數字世界:看統計學怎麼預測恐怖襲擊

What is the risk of a war on the Korean peninsula or South China Sea? Or, for that matter, of another terrorist attack on American soil? These are questions that western diplomats and security experts are asking themselves this spring. And as speculation grows, those officials have been duly scouring satellite feeds, intelligence reports and history books.

朝鮮半島或南中國海爆發戰爭的風險有多大？美國本土再次遭遇恐怖襲擊的風險又有多大？這是今年春天西方外交人士和安全專家一直在思索的問題。各種猜測四起，這些官員也在理所當然地查找衛星資料，翻閱情報報告和歷史書籍。

Over in Colorado, Aaron Clauset, a computational scientist, is pondering the dangers from a different perspective. Clauset, who teaches at the University of Colorado, Boulder and is part of the Santa Fe Institute, has spent the past decade on the frontier of computing and statistical research. But he has not focused on areas normally beloved by geeks, such as engineering, physics or biology.

而在科羅拉多州，計算科學家亞倫·克勞塞特(Aaron Clauset)正從截然不同的角度考量上述風險。克勞塞特在科羅拉多大學博爾德分校(University of Colorado, Boulder)任教，同時還是聖菲研究所(Santa Fe Institute)研究員，他在過去十年內一直奮戰在計算和統計研究的前沿。但工程、物理或生物等理科怪才們青睞的領域卻不是他的研究重點。

Instead, Clauset and other statisticians, such as Ryan Woodard of the Swiss Federal Institute of Technology in Zurich, have analysed the past 200 years of military conflicts. And this has produced a thought-provoking conclusion: if you look at the global pattern of war and terrorism, human violence has moved in surprisingly stable cycles.

相反，克勞塞特和瑞士聯邦理工學院蘇黎世分校(Swiss Federal Institute of Technology in Zurich)的瑞恩·伍達德(Ryan Woodard)等統計學家分析了過去200年的軍事衝突。他們得出的結論發人深思：對戰爭和恐怖主義的全球格局研究表明，人類暴力活動的變化規律出奇地穩定。

Indeed, it is so stable that Clauset sees strong parallels between human conflict and earthquakes – at least in statistical terms. He and other researchers are now borrowing models developed from seismology and physics to forecast future patterns of violence. The aim of this “terror physics” (as some dub it) is not to predict exactly where and when a terrorist attack may occur – doing that is as hard as pinpointing the next quake. Instead, these statisticians are working out the likely rate of attacks and wars – to tell when one seems statistically overdue.

確實，規律是如此穩定，以至於克勞塞特發現了人類衝突和地震之間存在密切關係——至少在統計學上是這樣的。現在，他在與其他研究人員借鑑由地震學和物理學發展而來的模型，預測未來暴力活動發生的規律。“恐怖物理學”（有些人如此稱呼這門學問）的宗旨不是預測恐怖襲擊的具體時間地點——這與精確地預測下一場地震一樣困難，而是研究出襲擊和戰爭的可能概率，從統計學角度推斷出何時可能會發生此類事件。

“The frequency and severity of wars has been pretty constant for 200 years despite all the massive changes in geopolitics, technology and population,” Clauset explains. On average the world sees one new international war every two years and a new civil war about every 1.5 years. And while terrorist attacks typically occur in clusters, with a few “mega” attacks accounting for large numbers of deaths, there are clear statistical rhythms there too. So much so that Clauset and Woodard argue that seemingly “rare” events, such as 9/11, are not actually that extraordinary after all. As they write in a 2012 paper: “Patterns observed in the frequency of severe terrorist events suggests that some aspects of this phenomenon, and possibly of other complex social phenomena, are not nearly as contingent or unpredictable as is often assumed.”

克勞塞特解釋道：“儘管200年來地緣政治、科技和人口均發生了劇變，但戰爭的頻率和烈程度一直較爲穩定。”世界上平均每兩年爆發一次國際戰爭，約每1.5年爆發一次內戰。另外，儘管恐怖襲擊通常集中爆發，且少數幾次“超級”襲擊造成大量死亡，但其中也存在頗爲明顯的統計規律。於是，克勞塞特和伍達德提出，類似“9·11”恐怖襲擊等看似“罕見”的事件其實並不特別。正如他們在2012年的一篇論文中寫道：“觀察嚴重恐怖襲擊事件得出的規律是，這一現象乃至其他複雜社會現象的某些方面,遠不如人們通常以爲的那樣不確定和難以預料。”

I daresay that some people would consider this analysis to be ridiculous or offensive. After all, we tend to think that the 21st century is a time of great flux, when we are reshaping the world. However, “terror physics” can only predict the future if you think that humans are doomed always to behave in consistent ways, without the capacity for change or progress. That is not a popular idea among governments. Some academics might question it too: the psychologist Steven Pinker, for example, argues that human violence is steadily declining in the world today, at least when measured in terms of violence per capita, as opposed to gross military casualties.

我敢說，有人會覺得這一結論荒誕不經或是招人反感。畢竟，我們傾向於認爲21世紀是激變的時代，是我們改造世界的時代。然而，“恐怖物理學”能夠預測未來的前提卻是，人類的行爲方式註定是前後一致的，缺乏改變或進步的能力。這一觀點得不到各國政府的認同，也引來一些學者的質疑。例如，物理學家斯蒂文·平克(Steven Pinker)認爲，人類暴力活動如今呈現出穩步減少的趨勢，至少以人均暴力衡量是如此（但軍隊總傷亡並非如此）。

In any case, diplomats usually study conflicts in terms of idiosyncratic social and historical factors, not cold data points. Or as Clauset says: “The conflict studies community usually wants to look at the motives of terrorists or their tactics, not the bigger pattern ... it’s like asking a weather forecaster to worry about climate change.”

無論如何，外交人士往往用來分析衝突的依據是特殊的社會和歷史因素，而不是冷冰冰的數據點。或如克勞塞特所說：“衝突研究領域通常希望研究恐怖分子的動機或行動手法，而不是總體趨勢……這就好比天氣預報員爲氣候變化操心一樣。”

But while military experts might be ambivalent about the value of terror physics, Clauset and Woodard’s research is now causing a buzz in the statistical world. It is also attracting serious interest from insurance companies and bankers, who are keen to work out the risks of terrorist attacks. Clauset and his fellow number-crunchers are hoping that the wider policy community starts to pay more attention too.

不過，雖然軍事專家對恐怖物理學的價值褒貶不一，但克勞塞特和伍達德的研究已經在統計學界引起轟動。熱衷於研究恐怖主義襲擊風險的保險公司和銀行家也對恐怖物理學產生了濃厚興趣。克勞塞特和統計學同仁們希望政策羣體也能予以更多關注。

If the number-crunchers can persuade governments to recognise that there is a statistical rhythm to violence, their argument goes, countries might be able to mobilise resources in preparation. And if policy makers acknowledge these cycles, they might also start to reflect on a fundamental question: what exactly drives those outbreaks of war or terrorism? Can we always blame violence on idiosyncratic personalities (be that the North Korean leaders, Osama bin Laden or anyone else)? Or is there something about the human condition – or our interaction with the environment – which dooms us to terrorism and war with such regularity?

在他們看來，如果統計學家能說服政府認識到暴力的統計學規律，各國或許能夠動用資源進行鍼對性的防範。如果政策制定者承認暴力發生週期的存在，他們或許還能開始反思一個根本的問題：究竟是什麼因素導致戰爭或恐怖主義活動的爆發？我們能否一成不變地將暴力歸咎於怪異的人格（不論是朝鮮領導人、奧薩馬·本·拉登(Osama bin Laden)還是別人）？還是說，與人類狀況有關的因素——或是我們與環境的互動——讓我們不可避免地遭遇規律性的恐怖主義活動和戰爭？

These are, of course, big philosophical issues. I don’t expect that any government will rush to discuss them publicly soon – not when politicians are busy fighting a “war on terror”, with the unspoken assumption that it is possible for humans to eradicate the scourge. But if nothing else, Clauset’s numbers put the recent past in perspective (by historical standards the Boston attack, for example, looks pretty small). And they should make us think about the future too. Clauset reckons that the chance of seeing another war this century on the same scale as the second world war (with 60m deaths) is 41 per cent. Meanwhile, the chance of a 9/11-size event this decade is between 19 per cent and 46 per cent. This is, of course, still irritatingly vague; but as predictions go, it seems too large to entirely ignore. Least of all in a place such as Boston, London – or even Korea.

當然，這些都是重大的哲學問題。我不指望哪國政府在短期內會公開討論這些問題——政客們正忙着打“反恐戰爭”，他們的想法不言自明：人類是可以根除恐怖主義災難的。但至少，克勞塞特的統計研究全面地檢視了近現代歷史（比如按歷史標準，波士頓遭受的襲擊似乎並不嚴重）。而且，這些研究也應當促使我們思考未來。克勞塞特估計，本世紀爆發一場與二戰規模（死亡6000萬人）相當的戰爭的概率是41%，而目前十年裏發生嚴重性堪比“9·11”恐怖襲擊的事件的概率介於19%至46%之間。當然有些惱人的是，這些數字仍然過於模糊。但從預測的角度來說，上述概率已大到讓我們很難置若罔聞的地步，對於波士頓、倫敦乃至朝鮮半島這些危險地帶就更不容忽視了。