Abstract Deep neural networks (DNNs) have had extraordinary successes in classifying photographic images of objects and are often described as the best models biological vision. This conclusion is largely based on three sets findings: (1) DNNs more accurate than any other model taken from various datasets, (2) do job predicting pattern human errors behavioral (3) brain signals response to datasets (e.g., single cell responses or fMRI data). However, these not test hypotheses regarding what...

Deep neural networks (DNNs) have had extraordinary successes in classifying photographic images of objects and are often described as the best models biological vision. This conclusion is largely based on three sets findings: (1) DNNs more accurate than any other model taken from various datasets, (2) do job predicting pattern human errors behavioral benchmark (3) brain signals response to datasets (e.g., single cell responses or fMRI data). However, most benchmarks report outcomes...

This study investigated whether dual processes theory or motivated reasoning better explains susceptibility to health-related misinformation. By using a working memory load task, we built conditions favoringeither deliberative heuristic while participants judged COVID19 headlines. We also assessed variables (unwarrantedbeliefs, naïve skepticism) and cognitive style measures (bullshit receptivityand reflection). Results showed no difference in correct answersbetween the loaded non-loaded...

People readily generalize knowledge to novel domains and stimuli.We present a theory, instantiated in computational model, based on the idea that cross-domain generalization humans is case of analogical inference over structured (i.e., symbolic) relational representations.The model an extension LISA (Hummel & Holyoak, 1997) DORA (Doumas et al., 2008) models learning.The resulting learns both content format structure) representations from non-relational inputs without supervision, when...

Same-different visual reasoning is a basic skill central to abstract combinatorial thought. This fact has lead neural networks researchers test same-different classification on deep convolutional (DCNNs), which resulted in controversy regarding whether this within the capacity of these models. However, most tests rely testing images that come from same pixel-level distribution as training images, yielding results inconclusive. In study, we tested relational DCNNs. series simulations show...

Researchers studying the correspondences between Deep Neural Networks (DNNs) and humans often give little consideration to severe testing when drawing conclusions from empirical findings, this is impeding progress in building better models of minds. We first detail what we mean by highlight how especially important working with opaque many free parameters that may solve a given task multiple different ways. Second, provide examples researchers making strong claims regarding DNN-human...

On several key issues we agree with the commentators. Perhaps most importantly, everyone seems to that psychology has an important role play in building better models of human vision, and (most) agrees (including us) deep neural networks (DNNs) will modelling vision going forward. But there are also disagreements about what for, how DNN-human correspondences should be evaluated, value alternative approaches, impact marketing hype literature. In our view, these latter contributing many...

Whether neural networks can capture relational knowledge is a matter of long-standing controversy. Recently, some researchers have argued that (1) classic connectionist models handle relational...

Abstract How a system represents information tightly constrains the kinds of problems it can solve. Humans routinely solve that appear to require structured representations stimulus properties and relations. Answering question how we acquire these has central importance in an account human cognition. We propose theory learn invariant responses instances similarity relative magnitude, relational be learned from initially unstructured inputs. instantiate DORA ( Discovery Relations by Analogy )...

Whether neural networks can capture relational knowledge is a matter of long-standing controversy. Recently, some researchers have argued that (1) classic connectionist models handle structure and (2) the success deep learning approaches to natural language processing suggests structured representations are unnecessary model human language. We tested Story Gestalt model, text comprehension, Sequence-to-Sequence with Attention modern architecture for processing. Both were trained answer...

Abstract Same-different visual reasoning is a basic skill central to abstract combinatorial thought. This fact has lead neural networks researchers test same-different classification on deep convolutional (DCNNs), which resulted in controversy regarding whether this within the capacity of these models. However, most tests rely testing images that come from same pixel-level distribution as images, yielding results inconclusive. In study we tested relational DCNNs. series simulations show...

We describe the MindSet benchmark designed to facilitate testing of DNNs against controlled experiments reported in psychology. will focus on a range low-, middle-, and high-level visual findings that provide important constraints for theory, materials DNNs, an example how assess DNN each experiment using ResNet152 pretrained ImageNet. The goal is not evaluate well accounts human vision, but rather, encourage researchers various account key phenomena.

Same-different visual reasoning is a basic skill central to abstract combinatorial thought. This fact has lead neural networks researchers test same-different classification on deep convolutional (DCNNs), which resulted in controversy regarding whether this within the capacity of these models. However, most tests rely testing images that come from same pixel-level distribution as images, yielding results inconclusive. In study we tested relational DCNNs. series simulations show models based...

This study investigated the individual influences of conventionality and designer's intent on function judgments possibly malfunctioning artifacts. Children aged 4 5 years 6 to 8 were presented with stories about an artifact two equally plausible functions, one labeled as either conventional or designed. Subsequently, a character attempted use for cued function, which resulted in malfunction successful use. The children's task was identify real artifact. When attempt succeeded, 4-...

Achieving visual reasoning is a long-term goal of artificial intelligence. In the last decade, several studies have applied deep neural networks (DNNs) to task learning relations from images, with modest results in terms generalization learned. However, recent years, object-centric representation has been put forward as way achieve within framework. Object-centric models attempt model input scenes compositions objects and between them. To this end, these use kinds attention mechanisms...

Multiple benchmarks have been developed to assess the alignment between deep neural networks (DNNs) and human vision. In almost all cases these are observational in sense they composed of behavioural brain responses naturalistic images that not manipulated test hypotheses regarding how DNNs or humans perceive identify objects. Here we introduce toolbox MindSet: Vision, consisting a collection image datasets related scripts designed on 30 psychological findings. experimental conditions,...

Humans readily generalize, applying prior knowledge to novel situations and stimuli. Advances in machine learning artificial intelligence have begun approximate even surpass human performance, but systems reliably struggle generalize information untrained situations. We describe a neural network model that is trained play one video game (Breakout) demonstrates one-shot generalization new (Pong). The generalizes by representations are functionally formally symbolic from training data, without...

On several key issues we agree with the commentators. Perhaps most importantly, everyone seems to that psychology has an important role play in building better models of human vision, and (most) agrees (including us) DNNs will modelling vision going forward. But there are also disagreements about what for, how DNN-human correspondences should be evaluated, value alternative approaches, impact marketing hype literature. In our view, these latter contributing many unjustified claims regarding...