Dot by Dot: Seeing Where The Children Take Us
When testing children in the laboratory, we often face the challenge of designing tasks that can both be completed quickly and produce rich data. Here we illustrate the workings of a novel task developed with the Child Brain Development Lab that made use of an adaptive algorithm to shorten assessment time. The algorithm ran in real-time, choosing the next item presentation based on the participant’s prior responses.
The novel task (you can find more information on the StudioLab website), aimed to assess cost-benefit decision making in children. When individuals make decisions, they consider both potential rewards and potential costs. For example, I will consider spending 50 minutes sweating away in the kitchen (cost) to make some warm, Irish soda bread for my lunch the next day (reward). In this case I make a favorable calculation – the benefits outweigh the estimated costs. However, if the cost of making this same bread was greater, say a recipe calling for over 60 minutes of effort, the outcome of my cost-benefit analysis would push me to just make a plain tuna sandwich instead, foregoing the potential reward of the Irish bread because the extra 10 minutes of sweating in the kitchen is too much.
The ACE project and task was designed to examine between-person differences in the ways that children make cost-benefit decisions. A sample of 8-year olds were presented with a series of cards, each worth a specific point value (reward) that ranged from 1 (low) to 10 (high), and were instructed to earn as many points as possible as the points could be exchanged at the end for a prize. When cards were played successfully, they won the points. However, playing a card also came with a cost, also ranging from 1 (low) to 10 (high). The costs for this particular part of the task were effort costs. More specifically, the higher the cost, the more nonsense words the participants had to alphabetize in order to earn the reward. On each trial, children decided if they wanted to play the card or skip to the next card. They were told that they could only skip for an undisclosed, limited number times before the deck ran out.
Figure 1: Task decision space containing 100 potential reward and cost value pairings.
In principle, each child’s cost-benefit decision space could be mapped with 100 decisions (10 reward options X 10 cost options; see fig. 1). However, the adaptive algorithm allowed us to get a good sense of what decisions participants would make using many fewer scenarios. To illustrate the workings of the algorithm, we created the plots below. These represent data from three participants and show the cards they were presented with and whether the participants chose to play or skip each card. They also show the order in which the task progressed, dot by dot showing us where the children’s responses pushed the algorithm.
Figure 2: Decision made by three different participants in the order in which they occurred during the task.
The plots above show how efficiently the algorithm was able to navigate the decision space to extract information about each child’s cost-benefit decision making. They also give a snapshot of just how different decision making is across children.