Finding the Cause: Examining the Role of Qualitative Causal Inference through Categorical Judgments - eScholarship
Michael Pacer, Woo‐kyoung Ahn · Proceedings of the Annual Meeting of the Cognitive Science Society · 2011
Finding the Cause: Examining the Role of Qualitative Causal Inference through Categorical Judgments Michael Pacer ([email protected]) Department of Psychology, University of California, Berkeley Berkeley, CA 94720 Woo-kyoung Ahn ([email protected]) Department of Psychology, Yale U., 2 Hillhouse Ave New Haven, CT 06520 Abstract Previous work showed that people‟s causal judgments are modeled better as estimates of the probability that a causal relationship exists (a qualitative inference) than as estimates of the strength of that relationship (a quantitative inference). Here, using a novel task, we present experimental evidence in support of the importance of qualitative causal inference. Our findings cannot be explained through the use of parameter estimation and related quantitative inference. These findings suggest the role of qualitative inference in causal reasoning has been understudied despite its unique role in cognition. Further, we suggest these findings open interesting questions about the role of qualitative inference in many domains. Keywords: Causal Categorization Reasoning, Qualitative Inference, Qualitative and Quantitative Causal Inference Studies have distinguished between two types of causal inferences: qualitative inferences (“whether an event X is or is not a cause of Y”) and quantitative inferences (“to what extent is X a cause of Y”; Griffiths & Tenenbaum, 2005; Waldmann & Martignon, 1998). Causal reasoning has been traditionally construed as making quantitative inferences of causal strengths (e.g., Cheng, 1997; Rescorla & Wagner, 1972), in which data on presence or absence of two events 1 are used to estimate parameters, which describe the degree of a relationship between the cause and effect. Within this causal strength paradigm, qualitative inferences are made by assessing the parameter estimates relative to a threshold. If the estimate is above that threshold, then it is a cause; if it is below, then it is not a cause. Such an approach casts qualitative causal inference as a process dependent upon quantitative inference. In the current study, we demonstrate that qualitative causal inference may play an important role such that people will make judgments which conflict with the judgments warranted by quantitative inference. Quantitative and Non-Quantitative Data in Causal Inference While quantitative data can be used in making qualitative We refer to the type of information captured in simple contingency tables as quantitative information. causal inferences of whether one event causes another, non- quantitative data 2 can also be used in making qualitative inferences. Previous studies (e.g., Griffiths & Tenenbaum, 2005; 2009; Kuhn 1997; Waldmann & Martignon, 1998) have proposed that qualitative inference utilizes information such as intuitive theories (e.g., an inductive bias such as “novel foods may have unusual causal features, but common foods do not”), temporal information (e.g., events exert their influence on the future, not the past), or explicit claims (e.g., “X causes Y and Z”). For instance, if a person‟s lips turn green after eating a novel fruit, she may conclude from this single observation that the fruit caused the discoloration. However, it is unlikely that she will infer that the symptoms caused her to eat the fruit or that a fruit with which she had a great deal of experience (e.g. an apple) caused it. Additionally, if no abnormal symptoms present after eating the novel fruit, she may not think the fruit has any unusual causal features. Low probability events (e.g. green lips) prompt people to seek out causal explanations (e.g., Weiner, 1985), and if there is a novel, preceding event, that event may be identified as the cause even if quantitative data do not support that inference (e.g., Hilton & Slugoski, 1986; Kahneman & Miller, 1986). The Role of Qualitative and Quantitative Inference What remain unknown are details of how qualitative and quantitative inferences interact with one another (Griffiths & Tenenbaum, 2009), or whether one is more fundamental to inference generally. Previous work on causal inferences could be thought of as assuming that quantitative causal inferences are more fundamental, with qualitative inferences generally treated as an afterthought to be computed using a threshold and prior quantitative inferences. Griffiths and Tenenbaum (2005) took another approach. Rather than estimating a single value for the strength of the causal relationship, Griffiths & Tenenbaum‟s model asks “whether or not a causal relationship exists” (i.e. a qualitative inference). In order to answer this, their model uses a large number of possible values for the strength of the causal relationship. In this way, their model gives the probability of the existence of a causal relationship without By non-quantitative we simply mean any type of data that cannot be captured in a simple contingency table.