Confidence Reasoning in Children’s Memory and Perceptual Decisions

• In Event: Poster Session 14
  In Poster Session: PS 14 Topic - Cognitive Processes

Abstract

Confidence is experienced in many decisions: choosing the biggest piece of cake or remembering where we left our keys. Each of these decisions relies on a different set of systems, representations, and units (perception and memory, respectively). Some theories about metacognition propose that confidence in each decision type is also distinct, develops independently, and slowly merges over time into a domain-general sense of confidence (Geurten et al., 2018; Vo et al., 2014). Other theories suggest that metacognition is domain-general throughout, even though the underlying decisions stem from different sources (Baer & Odic, 2020). Here, we contrast domain-specific versus domain general accounts by comparing metacognitive processes in children’s perceptual and memory decisions in two preregistered studies.
In Study 1, we examine whether metacognitive monitoring (e.g., indicated by higher confidence for correct answers than for incorrect answers) correlates across memory and perception in children. Seventy-four 4-7-year-olds (of pre-registered 132) performed a memory recognition task (was this item blue or yellow earlier?) and an area perception task (which shape is larger?) followed by three-point confidence judgments (Hembacher & Ghetti, 2014). Replicating past work (e.g., Hembacher & Ghetti, 2014), children reported higher confidence on correctly-answered items and lower confidence on incorrect items for both memory, F(1, 70) = 20.26, p < .001, ηp2 = 0.22, and perception tasks, F(1, 70) = 29.08, p < .001, ηp2 = 0.29. However, using four measures of metacognitive monitoring (difference scores, meta-d’, MRatio, and HMeta-d’, see Fleming, 2017; Maniscalco & Lau, 2012), we find no significant correlations between memory and perceptual tasks. Instead, confidence judgments appear to be computed domain-specifically for 4-7 year-olds’ memory and perceptual decisions.
In Study 2, we extend upon this work and examine whether children can integrate and compare confidence representations for memory and perception directly. Thirty-three 4-7-year-olds (of pre-registered 132) performed the same memory and area perception tasks as in Study 1, but in a forced-choice confidence paradigm (Baer & Odic, 2020). On each trial, children answered two questions, either from the same domain (e.g., memory and memory, “Within-Domain” condition) or from different domains (memory and perception, “Across-Domain” condition), and then indicated which answer they were most sure was correct. Children were more accurate on chosen than discarded items, F(1, 31) = 8.53, p = .006, ηp2 = .22, but this likely differs by condition, F(1, 31) = 2.97, p = .095, ηp2 = .09. Thus, although children can compare their confidence for memory and area, there is an additional cost in Across-Domain trials, suggesting distinct types of representations for the two tasks.
Together, these studies inform our understanding of how metacognitive judgments are computed in childhood, suggesting that metacognition may be computed independently and incomparably in memory and perceptual decisions. This work has not been accepted at another conference, and data collection and analysis are incomplete. Data collection will take approximately 5 months to complete the pre-registered sample size. Results of Study 2 may change, but are informative both if there is an effect or not.

Authors

• Carolyn Baer, University of California - Berkeley
  Presenting Author

• Simona Ghetti, University of California - Davis
  Non-Presenting Author

• Darko Odic, University of British Columbia (UBC)
  Non-Presenting Author