Search
Browse By Day
Browse By Time
Browse By Person
Browse By Mini-Conference
Browse By Division
Browse By Session or Event Type
Browse Sessions by Fields of Interest
Browse Papers by Fields of Interest
Search Tips
Virtual Exhibit Hall
Change Preferences / Time Zone
Sign In
X (Twitter)
Session Submission Type: Virtual Roundtable
An empirical datum of a ‘story’ (or a historical event) is an infinite set of facts, it is an information matrix of infinite size. Stories are often labeled as narratives. Here we separate the concept of a narrative, with a provisional definition of a narrative as a story told through the prism of causation.
Riker invited us to define the boundaries of what one is trying to explain.
The boundaries of the situation are the boundaries of a ‘narrative.’ They are not just temporal, but also substantive, and exclude ‘non-relevant’ persons and facts. The biases in choosing what to focus on imply the multiplicity of narratives, subsets of the initial set of facts. Bias in narratives create an advantage to test out some aspects of the original event.
Whether one should create a narrative as simply as necessary, make narratives as thin as possible depends on the objective. Hereby lies a difference between analytical narratives and process tracing approaches. In an analytic narrative the model answers the question “Is the conjecture as applied to the case truly logical, and can the case possibly be explained by this theory?” In the process-tracing framework, an applied model answers the question “Which is the best-fit theory to explain the outcomes in a particular case?” Both approaches are based on a single datum stylized into a narrative, but have different goals.
‘Thin’ narratives are more likely to yield generalizable models. They are more likely to yield statistically testable conjectures.
Unique events require non-statistical approach.
Should the narrative be repeated or revised after the model is completed? The ex-post narrative must not be ‘thinner than the model and the original narrative – conclusion should not claim greater generality than is achieved in the model.