Causality is central to developmental psychology, psychologists do not only want to identify developmental risks but also to understand mechanisms by which development can be fostered. But sometimes certain conditions or characteristics can't be assigned randomly. So, causal inference - inferring causal relationships - is difficult. An association alone do not reveal causal relationships. The last 30 years have produced superior methods for moving from association to causation. The aim of this article reflects the current state of developmental psychology and is guided by four premises: 

1. Causal inference is essential to accomplishing the goals of developmental psychologists. Causal inference should be the goal of developmental research in most circumstances.
2. In many analyses, psychologists unfortunately are attempting causal inference but doing so hardly, that is, based on many implicit and implausible assumptions. 
3. The assumption should be identified explicit and checked empirically and conceptually. 
4. Developmental psychologists will recognize the central importance of causal inference and naturally embrace the methods available.

This article also wants to promote broader thinking about causal inference and the assumptions on which it rests. But a characteristic of the broader literature, is that methodologists in different fields also differ substantially.

What is the confusion in current practice?

All the articles now lack in one of two directions: both are dissatisfying and potentially misleading. The authors hold causal inference as unattainable. A second group of authors embrace causality, there researchers often rely on the longitudinal nature of their data to make the leap form associations to causality. But authors often leave this assumptions unstated or may unaware of the assumptions themselves. There are also some authors who straddle the two groups, these authors often stray into interpretations of what are associations. But, the situation creates a swamp of ambiguity in which confusion thrives. This 'counterfactual' lies at the heart of causal inference. 

Why causal inference?

Causal thinking and therefor causal inference are unavoidable. One can support this claim in three ways:

1. A major goal of psychology is to improve the lives of humanity. Much of developmental science is devoted to understanding processes that might lead to interventions to foster positive development.
2. Causal analysis is unavoidable, because causal thinking is unavoidable.
3. If a researcher resists the urge to jump from association to causality, other researchers seem willing to do so on his or her behalf.

How is causal inference the goal of Developmental psychology?

It is not the case that causal relationships can never be established outside of random assignment, but they cannot be inferred from associations alone. This research is used to make causal inference as plausible as possible. As part of the proper use of this tools, the researcher should identify the key assumptions on which they rest and their plausibility in any particular application. But what is credible or plausible is not without debate. This paper cannot resolve this issue, but its broader purpose is to establish plausible causal inference as the goal of empirical research in development psychology.

What are the two frameworks for causal inference?

There are two frameworks that are useful for conducting causal inference, and two conceptual tools that are especially helpful in moving from associations to causal relationships. The first involves directed acyclic graph (DAG). This assists researchers in identifying the implications of a set of associations for understanding causality and the set of assumptions under which those associations imply causality. 

What is the DAG?

Computer scientists are also interested in causality, or in particular, in identifying the circumstances under which the association can be interpreted as causal. A DAG comprises variables and arrows linking them. It is directed in a sense that the arrows represent causal relationships. The model assumes a certain correspondence between the arrows in the graph and the relationships between the variables. If you can't trace a path from one variable to another variable, then the variables are not associated. This is the Markov assumption 'the absence of a path implies the absence of a relationship'. A key structure of the DAG is structural stability: an intervention on one component of the model does not alter the broader structure. It also has a preference for simplicity and probabilistic stability. 

The DAG looks like a path diagram but has some distinguish features:

• The DAG is not linear or parametric.
• It contains no bidirectional arrows implying simultaneity
• The essence of the DAG can be grasped by thinking about three variables X,Y and Z. You can think about Z as a common cause of X and Y, about Z as a common effect of X and Y, and about Z as a mediator of X on Y. 

The usefulness of the DAG is perhaps the most apparent when more than three variables are involved, especially when one is unmeasured (think about an unobserved determinant of the mediator).