Errors, fast and slow.

Area of Science:

• Cognitive Psychology
• Human Factors
• Computational Neuroscience

Background:

• Human errors in cognitive tasks exhibit variable response times (RTs), being faster or slower than correct responses.
• Existing models can fit observed error RT distributions but struggle with ex ante prediction of error speed.
• Predicting error RT characteristics is crucial for understanding underlying cognitive processes in decision-making.

Purpose of the Study:

• To empirically validate a simple nonparametric model for predicting when human errors are faster or slower than correct responses.
• To assess the model's applicability to generalized conflict tasks and its ability to predict error rate differences.
• To test the hypothesis of process multiplicity in cognitive tasks based on predictive performance.

Main Methods:

• Utilized 20 diverse datasets comprising 31 experiments across various domains.
• Employed a simple nonparametric model to predict error RTs relative to correct response times.
• Analyzed predictions for generalized Stroop effects and error rate differences in conflict tasks.

Main Results:

• The nonparametric model successfully predicted whether errors would be faster or slower than correct responses across datasets.
• Model predictions accurately encompassed generalized Stroop effects and error rate variations.
• The model's predictions were overwhelmingly supported by empirical data, validating its hypotheses.

Conclusions:

• A simple nonparametric model can reliably predict error RT characteristics ex ante in cognitive tasks.
• The model's success in generalized conflict tasks supports the assumption of process multiplicity.
• This approach provides a novel framework for generating and testing hypotheses about cognitive control and decision-making.