About this srcd poster session

---

Panel information
Panel 4. Cognitive Processes

Abstract

---

Causal reasoning is crucial for navigating the world, yet little is known about how causal relations are represented in the mind. One proposal is the “causal graphical models” (CGMs) position, which suggests that people represent causal events as structured models that follow the Markov condition. This condition states that a variable is conditionally independent of its non-descendants, given its parents. For instance, consider the relations between smoking, lung cancer, and yellow fingers. These can be represented as a “common cause” CGM, where smoking causes both lung cancer and yellow fingers. This pattern of conditional relations means that if we discover that a person smokes, the occurrence of yellow fingers provides no additional information about that person's likelihood of getting lung cancer (and vice versa). Understanding how children represent causal events is important theoretically because it can provide insight into the mechanisms that underlie those representations specifically and causal reasoning generally.

Currently, little is known about whether children represent causal events as CGMs and are sensitive to the Markov condition. To explore this, 3- to 4-year-old children (Ncurrent = 55; Ntarget = 66, meanage = 48.10 months, range = 36.67 to 59.3) were trained on one of three, 3-cause causal structures. Figure 1 illustrates the sequences used for one such structure. Each potential cause was represented by a box, and each box was positioned side by side on a screen. During three training trials, red semi-circles emerged either from all the boxes or from a subset of them. Children’s task was to determine which box caused the semi-circles to emerge from the other boxes, including itself.

At test, children were presented with two events: one that preserved the conditional relations from training (the Consistent event) and one that violated them (the Inconsistent event). These test events were shown simultaneously, and children were asked which event was more like what they observed during training. Preliminary results indicate that children chose the Consistent test event more often than the Inconsistent one, odds ratio = 1.82, 95%CI [1.24, 2.79], p < .005. This suggests children may be sensitive to the Markov condition.

However, it is possible children chose the Consistent event simply because it was familiar from training. Experiment 2 aims to address this limitation. Instead of seeing the red semi-circle emerge from two boxes at test, as in Experiment 1, in Experiment 2 it will emerge either from the sole cause’s box or from one of the caused object’s boxes. If children in Experiment 1 were responding based on the conditional relations they observed, they should continue to select the Consistent event; otherwise, they may choose randomly between the events, since neither was shown during training. Data collection for Experiment 2 has just begun; it is expected that data collection for both projects will be complete by the start of the SRCD conference. The proposed poster presentation will include both experiments.

Author information

Author	Role
Deon T. Benton, Vanderbilt University	Presenting author

⇦ Back to session