Read the paper to see how assumptions about initial-decisions (e.g. collapsing-boundaries) influence post-decision confidence model computations and comparisons, and how we went about simulating a CPP proxy signal (doi.org/10.1101/2025...)

Overall, we found and that post-decision confidence can be explained by a Single accumulation process that continues from initial decision until reaching a post-decision collapsing confidence-boundary, and comparing models' decision-dynamics to the CPP can distinguish behaviourall-similar models.

to get Certainty effects before the initial choice, as we have seen previously (e.g. Grogan et al., 2023).

No model could replicate a surprising result; that the CPP was larger when speed-pressure (short deadline) was applied to the confidence-ratings, suggesting additional mechanisms at play...

Boundary-Single models could replicate the effects of Certainty we saw on the CPP, while Boundary-Distinct could not, especially for pre-choice effects.
Having pre- and post-choice as a Single process allows evidence-info to carry over and inform certainty ratings, which seems to be necessary...

there was little difference between Boundary models where the pre- and post-decision accumulation processes were Distinct or a Single process, when looking at behavioural fits.
However, simulating Decision-Variable accumulation traces allowed us to compare these different mechanisms directly...

on a task with long or short post-decision deadlines to rate confidence, which induced a post-decision/confidence speed-accuracy trade-off.
Post-decision accumulation was better explained by accumulation to collapsing confidence-boundaries, than by a Time-based stopping rule, but...

Different confidence models can give very similar behavioural predictions, making it hard to compare them, but they often make different predictions for the decision dynamics. We simulated Decision Variable traces, and compared them to a neural metric of evidence accumulation, the CPP...