Poster #23 - Development of Metacognition in Multi-Choice Decision Making

• In Event: Poster Session 14

Abstract

In both children and adults, experiments on information seeking, decision making, changes of mind, and metacognition more broadly almost exclusively use two-alternative, forced-choice paradigms (Li & Ma, 2020). This methodological feature makes sense in the context of simple perceptual judgements and for the ease of implementing formal decision-making models. However, many decisions humans face throughout development involve more than two possible outcomes or courses of action, from selecting an ice-cream flavour to answering an exam question.
Prevailing theories suggest that metacognitive confidence is based on the overall probability that a selected option is optimal (Kepecs & Mainen, 2012; Mamassian, 2016). However, recent findings from three-alternative tasks instead suggest that adults’ confidence reflects the difference between the probabilities of the best and next-best options only, with the third-best option disregarded altogether (Li & Ma, 2020). But to our knowledge, no studies have used multiple-alternative confidence tasks in a developmental context. The planned study will therefore investigate how children’s confidence varies for judgements that involve more than two possible alternatives.
Following our previous work, 6- to 9-year-old children (N = 96) will first be introduced to the “Randomiser”: a digitised lottery machine that randomly selects one box from among multiple boxes containing coloured balls. Participants will also be introduced to a novel continuous scale in the form of a cup, which will allow them to rate how confident they are that a particular-coloured ball will exit the Randomiser. Following a preliminary phase in which children will be shown two balls entering the Randomiser, they will complete a 40-trial test phase in which 12 balls of four different colours enter the Randomiser. Factors to be varied include the number of balls for the best option (3 to 12), the total number of colours in the display (1 to 4), and the probability of each option (see Figure 1). On each trial, participants will be scored for whether they select the best (highest probability) option, as well as for implicit confidence (response latency) and explicit confidence (ratings on the continuous scale).
One hypothesis is that children’s implicit and explicit confidence will vary only as a function of the evidence for the best option, such that they will be just as confident on a trial with a 6-5-1 ball distribution as a trial with a 6-3-3 ball distribution (for example). A competing hypothesis, however, is that just like adults (Li & Ma, 2020), children’s confidence will vary as a function of the difference between evidence for the best and next-best options, such that they will be less confident on a trial with a 6-5-1 ball distribution than a trial with a 6-3-3 distribution (see Figure 1c). Exploratory analyses will further examine whether the number of colours, the arrangement of colours (e.g., clusters or scattered dispersion), and location (e.g., proximity to Randomiser opening) influence children’s predictions and confidence scores.

Authors

• Cate MacColl, University of Queensland
  Presenting Author

• Jonathan Redshaw, University of Queensland