Decomposing decision-making in preschoolers: Making decisions under ambiguity versus risk

Abstract

Decision-making in the real world involves multiple abilities. The main goal of the current study was to examine the abilities underlying the Preschool Gambling task (PGT), a preschool variant of the Iowa Gambling task (IGT), in the context of an integrative decision-making framework. Preschoolers (n = 144) were given the PGT along with four novel decision-making tasks assessing either decision-making under ambiguity or decision-making under risk. Results indicated that the ability to learn from feedback, to maintain a stable preference, and to integrate losses and gains contributed to the variance in decision-making on the PGT. Furthermore, children's awareness level on the PGT contributed additional variance, suggesting both implicit and explicit processes are involved. The results partially support the integrative decision-making framework and suggest that multiple abilities contribute to individual differences in decision-making on the PGT.

Fig 1. Set up for Preschool Gambling task.

Children chose from either the chick or giraffe decks (left). Losses and gains were marked by moving the magnetic marker up and down the stairs (right). Note that the illustration of materials is similar but not identical to the original materials used and is for illustrative purposes.

Fig 2. Set up for the Trash to Treasure task (passive avoidance).

Children could either pull up the treasure chest to see if there was a gold coin or pull down the trash can to move on to the next trial. Note that the illustration of materials is similar but not identical to the original materials used and is for illustrative purposes.

Fig 3. Set up for the Toy Choice task (choice consistency).

Following the rating of 6 toys, children pointed to their favorite of two toys. Note that the illustration of materials is similar but not identical to the original materials used and is for illustrative purposes.

Fig 4. Set up for the Risky Sticker task (risky decision-making).

Example (a) shows a trial using the gain frame while example (b) shows a trial using the loss frame. Children were asked to choose from either the safe or risky envelope. While the two possible options were illustrated on the envelope, children did not know which option they were getting until they chose one of the cards. Note that the illustration of materials is similar but not identical to the original materials used and is for illustrative purposes.

Fig 5. Waiting Game task (delay of gratification).

Children were given a paper bag for their accumulated toys. The transparent box was placed over the toys, which required a waiting period to earn. Note that the illustration of materials is similar but not identical to the original materials used and is for illustrative purposes.

Fig 6. Proportion of exploitation and exploration on the PGT.

The fitted models and means from the data (for comparison) are shown. (A) Model F from Table 3 is compared to the exploitation data means. Model F was fitted using data means. (B) Model E from Table 4 is compared to the exploration data means using the data means.

Fig 7. The prediction of exploitation using simpler decision-making tasks.

(A) Model F from Table 3 is fitted for children scoring 1 SD above and below the mean on PA1. The sample exploitation means are shown for children above (n = 60) and below the mean (n = 74) on PA1 for comparison. (B) Model F from Table 3 is fitted for children scoring 1 SD above and below the mean on DoG. The sample exploitation means are shown for children above (n = 62) and below the mean (n = 71) on DoG for comparison. (C) Model F from Table 3 is fitted for children scoring 1 SD above and below the mean on RDloss. The sample exploitation means are shown for children above (n = 37) and below the mean (n = 96) on RDloss.

Fig 8. The prediction of exploration using simpler decision-making tasks.

(A) Model E from Table 4 is fitted for children scoring 1 SD above and below the mean on Trans Err. The sample exploration means are shown for children above (n = 66) and below the mean (n = 68) on Trans Err for comparison. (B) Model E from Table 4 is fitted for children scoring 1 SD above and below the mean on RDloss. The sample exploration means are shown for children above (n = 37) and below the mean (n = 96) on PA1 for comparison. (C) Model E from Table 4 is fitted for children scoring 1 SD above and below the mean on PA1. The sample exploration means are shown for children above (n = 60) and below the mean (n = 74) on RDloss.

Fig 9. Prediction of awareness on the PGT.

Model E of Table 5 was fitted for children who were 1 SD above and below the mean on age in months on hunch and conceptual awareness questions. For comparison, the awareness test scores for the sample are plotted for children who were older (n = 71) and younger (n = 62) than the sample mean (M = 49.99, SD = 7.27).

Fig 10. Prediction of exploitation and exploration using decision-making tasks and awareness levels.

(A) Model G from Table 3 is fitted for exploitation as a function of awareness level (-1, 0, and 1). For comparison, group exploitation means for children classified in the conceptual level group (n = 50), the hunch level group (n = 35) and pre-hunch level (n = 49) are shown. (B) Model F of Table 4 is fitted for exploration as a function of awareness level. For comparison, group exploitation means for children classified in the conceptual level group, the hunch level group and pre-hunch level are shown. Abbreviations: PGT = Preschool Gambling task, CAL = conceptual awareness level, HAL = hunch awareness level, and PHAL = pre-hunch awareness.