Communicating Causality

Knowledge of causal processes is vital for all aspects of our lives, from mundane consumer choices to high-impact socio-political decision making. Much of our causal knowledge is acquired not from individual experience, but from cultural transmission via language. While theorists have amassed a large body of philosophical and psychological understanding about causation, surprisingly little research has been devoted to a theoretical understanding of the processes that underlie communication of causal information and the role of the linguistic signal in this transmission. The main objective of this project is to apply methods and tools from experimental pragmatics to shed light on a wide range of puzzles about causal language and cognition. We introduce a new pragmatic framework, the "CommuniCause" approach, which makes new and empirically testable predictions and offers a unified explanation for a number of disparate phenomena, both old and new. CommuniCause brings together two important strands of research that have been isolated until now, in a way that will benefit both communities. It draws on established philosophical theorizing and recent computational models of individual causal cognition, but is distinguished by its focus on linguistic factors: the speaker's choice among the variety of linguistic expressions available, and the impact that this choice has on a recipient's pragmatic interpretation. CommuniCause promises advances to linguistic pragmatics as well, by confronting theories and models with a range of intricate puzzles that have not previously been treated from a linguistic perspective. The project explores the extent to which this pragmatic approach can explain key features of causal inference. One is the problem of causal selection, how we identify one event as the cause of another. When two cars collide, why do we think that the driver that ran the red light was "the cause" of the accident, rather than the safe but unlucky driver? Another puzzle involves the interpretation of correlational evidence: Why do people interpret "Aspartame is linked to cancer" as implying that aspartame consumption causes cancer, and not the other way around? The CommuniCause perspective suggests treating these puzzles as traces of pragmatic reasoning: in interpreting what others have said, we can use what we know about how speakers make choices to infer what they are probably trying to convey. In a number of experiments, we explore the detailed predictions of the framework for how speakers make complex choices among various ways of expressing themselves, how causal and non-causal language is interpreted in context, and how linguistic framing influences causal reasoning processes. If these results validate our approach, we can conclude that language matters in causal thought and talk to a much greater extent than previously realized. A particularly interesting contribution of this project with important practical implications is the juxtaposition of causal and non-causal language. We investigate the conditions under which speakers choose non-causal expressions (e.g., "If A, then B" or "A is associated with B") to convey causal meaning, and the circumstances under which listeners infer causal information from expressions that are merely correlational. Contrary to widespread conception, we do not consider causal inference from non-causal language a fault or an irrationality. Rather, the CommuniCause approach is able to shed light on the efficiency and systematicity of this phenomenon, thereby providing a better grip on avoiding miscommunication between domain experts and lay audiences in important areas like health communication.