When asked to demonstrate the impact of their research, researchers build up statements that have a causal structure. However, as these statements have by nature a historical dimension, their credibility is under question. Historical statements have a genuine causal power only under certain conditions. We derive these conditions from the theory of historical causality and apply them to impact statements in two Medicine and Engineering (STEM) and Social Sciences and Humanities (SSH) areas of Research Excellence Framework. We then process the corpus with a novel text mining methodology called Semantic Hypergraphs. We identify the causal structure of statements and find that it is similar between STEM and SSH, but SSH makes systematically use of a larger number of actors. Making credible statements are more difficult in SSH than in STEM. We derive the policy implications for impact assessment and research policies.

中文翻译：

---

研究影响陈述的可信度：REF与语义超图的新分析

当被要求证明其研究的影响力时，研究人员会建立具有因果关系的陈述。但是，由于这些声明从本质上讲具有历史意义，因此其可信度受到质疑。历史陈述只有在某些条件下才具有真正的因果关系。我们从历史因果关系理论中得出这些条件，并将其应用于“研究卓越框架”的两个医学与工程（STEM）和社会科学与人文（SSH）领域的影响陈述。然后，我们使用一种称为语义超图的新型文本挖掘方法来处理语料库。我们确定了陈述的因果结构，发现在STEM和SSH之间它是相似的，但是SSH系统地使用了大量参与者。与STEM中相比，在SSH中做出可信的声明更加困难。