Reverse engineering of metacognition

  1. Matthias Guggenmos  Is a corresponding author
  1. Health and Medical University, Institute for Mind, Brain and Behavior, Germany
  2. Charité – Universitätsmedizin Berlin, Department of Psychiatry and Neurosciences, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Germany
9 figures, 1 table and 1 additional file

Figures

Computational model.

Input to the model is the stimulus variable x, which codes the stimulus category (sign) and the intensity (absolute value). Type 1 decision-making is controlled by the sensory level. The processing …

Figure 2 with 1 supplement
Psychometric functions for different settings of sensory model parameters.

Top left legends indicate the values of varied parameters, bottom right legends settings of the respective other parameters. (A) The sensory bias parameter δs horizontally shifts the psychometric …

Figure 2—figure supplement 1
Nonlinear transformation of the stimulus variable.

Early visual processing likely involves nonlinear transformations of stimulus signals, including processes such as contrast gain control nonlinearities or nonlinear transduction. In the toolbox, …

Figure 3 with 1 supplement
Effect of model parameters on the evidence-confidence relationship.

All metacognitive bias parameters and noise parameters affect the relationship between the sensory evidence |y| and confidence, assuming the link function provided in Equation 5. (A) Effect of …

Figure 3—figure supplement 1
Confidence link functions.

Alternative choices for link functions provided by the ReMeta toolbox describing the relationship between metacognitive evidence and confidence. Note that these link functions do not compute the …

Metacognitive bias parameters (φm, δm, λm, κm).

Gray shades indicate areas of true overconfidence according to the generative model. Gray stripes areas indicate additional areas that would be classified as overconfidence in conventional analyses …

Metacognitive noise.

Considered noise distributions are either censored, truncated or naturally bounded. In case of censoring, protruding probability mass accumulates at the bounds (depicted as bars with a darker shade; …

Figure 6 with 2 supplements
Comparison of Mratio and metacognitive noise σm.

Different performance levels were induced by varying the sensory noise of the forward model. Five different levels of metacognitive noise were simulated for a truncated normal noise distribution, …

Figure 6—figure supplement 1
Comparison of Mratio and metacognitive noise σm for constant stimuli.

This simulation mirrors the simulations in Figure 6 but is based on only a single stimulus intensity level for both stimulus categories. While parameter recovery improves for the noisy-readout model …

Figure 6—figure supplement 2
Type 1 dependency of Mratio and metacognitive noise σm for various settings of other parameters.

This simulation mirrors the simulations in Figure 6, while varying settings for other parameters (as indicated in the title for each column). Changed parameters: sensory threshold ϑs, sensory bias δs

Figure 7 with 6 supplements
Parameter recovery (500 trials per observer).

Linear dependency between generative parameters and fitted parameters for the six parameters of the noisy-report and noisy-readout model (σs, ϑs, δs, σm, φm, δm). Linear dependency between …

Figure 7—figure supplement 1
The figure mirrors the parameter recovery analysis in Figure 7 with 10,000 instead of 500 trials.

Sensory parameters: sensory noise σs, sensory threshold ϑs, sensory bias δs. Metacognitive parameters: metacognitive noise σm, multiplicative evidence bias φm, additive evidence bias δm.

Figure 7—figure supplement 2
The figure mirrors the parameter recovery analysis in Figure 7 for a model with metacognitive confidence biases (λm, κm) instead of metacognitive evidence biases and for either 500 or 10,000 trials.

Note that metacognitive confidence biases are incompatible with a noisy-readout model and hence this combination was omitted. Sensory parameters: sensory noise σs, sensory threshold ϑs, sensory bias …

Figure 7—figure supplement 3
Parameter recovery for a mix of evidence-related and confidence-related metacognitive bias parameters.

For simplicity and clarity, this figure shows only slope matrices for intermediate levels of sensory (σs = 0.7) and metacognitive (σm = 0.2) noise, and for 10,000 trials. Sensory parameters: sensory …

Figure 7—figure supplement 4
No indication of biases in parameter recovery.

For these analyses, the sample size was fixed to 10,000 trials. Sensory parameters: sensory noise σs, sensory threshold ϑs, sensory bias δs. At the metacognitive level the model was specified either …

Figure 7—figure supplement 5
Parameter recovery across a range of trial numbers (500 to 10,000).

Sensory parameters: sensory noise σs, sensory threshold ϑs, sensory bias δs. At the metacognitive level the model was specified either with evidence-related (middle row) or confidence-related …

Figure 7—figure supplement 6
Model recovery.

Data were generated for noisy-readout and noisy-report models with different settings for sensory noise (σs) and metacognitive noise (σm). Model recovery was quantified by the frequency/probability …

Figure 8 with 1 supplement
Application of the model to empirical data from Shekhar and Rahnev, 2021 (N=20).

(A) Posterior probability (choice probability for S+) as a function of normalized signed stimulus intensity. Model-based predictions closely follow the empirical data. Means and standard errors …

Figure 8—figure supplement 1
Empirical confidence distributions and generative models of all 20 subjects in Shekhar and Rahnev, 2021.

Empirical confidence distributions are depicted as gray histograms. Distributions of generative models are depicted as orange line plots for the winning model at the group level …

Visualization of a model fit for a single participant from Shekhar and Rahnev, 2021.

The applied model was a noisy-report model with a metacognitive noise distribution of the type truncated Gumbel and metacognitive evidence biases Each stimulus category in Shekhar and Rahnev, 2021

Tables

Appendix 2—table 1
Metacognitive noise distributions.

All distributions are parameterized such that z* is the mode and σm is the standard deviation of the distribution (the only exception is the Beta distribution, for which σm is a spread parameter …

Noisy-readoutNoisy-report
Censored
Normal
fm(z)={Φ(zσm)  if  z=0 1σmϕ(zzσm)   if  z>0fm(c)={Φ(cσm)   if  c=01σmϕ(ccσm)   if  0<c<1Φ(1cσm)   if  c=1
Censored
Gumbel
fm(z)={1exp(eηmz)   if  z=0ηmexp(ηm(zz)eηm(zz))   if  z>0fm(c)={1exp(eηmc)   if  c=0ηmexp(ηm(cc)eηm(cc))   if  0<c<1exp(eηm(c1))   if  c=1
Truncated
Normal
fm(z)=1σmϕ(zzσm)1Φ(zσm)fm(c)=1σmϕ(ccσm)Φ(1cσm)Φ(cσm)
Truncated
Gumbel
fm(z)=ηmexp(ηm(zz)eηm(zz))exp(eηmz)fm(c)=ηmexp(ηm(cc)eηm(cc))exp(eηmc)exp(eηm(c1))
Gamma/
Beta
fm(z)=βαΓ(α)zα1eβz
Parameterization:
α=z*2+2σm2+z*z*+4σm2
β=12σm2(z+z+4σm2)
fm(c)=cα1(1c)β1B(α,β)
Parameterization:
α=c(1σm2)+1
β=(1c)(1σm2)+1
Lognormalfm(z)=1σ^mπexp(Inzz^σ^m)2
Note: z^ and σ^m represent an analytic parameterization such that the lognormal distribution has mode z* and standard deviation σm. See the published code for details.

Additional files

Download links