A computational model combining informed and non-informed advice-taking with individual learning (Model 5) best predicts participants' choices¶

Claim · All claims
Evidence · 1 study · causal r2 · q2
Currency · newest study 2024 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Model 5, including individual learning, informed advice-taking and non-informed advice-taking, outperformed all nested alternatives in leave-one-block-out comparison. → Maayan Pereg 2024

Evidence¶

Maayan Pereg 2024¶

Maayan Pereg, Uri Hertz, Ido Ben-Artzi and Nitzan Shahar. (2024). Disentangling the contribution of individual and social learning processes in human advice-taking behavior. npj Science of Learning 9:4. https://doi.org/10.1038/s41539-024-00214-0

q2 · i? · causal · r2

Leave-one-block-out cross-validated model comparison (elpd via the loo package) of five prediction-error-based RL models on the empirical choice data; Table 1 shows the null model at −4147.2 (88.4) elpd difference versus Model 5, with Model 2 at −514.1 (33.4).

"We found Model 5 to be the winning model"

Discussion¶

Related Claims¶

• Simulated data show a double dissociation: early-trial reveal effects arise only from non-informed advice-taking, while outcome-dependent coherence repetition arises only from informed advice-taking — related
• Estimated parameters indicate simultaneous reliance on individual learning (ω ≈ 0.70) and a fixed non-informed bias to follow advice (φ ≈ 0.2) — related
• Mentalizing-network regions (TPJ, STS, dmPFC, MTG) signal social prediction errors beyond ToM tasks — related