patterns, it turns out computers are even better, able to discern small differences that we are unable to detect ourselves. Recently, researchers from the University of Groningen in the Netherlands have used the latest in AI technology to study the Great Isaiah Scroll (known as 1QIsaa), an ancient manuscript dating to the 2nd century BC that is considered one of the Dead Sea Scrolls. They wanted to see whether individual scribes could be identified.

The Dead Sea Scrolls are a series of ancient manuscripts found in caves near Qumran and at other Judaean Desert sites near the Dead Sea. Dating to between the 4th century BC and the 2nd century AD, they contain the oldest-known fragments of the Hebrew Bible (Old Testament), as well as previously unknown ancient Jewish texts. As such, these artefacts offer some of the only tangible evidence of religious scribes during this period. By analysing the texts that they wrote, it may be possible to identify certain aspects of scribal ‘culture’, answering questions such as: How many scribes typically wrote a manuscript? And if multiple scribes were used, did they write whole sections each or did they break up the text in a different way?

Image: Maruf A Dhali, University of Groningen/Rijksuniversiteit Groningen An 11×11 and a 12×12 Kohonen map of characters aleph and bet from the study. Each of the characters in the Kohonen maps are formed from multiple instances of similar characters (shown as the zoomed box with red lines). These maps are useful for chronological style-development analysis.

To start, small differences in handwriting need to be identified and confirmed as the work of particular scribes. This task is easier said than done, as it is difficult to discern whether any noticeable variation is due to a second writer or just normal differences within the handwriting of the same writer – but AI technology has advanced to the point that it can process large amounts of data, and produce quantitative analyses that are impossible for humans to perform. It is important to note, though, that human agency has not completely left this process. While AI can be used to provide statistical probabilities of certain identifications, it is still up to the human expert to decipher the likelihood of these different probabilities.

In this method, AI and pattern-recognition programmes process each individual letter in a text, analysing features such as the slant and curve of the character, as well as the way in which an individual writer produces each letter shape, such as when they slow down or speed up on a certain part of a letter. Each time a particular letter is used, the programme can compare it with the other times the same letter appears to create what are called Kohonen self-organising feature maps to be able to visualise these differences. These maps provide a centroid letter – one that is considered the most typical based on a computed average of all examples in the analysed text – and then maps how similar each occurrence is to this average. Statistical tests can then be used to assess whether a letter deviates from the expected random pattern of a particular writing style.

The team from Groningen used these types of analyses to assess the Great Isaiah Scroll. Measuring 7.34m in length, the scroll contains 54 columns of Hebrew text and there appears to be a break in the text between columns 27 and 28. There is also a change of sheet between these columns, with two pieces of parchment having been sewn together at this point. Although the entire scroll comprises 17 sheets, this is the only place where a sheet change coincides with a caesura in the text. The morphology of the Hebrew text also appears to be marginally different between these two halves. But, as mentioned above, does this change indicate the work of a different scribe, or just natural variation in the first scribe’s handwriting?

The results are now in, and the AI analysis did indeed suggest that two different scribes probably wrote the Scroll, with one writing the text in columns 1-27 and the other the text in columns 28-54. The success of this project opens up the ways in which other Dead Sea Scrolls – as well as ancient manuscripts in general – can be analysed.

As the researchers note: ‘In a way that was not possible before, our approach opens access to the tangible evidence of the hitherto almost completely inaccessible microlevel of the individual scribes of the Dead Sea Scrolls and the possibility to examine the different compositions copied by each of the scribes. The change of scribal hands in a literary manuscript or the identification of one and the same scribe in multiple manuscripts can be used as evidence to understand various forms of scribal collaboration that otherwise remain unknown to us. The number of literary manuscripts on which a scribe worked, either alone or with others, can serve as tangible evidence for understanding processes of textual and literary creation, circulation, and consumption.’