A single decision-and-confidence model with no condition-specific parameters fits decision weights across all three methods¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q2 · i3 large
Currency · newest study 2022 · not yet checked against its sources · not reviewed by a person · overdue for a recheck since 2025: tool and model accuracy changes quickly

Subclaims¶

q2 i? The full model accommodated the general pattern of decision weights without any parameter varying between the confidence-assessment methods. → Lorenz Weise 2022 q2 i3 The model's predicted proportion of high-confidence trials correlated significantly with the observed proportion (r = 0.63). → Lorenz Weise 2022 (2)

Evidence¶

Lorenz Weise 2022¶

Lorenz Weise, Saskia D. Forster, Siegfried Gauggel. (2022). Reverse-correlation reveals internal error-corrections during information-seeking. Metacognition and Learning (2022) 17:321–335. https://doi.org/10.1007/s11409-021-09286-4

q2 · i? · causal · r2

Simulation-based model fitting: 100,000 simulated trials per run, grid search plus Nelder-Mead fitting of cutoff and scale to each participant's weights. The article reports the full model accommodated the general pattern with no condition-specific parameters.

"the full model can accommodate the general pattern in the decision weights, without allowing any parameter to vary between the different methods of metacognitive judgment."

Lorenz Weise 2022 (2)¶

Lorenz Weise, Saskia D. Forster, Siegfried Gauggel. (2022). Reverse-correlation reveals internal error-corrections during information-seeking. Metacognition and Learning (2022) 17:321–335. https://doi.org/10.1007/s11409-021-09286-4

q2 · i3 · causal · r2

Out-of-sample validation: predicted high-confidence trial proportions, not fitted by the model, were correlated with each participant's observed proportion averaged across conditions, with printed effect size r = 0.63 (reported as r(25) = 0.63, p < 0.001), a large correlation.

"The model's prediction was correlated significantly with the empirically observed value (r(25) = 0.63, p < 0.001), further corroborating the model."

Discussion¶

Related Claims¶

• Reverse-correlation decision weights are positive for the chosen stimulus dimension and weaker on low-confidence trials across confidence-assessment methods — related
• Internal error-corrections, not correct-to-incorrect changes-of-mind, produce the inverted low-confidence decision weights — related
• Low-confidence info-seeking trials show inverted decision weights, unlike explicit reports and post-decision wagering — related
• The inverted low-confidence info-seeking pattern persists when only each participant's first condition is analyzed — reports the opposite
• Primary task accuracy is higher in the info-seeking condition than in the explicit reports and wagering conditions — related