Seeing Versus Doing: Two Modes of Accessing Causal Knowledge.

Top Cited Papers
Publisher Website
Google Scholar
Abstract
The ability to derive predictions for the outcomes of potential actions from observational data is one of the hallmarks of true causal reasoning. We present four learning experiments with deterministic and probabilistic data showing that people indeed make different predictions from causal models, whose parameters were learned in a purely observational learning phase, depending on whether learners believe that an event within the model has been merely observed ("seeing") or was actively manipulated ("doing"). The predictions reflect sensitivity both to the structure of the causal models and to the size of their parameters. This competency is remarkable because the predictions for potential interventions were very different from the patterns that had actually been observed. Whereas associative and probabilistic theories fail, recent developments of causal Bayes net theories provide tools for modeling this competency. Causal knowledge underlies our ability to predict future events, to explain the occurrence of present events, and to achieve goals by means of actions. Thus, causal knowledge belongs to one of our most central cognitive competencies. However, the nature of causal knowledge has been debated. A number of philosophers and statisticians, such as Bertrand Russell (1913) and Karl Pearson (1892), have dismissed the notion of causality altogether and tried to replace it with the idea of correlation. This idea may be traced back to Hume (1748/1977), who argued that causality is an illusion based on associations that are produced by the experience of constant conjunctions of events. A modern variant of this approach is represented by theories that attempt to reduce causal learning to the acquisition of associative links between event representations (e.g., Shanks & Dickinson, 1987). One fundamental problem of this view is that it collapses observational knowledge with interventional knowledge. Causal knowledge serves two different functions: It allows us to predict events on the basis of observed cues and at the same time underlies our ability to manipulate and control. For example, we can proba- bilistically predict the weather from readings of the barometer, and this prediction is driven by causal relations underlying the (spuri- ous) covariation. Nevertheless, we also know that artificially set- ting the barometer to a specific reading would do nothing to the weather. Causal knowledge not only allows us to predict events on the basis of observed cues, it also tells us whether and which effects our actions will have. Although both types of prediction are driven by a common underlying causal model, the predicted out- comes may differ depending on whether the events are merely observed or actively set.

This publication has 22 references indexed in Scilit: