A probabilistic approach to uncertainty in pre-modern history

概率是人类为理解和管理不确定性而开发的最强大的概念工具。从事预测未来的科学家、经济学家和其他研究人员经常使用概率技术来管理他们在估计全球气温或人口等参数的未来值时面临的巨大不确定性。他们不是试图猜测他们感兴趣的变量的确切数值，而是倾向于根据可能的数值范围内的“概率分布”进行思考。实际上，他们利用现代计算的能力来对所有可能的未来进行建模，以了解不同结果发生的频率，而不是试图对单一的“最佳估计”情景进行建模。对不确定变量的概率估计于20世纪70年代率先出现在核安全领域，此后成为自然科学和社会科学的主力。几十年的经验完善了支撑它的理论和实施它的方法。概率模型并不是万无一失的，但它们利用了计算的力量，并提供了实用主义和严谨性的平衡，这使它们远远优于它们所取代的确定性的“最佳估计”模型。概率方法现在在经济学、气象学和放射肿瘤学等不同领域都很流行。然而，它从未被开发为历史分析的工具，尽管历史上的许多量化问题与在其他领域采用概率方法的问题在结构上有很深的相似之处(最重要的是：存在多个独立的不确定性来源)。这一奖学金旨在展示它们作为历史分析工具的潜力，并努力向其他前现代历史学家传播使用它们所需的知识和技能。潜在的回报是，我们对许多模糊的历史参数的了解质量有所提高。在某些情况下，这将涉及对在目前的证据基础下似乎无法量化的参数进行稳健的估计；在其他方面，它将涉及扰乱现有的、混淆了全部不确定性的假设。该奖学金将产生(I)对古代历史中三个长期存在的问题的研究，以说明该方法及其潜在的广泛应用，(Ii)与其他职业生涯早期的研究人员合作，传播和完善该方法的研讨会，(Iii)一次会议，以激发对我们过去用来思考不确定性的方法的更广泛反思，以及(Iv)一本书-展示古代历史中参数模型的最新技术和新方法。

Probability is the most powerful conceptual tool that humans have yet developed to understand and manage uncertainty. Scientists, economists and other researchers engaged in forecasting the future regularly use probabilistic techniques to manage the massive uncertainties they face in estimating future values of parameters such as global temperature or human population. Rather than trying to guess the exact value of the variable they are interested in, they tend to think in terms of 'probability distributions' over ranges of possible values. In effect, they draw on the power of modern computing to model all possible futures in order to see how frequently different outcomes occur, rather than trying to model a single 'best estimate' scenario.Probabilistic estimation of uncertain variables was pioneered in the field of nuclear safety in the 1970s and has since become a workhorse of the natural and social sciences. Decades of experience have refined both the theory that underpins it and the methods for implementing it. Probabilistic models are not infallible, but they capitalise on the power of computing and offer a balance of pragmatism and rigour that makes them far superior to the deterministic 'best estimate' models they displaced. The probabilistic approach is now de rigeur in fields as diverse as economics, meteorology and radiation oncology. Yet it has never been developed as a tool of historical analysis, although many quantitative problems in history share deep structural similarities with the problems where probabilistic approaches are deployed in other fields (most importantly: the presence of multiple independent sources of uncertainty). This Fellowship aims to demonstrate their potential as tools of historical analysis and to work towards disseminating the knowledge and skills needed to use them to other pre-modern historians. The potential reward is an improvement in the quality of our knowledge about many obscure historical parameters. In some cases this will involve generating robust estimates for parameters that have seemed unquantifiable given the current evidence base; in others it will involve unsettling established hypotheses that obfuscate the full extent of uncertainty.The Fellowship will produce (i) studies of three long-standing problems in ancient history, to illustrate the method and its potentially wide application, (ii) a workshop to disseminate and refine the methodology in collaboration with other early-career researchers, (iii) a conference to stimulate wider reflection on the methods we use to think about uncertainty in the past and (iv) a book show-casing the state of the art in, and new approaches to, parametric modellin in ancient history.