Model-based strategy use partially mediates the effect of WM load on RL learning performance¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q2–q3
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Bootstrapped mediation showed WM load negatively predicted mixing weight (path a = −0.232), which positively predicted reward rate (path b = 0.090); the indirect effect (ACME = −0.021) accounted for 24% of the total effect, with the direct effect remaining significant. → Chengyan Yang 2025 q2 i? A Monte Carlo simulation with 1000 replications estimated the mediation analysis had 96.4% statistical power to detect the ACME at the study's sample size. → Chengyan Yang 2025 (2)

Evidence¶

Chengyan Yang 2025¶

Chengyan Yang, Tongran Liu, Mengxin Wen & Xun Liu. (2025). The distinct functions of working memory and intelligence in model-based and model-free reinforcement learning. npj Science of Learning 10:72. https://doi.org/10.1038/s41539-025-00363-w

q3 · i? · causal · r2

Exploratory mediation analysis (1000 bootstrap samples) testing whether mixing weight mediated the WM load effect on corrected reward rate. The direct effect (ADE = −0.066) remained significant, indicating partial mediation; total effect = −0.087.

"WM load negatively predicted mixing weight (path a= −0.232, p < 0.001), which in turn positively predicted reward rate (path b = 0.090, p < 0.001). The indirect effect (ACME =−0.021, 95% CI [−0.037, −0.008], p = 0.002) accounted for 24% of the total effect (Proportion Mediated)"

Chengyan Yang 2025 (2)¶

Chengyan Yang, Tongran Liu, Mengxin Wen & Xun Liu. (2025). The distinct functions of working memory and intelligence in model-based and model-free reinforcement learning. npj Science of Learning 10:72. https://doi.org/10.1038/s41539-025-00363-w

q2 · i? · causal · r2

Monte Carlo simulation with 1000 replications using the observed path coefficients and N = 100 (50 per condition) to estimate power for the mediation test. The article cautions the mediation results should be interpreted with caution given the relatively small sample size.

"The results demonstrated that the current media- tion analysis achieved a statistical power of 96.4% (α = 0.05) for detecting the ACME, given the speci fied model parameters and sample size."

Discussion¶

Related Claims¶

• High WM load shifts learners from model-based toward model-free reinforcement learning strategies — related
• Intelligence correlates with RL performance under low WM load but not under high WM load — related
• Participants earned higher corrected reward rates under low WM load than high WM load in a two-stage RL task — related
• Model-based strategy use and intelligence positively predict RL learning performance, while WM load negatively predicts it — related
• The effect of tool type on hypervideo product quality is partially mediated by performing fewer macro-actions — related