Dobromir Rahnev

The paper tests AI models on standardized face recognition tasks and finds most perform worse than humans, with some showing face-blindness similar to human prosopagnosia.

The paper compares bias-aware and bias-blind confidence in humans and machines, showing humans exhibit bias-aware confidence while standard ANNs do not, but this can be achieved with metacognitive modules.

This study shows that individual differences in artificial neural networks mirror human behavioral differences in accuracy, confidence, and response time across tasks.

This study shows that different random initializations of neural networks produce individual differences similar to humans, and using these differences improves predictions of human responses.

The paper shows that multisensory integration in perception is driven by subjective confidence rather than objective accuracy of the sensory cue.

The paper identifies multiple stable brain activation patterns for the same perceptual decision-making task using functional MRI and clustering analysis.

RTNet is a neural network that mimics human perceptual decision-making by generating stochastic decisions and human-like response time distributions, outperforming existing models.

The paper shows that perceptual similarity judgments are linked to an individual's discrimination capacities, with similarity reflecting their ability to discern stimuli.

The paper tests whether threat cues improve perceptual discrimination and finds no improvement despite increased neural activity in the frontoparietal network.

This study shows that convolutional neural networks exhibit confidence-accuracy dissociations similar to humans when stimulus energy is manipulated, suggesting these effects arise from low-level representation changes rather than cognitive heuristics.

The paper shows that a frontoparietal brain network is domain-general for task learning, becoming more connected as learning progresses.

The paper decodes reaction time and confidence across the cortex at the individual level, finding broader brain-behavior relationships than previously observed.

This study investigates the timing of confidence computations in human prefrontal cortex by using TMS to disrupt DLPFC and observe effects on confidence ratings.

The paper quantifies the contribution of subject and group factors in brain activation using a new method applied to a perceptual decision-making task.

The paper identifies trial-to-trial variability in EEG patterns during a perceptual decision task and links these patterns to decision-making parameters.

The paper investigates whether awake reactivation after a brief task depends on task novelty, finding that novel tasks show reactivation while trained tasks show suppression.

The paper investigates neural circuits involved in perceptual decision making and confidence generation using fMRI, finding overlapping and some unique brain regions activated during these processes.

This study investigates suboptimal perceptual decision making in tasks with multiple alternatives and finds that most sensory information is lost during the transformation to decision-level representations.

This study shows that TMS alters multivoxel patterns in the brain without changing overall BOLD activity.

The paper shows that increasing sensory noise improves metacognitive efficiency by manipulating confidence ratings in a visual decision task.

The paper tests the hypothesis that long-term potentiation (LTP) competition causes interference in consecutive visual learning tasks and finds that disrupting LTP in the visual cortex reduces interference.

The paper shows that specific visual features, like Gabor orientation, are reactivated in human V1 after visual training, and this reactivation correlates with performance improvement.

The paper reviews suboptimal perceptual decisions and argues that optimality claims are meaningless without knowing likelihood functions, priors, and cost functions, proposing an LPCD approach instead.

The paper challenges the assumption that perceptual decisions rely on full probability distributions by reviewing evidence that supports simpler representation schemes.