Causal reasoning with ambiguous information

Ambiguous observations result in imprecise estimations of subjective probabilities for rule-based causal reasoning. To date, there is a lack of research on causal reasoning from covariance information involving ambiguous observations. This dissertation examines the impacts of ambiguous information on causal reasoning, and how subjects may treat ambiguous observations. It started with exploring constructs in causal reasoning and evaluating existing models of causal reasoning. This was then followed by a series of experimental studies that investigated the impacts of ambiguous observations on causal reasoning, and how the treatment of ambiguous observations can be influenced by a range of factors. The factors have been explored included external factors such as the distribution of ambiguity, the base rate of the effect, causal valence, the expected causal strength and the interpretation of the ambiguous observations, as well as internal factors such as cognitive resources, working memory capacity and prior beliefs. Finally, a Bayesian hierarchical model was developed to account for the processes of causal reasoning and ambiguity treatment, and was validated in four sets of empirical data. Findings suggested that most people tended to treat ambiguous observations as negative cues against causal associations. The tendency was stronger when the cause was present in the ambiguous observations than when the cause was absent. In addition, the extent to which people were influenced by the base rate of the effect and the expected causal strength depended on the distribution of the ambiguous information. Furthermore, the tendency of people to treat ambiguous observations as negative cues declined with increase in the cognitive load. It was also found that the contribution of working memory capacity to the comprehensiveness in evidence retention in causal reasoning was reduced when there were ambiguous observations. Moreover, people seemed to treat the ambiguous observations as not supporting the prior causal valence provided in the reasoning context. The effects of ambiguity on causal strength were mostly independent of prior belief. Finally, although most people seemed to treat ambiguous information as negative cues, there was individual difference in selecting strategies to treat ambiguous observations. The current findings imply that people become more cautious about causations when being presented with ambiguous information. People are responsive to the characteristics of the ambiguous information and are sensitive to the information available in the unambiguous evidence. Moreover, evaluating ambiguous observations may require extra cognitive resources. Processing ambiguous information may be a deliberate process that is relatively independent from causal reasoning. The treatment of ambiguous observations may occur prior to causal strength estimates, or may be a process that is independent of reasoning about causal strength. The Bayesian hierarchical model accounting for both causal reasoning and strategies of ambiguity processing is ready to be extended and applied to further understand causal reasoning with ambiguous observations