Three years ago, a small group of academics at a German university launched an unprecedented collaboration with the military – using novels to try to pinpoint the world’s next conflicts. Are they on to something?
As the car with the blacked-out windows came to a halt in a sidestreet near Tübingen’s botanical gardens, keen-eyed passersby may have noticed something unusual about its numberplate. In Germany, the first few letters usually denote the municipality where a vehicle is registered. The letter Y, however, is reserved for members of the armed forces.
Military men are a rare, not to say unwelcome, sight in Tübingen. A picturesque 15th-century university town that brought forth great German minds including the philosopher Hegel and the poet Friedrich Hölderlin, it is also a modern stronghold of the German Green party, thanks to its left-leaning academic population. In 2018, there was growing resistance on campus against plans to establish Europe’s leading artificial intelligence research hub in the surrounding area: the involvement of arms manufacturers in Tübingen’s “cyber valley”, argued students who occupied a lecture hall that year, brought shame to the university’s intellectual tradition.
Yet the two high-ranking officials in field-grey Bundeswehr uniforms who stepped out of the Y-plated vehicle on 1 February 2018 had travelled into hostile territory to shake hands on a collaboration with academia, the like of which the world had never seen before.
The name of the initiative was Project Cassandra: for the next two years, university researchers would use their expertise to help the German defence ministry predict the future.
The academics weren’t AI specialists, or scientists, or political analysts. Instead, the people the colonels had sought out in a stuffy top-floor room were a small team of literary scholars led by Jürgen Wertheimer, a professor of comparative literature with wild curls and a penchant for black roll-necks.
After the officers had left, the atmosphere among Wertheimer’s team remained tense. A greeting gift of camouflage-patterned running tops and military green nail varnish had helped break the ice, but there was outstanding cause for concern. “We’d been unsure about whether to go public over the project,” recalls Isabelle Holz, Wertheimer’s assistant. The university had declined the opportunity to be formally involved with the defence ministry, which is why the initiative was run through the Global Ethic Institute, a faculty-independent institution set up by the late dissident Catholic, Hans Küng. “We thought our offices might get paint-bombed or something.”
They needn’t have worried. “Cassandra reaches for her Walther PPK” ran the headline in the local press after the project was announced, a sarcastic reference to James Bond’s weapon of choice. The idea that literature could be used by the defence ministry to identify civil wars and humanitarian disasters ahead of time, wrote the Neckar-Chronik newspaper, was as charming as it was hopelessly naive. “You have to ask yourself why the military is financing something that is going to be of no value whatsoever.”
In the end, the launch of Project Cassandra saw neither paint bombs nor sit-ins. The public, Holz says, “simply didn’t take us seriously. They just thought we were mad.”
Charges of insanity, Wertheimer says, have forever been the curse of prophets and seers. Cassandra, the Trojan priestess of Greek myth, had a gift of foresight that allowed her to predict the Greek warriors hiding inside the Trojan horse, the death of Mycenaean king Agamemnon at the hands of his wife and her lover, the 10-year wanderings of Odysseus, and her own demise. Yet each of her warnings was ignored: “She’s lost her wits,” says Clytaemestra in Aeschylus’ play Agamemnon, before the chorus dismiss her visions as “goaded by gods, by spirits vainly driven, frantic and out of tune”.
The idea that novelists are modern-day Cassandras – “speaking always truths, never grasped as true” – may sound positively esoteric. There are tons of listicles on the internet hailing books that predicted events before their time, but most of these are accidental acts of clairvoyance by sci-fi writers describing futuristic technological equipment.
In 1914’s The World Set Free, HG Wells wrote about atomic bombs whose radioactive elements contaminate battlefields – three decades before Hiroshima and Nagasaki. British author John Brunner’s Stand On Zanzibarfrom 1968imagined Europe’s states forming a collective union, China’s rise as a global power, the economic decline of Detroit and the inauguration of a “President Obomi”.
And, naturally, there is George Orwell’s Nineteen Eighty-Four, in which a one-party state uses “telescreens” to identify people from their expressions and heart rate – written more than half a century before the NSA’s Prism surveillance programme and China using facial recognition software to track its citizens.
But Wertheimer says great writers have a “sensory talent”. Literature, he reasons, has a tendency to channel social trends, moods and especially conflicts that politicians prefer to remain undiscussed until they break out into the open.
“Writers represent reality in such a way that their readers can instantly visualise a world and recognise themselves inside it. They operate on a plane that is both objective and subjective, creating inventories of the emotional interiors of individual lives throughout history.”
We were interested in what hit a nerve. Was a book heaped with awards? Or banned and the author had to leave the country?
His favourite example of literature’s ability to identify a social mood and cast it into the future is a retelling of the Cassandra myth by the East German novelist Christa Wolf. Kassandra, published in 1983, casts Troy as a state not unlike the late-stage German Democratic Republic, succumbing to the paranoia of a Stasi-like secret police as it veers towards a not-so-cold war. Kassandra, cursed with the gift of prophecy, is also a cipher for the author’s own predicament: she foresees the decline her society is heading for, but her warnings are ignored by the military patriarchy.
If states could learn to read novels as a kind of literary seismograph, Wertheimer argues, they could perhaps identify which conflicts are on the verge of exploding into violence, and intervene to save maybe millions of lives.
At first glance, Jürgen Wertheimer looks like the least ideal candidate to drag literature out of the ivory tower. Born in Munich in 1947, to a Jewish father who had escaped from Dachau concentration camp and spent the rest of the war in hiding, he started university at the height of the 1968 student movement that energised a generation of young postwar Germans, but steered clear of the political activism that came with it.
Wertheimer, who says he is scared of crowds, has never joined a march in his life. Instead, he has spent time in libraries to work on his dissertation on Stefan George, the poet of the German wing of the l’art pour l’art movement: the idea that art must serve no function other than that of being art. Apart from a doomed love affair with Munich’s less successful football team, 1860 Munich, literature remains both his vocation and sole hobby, the 74-year-old says with a husky Bavarian drawl so melancholy, it suggests that passions can also become burdens.
Colleagues say Wertheimer’s air of apolitical world-weariness belies an agile, pragmatic mind. “I think he simply didn’t take part in 1968 because everyone else was,” Holz says. “He’s always been a bit of an outsider who liked doing things his own way.” At the end of the 2000s, when students across Germany rose in protest against plans to introduce university fees and a more efficient, UK-style bachelor’s-master’s system, Wertheimer found himself on the barricades, speaking up in support of the protesters.
As more and more university funding was diverted to Tübingen’s “cyber valley” rather than the arts and humanities departments, and literary study was increasingly under pressure to prove it could be socially useful, Wertheimer discovered his activist zeal.
On 15 December 2014, he posted a letter to Ursula von der Leyen – now president of the European Commission, then Germany’s defence minister. He drew her attention to the fight against Boko Haram in northern Nigeria – a terrorist group whose name is often translated as “western education is forbidden” and which has targeted schools and burned down libraries. Clashes of arms, he wrote, were usually preceded by wars of – and sometimes on – words, and therefore words could also be used to prevent them. He would relish the opportunity to put the theories he had developed in this field “into practical use in the framework of the German military’s foreign deployments”.
Von der Leyen never personally responded to Wertheimer’s offer. But in the spring of 2015 he received a letter from the defence ministry’s politics directorate, inviting him for a meeting. Two years of talks and meetings later, it commissioned the Tübingen professor to run a pilot project to test his idea of literature as an “early warning system”. Cassandra’s dry run was to demonstrate how the war in Kosovo and the rise of Boko Haram could have been predicted through the study of literary texts.
In the summer of 2017, Wertheimer hired a small team of researchers, including Florian Rogge, an unassuming master’s student working on anti-democratic satire in the Weimar Republic, and Holz, who was writing her PhD on the literary inspirations of the Baader-Meinhof group. A room where the team would convene each week was fitted with giant military maps and shelves of literature from Nigeria, Algeria and Kosovo. “We felt we had been handed a unique opportunity,” Holz recalls, “to show literature was capable of more.”
Predicting unpredictable crises has become the west’s great obsession in the 21st century. It stumbled into the millennium on the back of two genocides, in Bosnia and Rwanda, that were later widely thought to have been preventable if the international community had responded with more resolute shows of force. It then rushed into a war in Afghanistan, and launched an ill-conceived intervention in Iraq, often riding roughshod over the advice of experts who had studied political division and sectarian conflicts on the ground.
The terrorist attack on New York’s twin towers that triggered the rush to Afghanistan is an example of what Nassim Nicholas Taleb described in his 2007 book The Black Swan. “Black swan events” are rare, have an extreme impact and seem predictable – if only in retrospect. In a world becoming ever more interconnected, Taleb argued, such events were increasing exponentially.
The book would become a bestseller, not least because the unpredictable crises kept coming: the Lehman Brothers collapse in 2008, the European debt crisis in 2009, the refugee crisis of 2015, Brexit and Trump in 2016, a global pandemic in 2020. Even when Covid-19 has been hemmed in, it seems inevitable that another crisis will follow. The only thing governments can do is try to anticipate which direction it will come from, and shovel resources in that direction in advance. “With the wisdom of hindsight” was the go-to excuse of the Bush and Blair era; the next generation of global leaders want to attain the gift of foresight.
To an extent, this is nothing new: giving warning of impending crises is what foreign intelligence agencies are meant to do. What is new is that information that was once accessible only to spies is now freely available online and can be openly harvested by other arms of state. Big data is where most governments of today hope to find the tea leaves that allow them to read the future. America and the Scandinavian countries lead the field in conflict prediction through machine learning. In Britain, the Alan Turing Institute is trying to harness AI to give policymakers advance warnings through its Global Urban Analytics for Resilient Defence – or Guard – project.
Germany, which had contributed to the Afghan war in only a modest capacity and stayed out of Iraq, has invested some £43m into figuring out if it can use data tools to predict international conflicts. It has set aside a further £2.6bn to expand the approach until 2025. The centrepiece of its attempt at geopolitical clairvoyance is a megadata management platform developed at Munich’s University of the German Federal Armed Forces. It is called Preview – a forced acronym for “Prediction, Visualisation, Early Warning” – overseen by Carlo Masala, a professor of international politics. Preview sucks up information that could give a hint about where on the planet a crisis is about to erupt: RSS feeds of news websites, data banks tracking military conflicts, civil protests or car bombs going off. Broader structural clues are thrown into the mix: GDP per capita, regional educational structures, climate change data.
All this raw information is fed into Watson, IBM’s artificial intelligence platform, which helps convert it into maps highlighting potential trouble spots: green indicates stability, orange highlights instability, red warns of a conflict on the verge of escalation. One German official says the AI prediction system had already given Angela Merkel’s government a few months’ warning of the rebel insurgency in Mozambique’s northern Cabo Delgado province, where security forces are battling with militants trying to set up an Islamic state. But the early warning system is still in development: the aim is to eventually be able to predict conflicts 12-18 months in advance.
Germany remains more wary than other nations of outsourcing strategic assessments to algorithms: “There are some analysts in the US who believe that AI can eventually replace human prognostics altogether – we don’t believe that,” says Masala, a lively and communicative German-Italianwho enjoys poking fun at his country’s earnest moralising on matters military on social media (his Twitter profile is a picture of Tyrion Lannister, Game Of Thrones’s permadrunk military strategist). “The algorithm may be able to do 84% of the work, but we see it as a support system for human analysts, not a replacement.”
Watson, the question-solving supercomputer that sits at the heart of Preview’s crisis prediction effort, takes a second to process 500 gigabytes, the equivalent of a million books. But Watson isn’t great at reading between the lines. A literature professor seemed perfectly placed to complement the computer’s linear skill set.
Even with his team of three researchers, the challenge Wertheimer had been set seemed overwhelming. “At the beginning of the project, we thought we’d spend most of our time sitting in libraries reading books,” recalls research assistant Florian Rogge.“But we realised quite quickly that there were simply too many books, written in languages we didn’t speak.”
The team considered text mining: scanning books for emotive words and phrases to build a cluster map of feelings associated with specific issues, geographical regions or political figures. They consulted the IT department. The problem was that text mining requires books to be digitised, with the researchers’ chosen search terms defined in advance. “It would have bulldozed over irony, ambivalence or metaphors, all the literary aspects we were interested in,” Rogge says. “We wanted the books to tell us something we didn’t already know.”
The group decided instead to focus on what it calls “literary infrastructure”: what happens around the text? How is it being received? “We became interested in what hit a nerve,” Rogge says. “Was a book heaped with awards and state prizes? Or was it banned and the author had to leave the country?” Kuwait, for example, saw a rise of novels about the situation of the stateless Bidoon minority after 2010. Many of them were censored or banned shortly after their publication, prefiguring the crackdown on Bidoon protesters in 2019.
Reading books in translation proved an inefficient way to pick up such trends. Rogge says he ended up skim-reading no more than 30 novels over the course of the project. Instead, Wertheimer’s team reached out to writers and literary critics in regions they were interested in. The response was surprisingly enthusiastic. The novelist Wole Soyinka sent links to articles in the Nigerian press and supplied contacts to other writers. Kosovar writer Beqë Cufaj organised a colloquium at his country’s embassy in Berlin. Hearings in Paris and Madrid were attended by novelists from Algeria, Morocco, Egypt, Israel and France, most of whom volunteered to pay for their attendance out of their own pockets.
Our systems could already predict a conflict a year in advance. But five years – that was something new
In 2018, weeks after the Bundeswehr officers had travelled to Tübingen, Wertheimer presented his initial findings at the defence ministry in Berlin. He drew attention to a literary scandal around Jovan Radulović’s 1983 play Dove Hole, about an Ustashe massacre against their Serbian neighbours, and the expulsion of non-Serbian writers from the Serbian Writers’ Association in 1986. In the years that followed, he showed, there was an absence of tales about Albanian-Serbian friendships or love stories, and a rise in revisionist historical novels. Literature and literary institutions, he told the military men, had “paved the way for war” a good decade before the start of the bloodshed of the Kosovo war in 1998.
Carlo Masala was at the presentation. “At the beginning, I thought: this is crazy shit,” he recalls. “It won’t fly.” But Masala, who had spent a part of his academic career studying the conflict in Bosnia, remembered how the hardening tensions in the regions had been preceded by a decline in interfaith marriages. “In Kosovo, it seemed, you could detect similar early warning signs in the literary scene.”
“It was a small project that created a surprising amount of useful results,” says one defence ministry official who attended the presentation. “Against our initial instincts, we were excited.”
In its bid for further government funding, Wertheimer’s team was up against Berlin’s Fraunhofer Institute, Europe’s largest organisation for applied research and development services, which had been asked to run the same pilot project with a data-led approach. Cassandra was simply better, says the defence ministry official, who asked to remain anonymous.
“Predicting a conflict a year, or a year and a half in advance, that’s something our systems were already capable of. Cassandra promised to register disturbances five to seven years in advance – that was something new.”
The German defence ministry decided to extend Project Cassandra’s funding by two years. It wanted Wertheimer’s team to develop a method for converting literary insights into hard facts that could be used by military strategists or operatives: “emotional maps” of crisis regions, especially in Africa and the Middle East, that measured “the rise of violent language in chronological order”.
hey had the books; but how could they feed their findings into a machine?The challenge to bridge the intellectual divide between science and the humanities in the space of 12 months fell to Julian Schlicht, who was brought in to Project Cassandra in September 2019 to work out how literary criticism could be converted into data points. A 30-year-old student of politics, sociology and Islam studies who was researching the Taliban’s innovation strategies for his MA, he was the only member of the team without a background in literary studies.
“I was probably among the sceptics when I joined,” Schlicht says in hindsight. “Coming into this project with a background in politics, I thought their approach was… quite cheeky. There was a moment after the first meeting when I thought: how is this meant to work?”
Wertheimer’s researchers had become increasingly interested in Algeria, a country that had remained mostly silent during the Arab spring. Only 51.7% of voters in the north African country had turned out to vote in its 2014 presidential election, suggesting political apathy or weariness, borne of violent memories of the civil war that shook the region in the 1990s: Algeria was broadly classified as a “stable” state.
Trends in Algerian book publishing, however, hinted that something was about to change beneath the surface. Amar Mezdad’s 2014 novel Yiwen Wass Deg Tefsut (“A Spring Day”) follows a group of people who join a demonstration that is violently dispersed. Saïd Sadi’s 1991 diary novel Algérie, L’Échec Recommence was reissued in the northern Algerian Kabylia region in 2015, revisiting the Berber spring of the 1980s. And there was Boualem Sansal’s 2015 novel 2084: La Fin Du Monde, an Orwell-referencing dystopia in which an Islamist dictator uses religion to control the language and minds of his people. A former high-ranking government official who has criticised the rise of political Islam in Algeria, Sansal’s books had been banned in his home country since 2006, but were still widely read – making this a prime example of literature’s ability to touch a nerve.
But turning these atmospheric signals into information that could be of use to military or political decision-makers proved a challenge. “We realised there was a clash between how someone from the humanities and a scientist would go about drawing up a map,” Schlicht recalls. “A cultural historian will say: my expertise tells me this region is red and that region yellow. A scientist, however, asks: how do we work out when the yellow region turns orange?”
The researchers developed a risk score system with nine indicators for each book: thematic reach, censorship of the text, censorship of the author, media response, scandals around the text, scandals around the author, literary awards for the author, literary awards for the the text, and narrative strategy. In each category, the book was assigned a score between –1 and +3: the higher the score, the more “dangerous” the text.
In some cases, negative scores were necessary. A book told from several perspectives, such as from two opposing enemy camps, was assigned zero or –1 points in the narrative strategy category. Zoltán Danyi’s 2015 novel The Carcass Remover, for example, scored only 12 points, mainly because it reflected on the Yugoslav wars without using black-and-white depictions of heroes or villains. “We realised that literature can also help solve or lessen conflicts,” Schlicht says. “Not every book divides opinion.”
Dystopian fiction from Algeria scored much higher. Body Writingby Mustapha Benfodil, a 2018 collage novel made up of the diaries and scribbles of a fictional astrophysicist killed in a mysterious car accident on the day of the presidential election, spoke of the desire to create order out of the chaotic memories of the Algerian civil war in the 90s, expressing a yearning for democratic change. Twenty points, the Project Cassandra researchers decided.
Had the book had more impact, it would have scored higher: La Faille by Mohamed-Chérif Lachichi, a 2018 thriller portraying violence in Algerian prisons, a corrupt legal system and a growing protest movement, scored 22 points because it represented a case of a well-known and widely reviewed author questioning the status quo. The highest-scoring work in the project’s 300-book data bank was Sansal’s 2084: Wertheimer’s team assigned it 25 points.
Project Cassandra’s scoring system had several flaws. Its indicators were malleable: should rave reviews or hatchet jobs contribute to a book’s score only when they were published in its country of origin? Was a book with 24 points really twice as dangerous as a book with 12 points? Converting scores into emotional heat maps came with additional headaches: what does a dystopian Algerian novel set in a parallel universe really tell you about the different political allegiances of Algiers and Oran? “Translating literature into numbers is difficult; you constantly end up making compromises,” Rogge says. “You need to score a lot of books in order to even out the imprecisions.”
When Azerbaijan gave anti-Armenian books to Georgian libraries, the project predicted conflict. A year later, war broke out
But the literary seismograph’s instincts proved reliable. In February 2019, two years after Wertheimer’s team had identified Algeria as a region of interest, civil protests broke out in Algiers and several other cities, culminating in the resignation of President Abdelaziz Bouteflika. When Wertheimer handed over his team’s findings in the summer of 2020, they met the defence ministry’s formal requirements for taking their methodology to the next level. Project Cassandra had established a tangible link between literature and empirical historical events. Plus the cancellation of trips and seminars because of the pandemic meant the study had come in almost £34,000 under budget.
A meeting with Operations Command was scheduled. “We all had our hopes up,” Holz remembers. “We couldn’t wait to apply our findings in the field.”
Then Wertheimer received a phone call. Project Cassandra would be discontinued in winter 2020.
In the fallout, some blamed the pandemic, which had placed unforeseen strain on government budgets. Others pointed the finger at the five-year reshuffle cycle for government posts, which meant one of the most supportive officers had been dispatched to Moscow and the new guard knew little about Cassandra’s original promise. Via official channels, the defence ministry is tight-lipped: “Retrospectively we like to emphasise how much we appreciated this innovative approach,” is all a spokesperson could say on the matter.
“Of course it was a massive blow, even more so because we hadn’t expected it,” Holz says. Behind the scenes, some officials remain baffled. Crisis prediction through literature, one says, would have been a “self-fulfilling success story”. It came with a free network of enthusiastic, expert volunteers, who could not only help identify conflicts several years in advance but provide diplomatic corps with the right counter-narratives to quell it. Perhaps, another official pondered, Project Cassandra was simply too good to be true, and would have risked showing up the more costly undertaking of data-driven programmes.
In the Greek myth, Cassandra’s warnings go unheeded because the Trojan priestess has been cursed by the god Apollo, angered after being turned down for sex. In Christa Wolf’s modern adaptation, the Trojan generals know she is speaking the truth, but ignore her regardless. “King Priam prefers to remain ignorant out of political calculation,” Wertheimer says. “I used to believe modern politicians were different, that they simply didn’t know better. It turns out they are much like their ancient counterparts: they prefer not to know.”
In one of his last reports to the defence ministry, towards the end of 2019, Wertheimer had drawn attention to an interesting development in the Caucasus. The culture ministry of Azerbaijan had recently supplied libraries in Georgia with books carrying explicit anti-Armenian messages, such as the works of poet Khalil Rza Uluturk. There were signs, he warned, that Azerbaijan was ramping up propaganda efforts in the brewing territorial conflict with its neighbouring former Soviet republic.
War broke out a year later: 6,000 soldiers and civilians died in a six-week battle over Nagorno-Karabakh, an enclave of Azerbaijan populated by ethnic Armenians. Turkish president Recep Tayyip Erdoğan used the war to bolster his strongman image, hailing Armenia’s defeat in December as a “glorious victory”. Russia, traditionally allied with Armenia, successfully leveraged the conflict to consolidate its influence in the region. Germany and the EU, meanwhile, looked on and stayed silent: being able to predict the future is one thing, knowing what to do with the information is another.
Wertheimer isn’t bitter about the end of Project Cassandra, he says: “I’m not a bitter person, only melancholy.” At any rate, the project may still have a future after all. The German interior ministry has commissioned his team to investigate the hidden scars of the country’s reunification process. There have been talks with the EU representative for foreign affairs and security policy, Josep Borrell, about docking Cassandra in Brussels. Wertheimer says he is interested in applying his method to analyse geopolitical tensions in Ukraine, Lithuania and Belarus.
Above his writing desk in Tübingen hangs a map of the world with travel visas pinned to the white spaces next to the continents. “Literature is at the centre of my life,” he says, “but only because I believe it can be a springboard into the real world.”