Toward a Conception of Causality Applicable to Experimentation and Causal Modeling

Abstract

MULAIK, STANLEY A. Toward a Conception of Causality Applicable to Experimentation and Causal Modeling. CHILD DEVELOPMENT, 1987, 58, 18-32. Misuses, as well as undeserved criticism, of causal modeling have resulted from inappropriate conceptions of the assumptions about causality that underlie the causal modeling methods. In an attempt to bring more clarity to the issue, I examine and reject common criticisms of the causality concept and focus on showing how causality is a relation implied in the grammar of a language about objects. This relation concerns how, by means of functional relations, variable properties of objects determine the variable properties of other objects. Next I emphasize how conceptions of causal relations must be objective by uniting all our diverse observations of these relations according to rules. Then I show how the functional relation concept of causality may be modified to have causes determine not specific outcomes but the probabilities of outcomes, thereby synthesizing determinism with probabilism. This result unifies numerous probabilistic models in psychology as causal models. I then consider how these fundamental considerations about causality may be translated into a network of assumptions that must be explicitly considered in any attempt at causal modeling. This reveals how the same principles underlying experimental efforts to establish causality also underlie causal modeling. I also consider the role of a priori assumptions in causal modeling.