Age-group differences in trust-related decision-making and learning

Facial impressions contribute to evaluations of trustworthiness. Older adults are especially vulnerable to trust violations, incurring risks for deception and exploitation. Using the newly developed social Iowa Gambling Task (S-IGT), we examined age-group differences in the impact of facial trustworthiness on decision-making and learning. In the congruent condition (CS-IGT), advantageous decks were paired with trustworthy faces and disadvantageous decks with untrustworthy faces. In the incongruent condition (IS-IGT), this pairing was reversed. Younger (n = 143) and older (n = 129) participants completed either the standard Iowa Gambling Task (IGT), CS-IGT, or IS-IGT. Both age groups preferred trustworthy faces in their initial choices. Older adults performed worse than younger adults across all tasks over time. Further, compared to younger adults, older adults performed worse on the IS-IGT, suggesting that incongruent facial cues interfered with older adults’ performance, which aligns with reduced sensitivity to negative social reputations in aging. Multilevel modeling also indicated that age-group differences were most pronounced across all tasks in the last 40 trials. Together these findings suggest that differences between younger and older adults in experience-dependent decision-making are magnified in social contexts that involve a “wolf in sheep’s clothing,” which may reflect age-related difficulties in integrating incongruent information.


Results for Blocks 1-4 with Outliers Removed
As reported in the manuscript, no outliers were identified as exceeding +/-3 SD from the mean total performance across blocks within each age group; however, four mean performance scores within blocks were identified as outliers.The pattern of results did not change when these outliers were excluded from the analysis, as reported below.

Full-Factorial Multilevel Model of Effects for Age Group, Task Condition, and Block (Hypotheses 2a, 2b, 2c, 3a, 3b)
Mirroring the results reported in the manuscript, the interaction between Age group and Task condition for overall performance was significant (χ 2 (2) = 11.27;p < .01).
Pairwise comparisons indicated that younger, compared to older, adults performed better overall on the IGT (z = 3.56, p < .001)and the IS-IGT (z = 4.26, p < .001).However, overall performance was comparable between the two age groups in the CS-IGT (z = -0.66,p = .51).Together, these findings supported Hypotheses 2a, 2b, and 2c that, while there were no significant age-group differences in overall performance for the CS-IGT, younger adults significantly outperformed older adults in the IGT and IS-IGT.
In addition, a significant interaction between Age group and Block was found (χ 2 (3) = 13.56;p < .01).Post-hoc examination of age-group differences by block showed that across all task conditions, younger and older adults performed comparably during Block 1 (z = 0.58, p = .56),but younger adults significantly outperformed older adults in Block 2 (z = 2.38, p = .02),Block 3 (z = 3.90, p < .001),and Block 4 (z = 3.95, p < .001).This provides support for Hypothesis 3a, that older age would be associated with less improvement over time across all task conditions, consistent with well-established agerelated impairments in learning.
In addition, the interaction between Task condition and Block was significant (χ 2 (6) = 28.45;p < .001).Post-hoc analysis revealed that Block 1 performance was significantly lower in the IS-IGT than the CS-IGT (z = -4.94,p < .001)and the IGT (z = -2.80,p = .01)across all participants.Block 4 performance was highest in the IGT compared to both the CS-IGT (z = 3.20, p < .01)and the IS-IGT (z = 2.71, p = 0.01).Block 4 performance was comparable between the CS-IGT and the IS-IGT (z = 0.40, p = .69).As discussed in the manuscript, these results suggest that, over time, participants in the IS-IGT learned to ignore the facial cuescomparable to performance levels of participants in the CS-IGT.
However, learning was greatest in the IGT which did not include additional congruent or incongruent facial cues to process.Age group did not significantly interact with Task condition and Block (χ 2 (6) = 5.75; p = .45).

Results with Blocks 1-5
The analyses reported in the manuscript were conducted on the first four of five blocks due to the non-replenishing deck design for the IGT and the S-IGT.Results on the full task (Blocks 1-5) were parallel to those reported in the manuscript, however learning and performance plateaued after Block 4 likely due to the non-replenishing deck design.
Full results for the main analysis with Block 5 included are reported below.

Full-Factorial Multilevel Model of Effects for Age Group, Task Condition, and Block (Hypotheses 2a, 2b, 2c, 3a, 3b)
The interaction between Age group and Task condition was significant (χ 2 (2) = 10.82;p = .005).Pairwise comparisons indicated that younger, compared to older, participants performed better overall on the IGT (z = 4.13, p < .001)and the IS-IGT (z = 4.88, p < .001).However, overall performance was comparable between the two age groups in the CS-IGT (z = 0.18, p = .86).As reported in the main manuscript, these findings supported Hypotheses 2a, 2b, and 2c that while younger and older adults performed comparably on the CS-IGT, younger participants significantly outperformed older participants in the IGT and the IS-IGT.
There was also a significant interaction between Age group and Block (χ 2 (4) = 10.03;p = .04).Post-hoc examination of age-group differences by block showed that across all tasks, younger and older adults performed comparably during Block 1 (z = 0.38, p = .71),but younger adults significantly outperformed older adults in Block 2 (z = 2.21, p = .03),Block 3 (z = 3.58, p < .01),Block 4 (z = 3.58, p < .001),and Block 5 (z = 3.81, p < .001).This again provides support for Hypothesis 3a, that older age would be associated with less improvement over time across all task conditions, consistent with wellestablished age-related impairments in learning.
In addition, the interaction between Task condition and Block was significant (χ 2 (8) = 26.38;p < .001).Post-hoc examinations revealed that Block 1 performance was significantly lower in the IS-IGT than the CS-IGT (z = -4.51,p < .001)and IGT (z = -2.56,p = .01)across all participants.Block 4 performance was highest in the IGT compared to both the CS-IGT (z = 2.76, p < .01)and the IS-IGT (z = 2.30, p = 0.02).Block 4 performance was comparable between the CS-IGT and the IS-IGT (z = 0.41, p = .69).In Block 5, performance in the IGT was greater than performance in the CS-IGT (z = 2.53, p = .01)but not in the IS-IGT (z = 1.53, p = .13).There was no difference in performance between the CS-IGT and the IS-IGT (z = -1.16,p = .25).These findings suggest that learning was greatest in the non-social IGT, but performance plateaued after Block 4. Age group again did not significantly interact with Task condition and Block (χ 2 (8) = 8.24; p = .41).

Table 1 .
Design Factors for Each Hypothesis and Statistical Test All analyses controlled for Data source (Study 1 vs. Study 2) where applicable.