Workshop: Causality in Psychological Modeling

Organized by Jan-Willem Romeijn (University of Groningen) and Markus Eronen (Groningen / Leuven)

University of Groningen, Department of Theoretical Philosophy

Programme May 15, 2017

Location

Oude Boteringestraat 52, room Omega

Registration

There is no fee, but if you would like to attend, please send an email to markus.eronen kuleuven.be

Abstracts

Roland Poellinger (MCMP/LMU)

Causal Similarity, Evidential Relevance, and Analogical Inference

Analogical arguments are ubiquitous vehicles of knowledge transfer in science and medicine. This talk builds on a Bayesian evidence-amalgamation framework for the purpose of formally exploring analogy-based inference patterns in causal assessment, especially when direct tests on the target are not feasible. By relating formal explications of relevance, similarity, and analogy, different sources of confirmatory support for a causal hypothesis are distinguished in reconstruction.

Laura Bringmann (Groningen)

Common cause vs. network models in psychopathology: A needless dichotomy

The network approach is becoming increasingly popular in clinical psychology. In most papers, the key rationale for this approach is the contrast between common cause or latent variable models and network models. We argue that this way of justifying the network approach is misguided.

Denny Borsboom (UvA Amsterdam)

The problem with subscript i: What psychometric models do and do not imply about individuals

To analyze data from IQ-tests, personality questionnaires, and other measures, psychologists commonly utilize psychometric models. In these models, observable behaviors (e.g., a given response to the question "do you like parties?") are related to an underlying construct (e.g., one's level of extraversion). On the face of it, it looks like the relation between the construct and the items is causal: the construct acts as a common cause with respect to the item responses. This interpretation is consistent with the emerging literature that frames causality in terms of statistical conditional independence relations and with the way the models are used in practice. However, as I will argue in this talk, the causal relations posited in psychometric models have far less to say about individual people than is commonly supposed: constructs as represented in psychometric models are composed entirely of differences between people that are not linked to psychological attributes as they exist at the level of the individual person. For this reason, I suggest that causal claims in these models should not be understood as indirect claims about the psychology of the individual. A corollary of this viewpoint is that knowledge about the causal effects of differences between people cannot be reduced to knowledge about the causal processes that take place within people.

Naftali Weinberger (Tilburg)

Rethinking the distinction between Between-Subject and Within-Subject Causes

In an influential paper, Borsboom, Mellenbergh and van Heerden (2003) argue that latent variables such as intelligence should be understood as 'between-subject' causes, rather than 'within-subject' causes. Weinberger (2015) criticizes the tenability of this distinction, while Borsboom (2015) responds by noting ways that Weinberger's argument overgeneralizes. Here I propose ways to make headway in this debate. I explain why Borsboom's response cannot by itself save the relevant distinction, and propose alternative distinctions that might be more useful for psychometrics.

Markus Eronen (Groningen)

Interventionism and within-person causal inference in psychology

In philosophy of psychology and psychological research, it is generally assumed that it is possible to discover within-person causes: Causal relationships that describe the workings of individual human minds. However, from the perspective of interventionism, which is the state-of-the-art theory of causation in philosophy of science, finding such causes faces a fundamental problem: How could we make targeted interventions on particular psychological states of human subjects? In this talk, I discuss the challenges to inferring within-person interventionist causes, as well as possible ways to overcome them.