Elsevier

Cognition

Volume 82, Issue 1, November 2001, Pages 27-58
Cognition

Estimating causal strength: the role of structural knowledge and processing effort

https://doi.org/10.1016/S0010-0277(01)00141-XGet rights and content

Abstract

The strength of causal relations typically must be inferred on the basis of statistical relations between observable events. This article focuses on the problem that there are multiple ways of extracting statistical information from a set of events. In causal structures involving a potential cause, an effect and a third related event, the assumed causal role of this third event crucially determines whether it is appropriate to control for this event when making causal assessments between the potential cause and the effect. Three experiments show that prior assumptions about the causal roles of the learning events affect the way contingencies are assessed with otherwise identical learning input. However, prior assumptions about causal roles is only one factor influencing contingency estimation. The experiments also demonstrate that processing effort affects the way statistical information is processed. These findings provide further evidence for the interaction between bottom-up and top-down influences in the acquisition of causal knowledge. They show that, apart from covariation information or knowledge about mechanisms, abstract assumptions about causal structures also may affect the learning process.

Introduction

Causal learning is central for our survival. Causal knowledge allows us to anticipate harmful or gratifying events, and to plan actions to achieve goals. Despite the fact that a large number of philosophers and psychologists agree on the importance of research on causality, no unitary concept has evolved so far. The main question of how we distinguish causal relations from accidental sequences of events remains highly debated.

Section snippets

Covariation view

In the past 30 years, philosophers and psychologists have become increasingly interested in probabilistic notions of causality. Our knowledge about causal relations, such as “Smoking causes lung cancer”, is often based on the observation of covariations between causes and effects. A number of philosophers have proposed a notion of causality that reduces causal relations to observable statistical laws (e.g. Eells, 1991, Salmon, 1971, Suppes, 1970). Roughly, it has been proposed that causes alter

Overview of experiments

The goal of the experiments that will be presented in this article is to show that prior knowledge and processing effort guide the strategies of causal strength estimation. A number of previous studies have investigated causal structures in which multiple causes converge on a common effect, and have demonstrated that learners take potential causal co-factors into account (e.g. Baker et al., 1993, Price and Yates, 1993, Spellman, 1996a, Spellman, 1996b). In contrast, the aim of the present

Experiment 1

The Cartwright (1983) analysis of the Berkeley admission problem has demonstrated that in situations with multiple potential causes of a common effect, conditional contingencies should be assessed which control for causally relevant co-factors. Causally irrelevant co-factors should be ignored. Controlling for causally irrelevant co-factors may lead to distortions of the underlying causal relations. Because the causal relevance of the co-factors has to be established prior to the induction task,

Experiment 2

In Experiment 1 two different grouping variables were compared, type of fruit (Mamones and Taringes) and investigators (A and B). Experiment 2 attempted to replicate the results of Experiment 1 with a grouping variable that was kept constant across the two conditions. Thus, all participants saw identical cases, received identical rating instructions, and were informed about identical subcategories. The only manipulation involved a hint about the potential causal relevance of the co-factor

Experiment 3

The previous experiments have explored the conditions in which participants calculate contingencies between a potential cause and an effect conditional upon a third correlated event. This third event was either described as causally relevant or irrelevant for the cause–effect relation in question. In these experiments, the co-factor was typically introduced as a variable that potentially interacted with the target cause in producing the effect. This is a situation in which it is appropriate to

General discussion

The goal of causal induction is to arrive at representations of objective causal relations. Typically causal relations cannot be observed directly but must be inferred on the basis of statistical relations between observable events (see Cartwright, 1989, Cheng, 1997). The presented studies focus on the problem that there are multiple ways of extracting statistical information from a set of events. Which method is appropriate is partly determined by the hypothesized causal structure underlying

Acknowledgements

We would like to thank Merideth Gattis for helpful comments. Experiment 1 was presented at the 1995 meeting of the Cognitive Science Society, Pittsburgh, PA. The first two experiments were planned and conducted while the authors were affiliated with the Max Planck Institute for Psychological Research, Munich, and the University of Tübingen.

References (62)

  • B.J Bickel et al.

    Sex bias in graduate admissions: data from Berkeley

  • C.R Blyth

    On Simpson's paradox and the sure-thing principle

    Journal of the American Statistical Association

    (1972)
  • M Bullock et al.

    The development of causal reasoning

  • N Cartwright

    How the laws of physics lie. Essay 1

    (1983)
  • N Cartwright

    Nature's capacities and their measurement

    (1989)
  • G.B Chapman et al.

    Cue interaction in human contingency judgment

    Memory & Cognition

    (1990)
  • L.J Chapman et al.

    Genesis of popular but erroneous psychodiagnostic observations

    Journal of Abnormal Psychology

    (1967)
  • L.J Chapman et al.

    Illusory correlation as an obstacle to the use of valid diagnostic signs

    Journal of Abnormal Psychology

    (1969)
  • P.W Cheng

    From covariation to causation: a causal power theory

    Psychological Review

    (1997)
  • P.W Cheng et al.

    Covariation in natural causal induction

    Psychological Review

    (1992)
  • C Cherniak

    Minimal rationality

    (1986)
  • R.M Dawes

    Rational choice in an uncertain world

    (1988)
  • J Dupré

    The disorder of things. Metaphysical foundations of the disunity of science

    (1993)
  • E Eells

    Probabilistic causality

    (1991)
  • E Eells et al.

    Probabilistic causality and the question of transitivity

    Philosophy of Science

    (1983)
  • Fiedler, K., Walther, E., Freytag, P., & Stryczek, E. (in press). Playing mating games in foreign cultures: a...
  • G Gigerenzer et al.

    Reasoning the fast and frugal way: models of bounded rationality

    Psychological Review

    (1996)
  • Jenkins, H. M., & Ward, W. C. (1965). Judgment of contingency between responses and outcomes. Psychological Monographs,...
  • B Koslowski

    Theory and evidence. The development of scientific reasoning

    (1996)
  • J.W Payne et al.

    The adaptive decision maker

    (1993)
  • Cited by (94)

    • The role of action tendencies in expert anticipation

      2023, Asian Journal of Sport and Exercise Psychology
    • How people learn about causal influence when there are many possible causes: A model based on informative transitions

      2018, Cognitive Psychology
      Citation Excerpt :

      One of the goals of the current research was to better understand how people statistically ‘control’ for alternative cues, since prior experiments mainly focused on whether participants controlled for alternative cues, but not how they did so (e.g., Spellman, 1996, p. 342). Many of the previous experiments that tested whether participants controlled for alternative cues used a design involving Simpson’s paradox with two cues (e.g., Spellman, 1996; Spellman et al., 2001; Waldmann & Hagmayer, 2001). However, we could not use this paradigm in the current experiements: the design was developed for two cues, and even if we could figure out a way to generalize the Simpson’s paradox design to eight cues, it still would not allow us to test the hypotheses about one-change transitions.

    View all citing articles on Scopus
    View full text