Search

The current research proposes a novel approach to identity-based choice that focuses on consumers' representations of the self-concept, as captured by the perceived cause–effect relationships among features of an individual consumer's self-concept. More specifically, the studies reported here test the proposal that the causal centrality of an identity—the number of other features of a consumer's self-concept that the consumer believes influenced or were influenced by the identity—underlies identity importance and is a determinant of identity-based consumer behaviors. Across seven studies, using both measured and manipulated causal centrality, the current research provides evidence for the role of causal centrality in identity-based choice. Among consumers who share an identity (belong to the same social category), those who believe that the identity is more causally central perceive the identity as more important and are more likely to engage in behaviors consistent with the norms of the social category. [ABSTRACT FROM AUTHOR]

Null hypothesis significance testing (NHST) is the default approach to statistical analysis and reporting in marketing and the biomedical and social sciences more broadly. Despite its default role, NHST has long been criticized by both statisticians and applied researchers, including those within marketing. Therefore, the authors propose a major transition in statistical analysis and reporting. Specifically, they propose moving beyond binary: abandoning NHST as the default approach to statistical analysis and reporting. To facilitate this, they briefly review some of the principal problems associated with NHST. They next discuss some principles that they believe should underlie statistical analysis and reporting. They then use these principles to motivate some guidelines for statistical analysis and reporting. They next provide some examples that illustrate statistical analysis and reporting that adheres to their principles and guidelines. They conclude with a brief discussion. [ABSTRACT FROM AUTHOR]

This article is a response to a continuing discussion on the original article by Wood, Bakker, and Fisher “Back to the Future: A Time Calibrated Theory of Entrepreneurial Action” and a commentary on it by Professors Van Lent, Hunt, and Lerner. This article, written by the authors of the original article, comments on Van Lent, Hunt and Lerner’s ideas concerning entrepreneurial action theory.

In real-world marketplaces, one may encounter an alternative that is inferior to another one in the assortment. While the presence of such seemingly irrelevant inferior alternatives should ostensibly have no influence on consumers' decisions, an extensive literature using stylized lab experiments has found that, surprisingly, their presence matters. Specifically, the dominance effect suggests that the presence of an inferior alternative shifts consumer's preferences toward the alternative made to be superior. However, null results in some recent lab studies and a lack of real-world evidence call into question whether, when, and how the effect exists. In this work, we find clear evidence that dominance matters in the wild. We also identify an important moderator for the dominance effect—preference uncertainty—and test it in both a real-world marketplace for digital freelance services and a lab experiment. Further, we find evidence for additional moderators that help explain how the effect works, including the count of dominated alternatives and the magnitude of dominance, consistent with a perceptual mechanism. This work is the first to use consequential field data to shed light on when and why the dominance effects occur, with implications for marketers, choice architects, user interface designers, and policymakers. [ABSTRACT FROM AUTHOR]

Many human behavioral and brain imaging studies have used narratively structured stimuli (e.g., written, audio, or audiovisual stories) to better emulate real-world experience in the laboratory. However, narratives are a special class of real-world experience, largely defined by their causal connections across time. Much contemporary neuroscience research does not consider this key property. We review behavioral and neuroscientific work that speaks to how causal structure shapes comprehension of and memory for narratives. We further draw connections between this work and reinforcement learning, highlighting how narratives help link causes to outcomes in complex environments. By incorporating the plausibility of causal connections between classes of actions and outcomes, reinforcement learning models may become more ecologically valid, while simultaneously elucidating the value of narratives.

Large language models (LLMs) like OpenAI's GPT-4 (OpenAI et al., 2023), or Google's PaLM (Chowdhery et al., 2022) generate text responses to user-generated text prompts. In contrast to work that evaluates the extent to which model-generated text coheres with linguistic rules (i.e., formal competence) (Chomsky et al., 2023; Piantadosi, 2023), the present symposium discusses the work of cognitive scientists aimed at assessing the extent and manner in which LLMs show effective understanding, reasoning and decision making, capacities associated with human higher cognition (i.e., functional competence) (Binz & Schulz, 2023; Mahowald et al., 2023; Webb et al., 2023). Given both their expertise and their interest in clarifying the nature of human thinking, cognitive scientists are in a unique position both to carefully evaluate LLMs' capacity for thought (Bhatia, 2023; Han et al., 2024; Mitchell, 2023) and to benefit from them as methodological and theoretical tools. This symposium will thus be of interest not only to cognitive scientists concerned with machine intelligence, but also to those looking to incorporate advances in artificial intelligence with their study of human intelligence.

Human learning does not stop at solving a single problem. Instead, we seek new challenges, define new goals, and come up with new ideas. What drives people to disrupt the existing conceptual landscape and create new things? Here, we examine the decision to create new things under different levels of expected returns. We formalize innovation as stochastically recombining existing ideas, where successful and more complex combinations generate higher returns. This formalization allows us to cast innovation-seeking as a Markov decision process, and derive optimal policies under different settings. Data collected through an online behavioral experiment confirm our prediction that people should invest more time and effort in seeking innovations when they know the chances of success are high and the potential new ideas would be rewarding. However, people also deviate from being optimal, both innovating more and less than they should in different settings.

Counterfactual reasoning is at the centre of human daily life and plays a key role in shaping our moral and social judgments. Its effect on moral judgment in adulthood, such as justifying immoral behavior (e.g., “If you had not left your phone on the table, it would not have been stolen.”), has been studied for years. However, we still know very little about when counterfactual reasoning starts to affect humans' moral judgments. To test this, we examined the effect of better and worse counterfactual alternatives on 5-6-year-olds' (N = 91) moral and happiness judgments. We found that children judged social exclusion (e.g., a new kid has to play alone while other children play together) as less morally acceptable after imagining how it could have been better (e.g., the new kid and other children at the school could have played all together), but, contrary to past work with adults, they did not justify it after imagining how it could have been even worse (e.g., the other children could have broken the new kid's toy). However, children's happiness judgments showed the opposite effect: they reported feeling happier about reality after imagining a worse counterfactual alternative compared to children who only thought about what actually happened.Keywords: counterfactuals; moral judgment; children; happiness judgment

Comparison of visibility inferences across congenitally blind and sighted people provides insight into the contribution of first-person sensory experience to intuitive theories. We hypothesized that both groups understand others' visual experiences via an intuitive theory incorporating variables known to influence visual psychophysics (distance, looking duration, and feature size). Adults born blind (n=20) and sighted (n=40) listened to short scenarios that described an observer looking at another person from different distances and for varying durations. Participants rated how likely the observer would perceive appearance features of the person that varied in size (e.g., eye color vs. hat). A probabilistic formalization of intuitive visibility fit the ratings with high accuracy across scenarios and features. Model parameters were qualitatively identical across groups but blind adults weighted distance and size less. A quantitative and generative intuitive theory of vision develops without first-person sensory access, possibly through linguistic communication, and is fine-tuned by visual experience.

As AI systems like language models are increasingly integrated into decision-making processes affecting people's lives, it's critical to ensure that these systems have sound moral reasoning. To test whether they do, we need to develop systematic evaluations. We provide a framework that uses a language model to translate causal graphs that capture key aspects of moral dilemmas into prompt templates. With this framework, we procedurally generated a large and diverse set of moral dilemmas---the OffTheRails benchmark---consisting of 50 scenarios and 400 unique test items. We collected moral permissibility and intention judgments from human participants for a subset of our items and compared these judgments to those from two language models (GPT-4 and Claude-2) across eight conditions. We find that moral dilemmas in which the harm is a necessary means (as compared to a side effect) resulted in lower permissibility and higher intention ratings for both participants and language models. The same pattern was observed for evitable versus inevitable harmful outcomes. However, there was no clear effect of whether the harm resulted from an agent's action versus from having omitted to act. We discuss limitations of our prompt generation pipeline and opportunities for improving scenarios to increase the strength of experimental effects.

Functional relations are prevalent in everyday life and science. Do children have intuitive knowledge of functional relations, and can they learn these relations by active information gathering (i.e., choosing a few input values and observing the corresponding outputs)? We found that 6- to 9-year-olds can learn different families of functions (linear, Gaussian, and exponential) through both informative data provided by an experimenter and data they gather from the environment for themselves. Overall, children learn linear functions more accurately than non-linear functions. When choosing data points to learn about, some children select highly similar points that only shed light on a narrow region of a function, while others choose more variable inputs and gain a more holistic view of a function. Children who use this latter, globally informative strategy have higher learning accuracy, particularly for non-linear functions. Results suggest that children are in the process of developing effective strategies for active function learning.

Philosophers have long been arguing the precise semantics of different deontic terms within normative statements. However, little research has been done on the human reasoning side of understanding such terms. In this paper, we propose a normative scheme with bitstring semantics that is expressive enough to cover the basic normative concepts in most mainstream schemes proposed in deontic logic research. Even though further confirmation is needed, our explorative experiments on human deontic reasoning have shown results that are consistent with our proposed scheme.

When reasoning about a claim, it makes sense to be more persuaded if lots of other people agree. But, there are many factors that make weighing the evidence behind a consensus complicated. For example, a consensus might be more or less informative depending on the type of claim, or whether each consensus member formed their opinions independently. These factors might also influence people differently depending on their own assumptions or preferences. In this study we used a mock social media paradigm to assess how persuaded people were by two factors: the presence of consensus (no consensus vs. consensus), and source independence (a consensus based on independent information sources vs. a consensus formed off shared, dependent sources). We varied these factors at both the group and individual level. At the group level, we assessed a third factor: whether people were influenced by the type of claim being reasoned about (we assessed 60 different claims divided into 4 categories). Almost everyone was more persuaded by consensus trials compared to no consensus trials. However, the strength of this effect was credibly stronger if the claim was likely to have a ground truth. We found that around one third of participants were sensitive to source independence. Of these, three quarters were more persuaded by a consensus based on independent sources, but the quarter who were more persuaded by dependent sources were persuaded just as strongly.

The Abstraction Reasoning Corpus (ARC) is a visual analogical reasoning test designed for humans and machines (Chollet, 2019). We compared human and large language model (LLM) performance on a new child-friendly set of ARC items. Results show that both children and adults outperform most LLMs on these tasks. Error analysis revealed a similar "fallback" solution strategy in LLMs and young children, where part of the analogy is simply copied. In addition, we found two other error types, one based on seemingly grasping key concepts (e.g., Inside-Outside) and the other based on simple combinations of analogy input matrices. On the whole, "concept" errors were more common in humans, and "matrix" errors were more common in LLMs. This study sheds new light on LLM reasoning ability and the extent to which we can use error analyses and comparisons with human development to understand how LLMs solve visual analogies.

Disparate impact rules are formally neutral but indirectly discriminate against protected groups (i.e., by targeting a characteristic that is more prevalent in a given group). Because these rules are not obviously malicious, they have been widely enacted to circumvent policies against explicit discrimination. In a series of four experiments, we show that adults and children are sensitive to the moral implications of disparate impact rules. However, we also find that they are more accepting of these rules when strong justification is provided, compared to rules with no justification. Crucially, demographic differences also impact people's judgments of disparate impact rules and their creators. We find that conservatives and those from groups not directly affected by the rule tend to be more accepting of it. By studying people's reasoning about disparate impact rules, this work aims to identify the mechanisms by which these rules may evade detection. Finally, we discuss how these insights may inform the development of interventions that highlight the problematic effects of indirectly discriminatory policies.

While Vector Symbolic Architectures (VSAs) are promising for modelling spatial cognition, their application is currently limited to artificially generated images and simple spatial queries. We propose VSA4VQA – a novel 4D implementation of VSAs that implements a mental representation of natural images for the challenging task of Visual Question Answering (VQA). VSA4VQA is the first model to scale a VSA to complex spatial queries. Our method is based on the Semantic Pointer Architecture (SPA) to encode objects in a hyperdimensional vector space. To encode natural images, we extend the SPA to include dimensions for object's width and height in addition to their spatial location. To perform spatial queries we further introduce learned spatial query masks and integrate a pre-trained vision-language model for answering attribute-related questions. We evaluate our method on the GQA benchmark dataset and show that it can effectively encode natural images, achieving competitive performance to state-of-the-art deep learning methods for zero-shot VQA.

There has been exciting recent progress in computational modeling of moral cognition. Work in this area tends to describe the cognitive mechanisms of human moral judgment using symbolic models, which are interpretable and written in terms of representations that carry meaning. However, these models fall short of capturing the full human capacity to make moral judgments in that they fail to process natural language inputs but instead rely on formal problem specifications. The inability to interface with natural language also limits the usefulness of symbolic models. Meanwhile, there have been steady advances in conversational AI systems built using large language models (LLMs) that interface with natural language. However, these systems fall short as models of human reasoning, particularly in the domain of morality. In this paper we explore the possibility of building neuro-symbolic models of human moral cognition that use the strengths of LLMs to interface with natural language (specifically, to extract morally relevant features from it) and the strengths of symbolic approaches to reason over representations. Our goal is to construct a model of human moral cognition that interfaces with natural language, predicts human moral judgment with high accuracy, and does so in a way that is transparent and interpretable.

People who have no experience with programming can create informal programs to rearrange the order of cars in trains. To find out whether they rely on kinematic mental simulations, the current studies examined participants' eye movements in two experiments in which participants performed various moves and rearrangements on a railway consisting of a main track running from left to right and a siding entered from and exited to the left track. In Experiment 1, they had to imagine different sorts of single moves of cars on the railway. The sequences of their fixations resembled iconic gestures: they tended to look at the starting location of the imagined move, and then at its final location. In Experiment 2, the task was to create descriptions of how to solve four sorts of rearrangements that differ in their Kolmogorov complexity. It predicted the time to find the correct solution and the relative number and duration of fixations recorded during the description of each move for rearrangements of different complexity. Participants were more likely to fixate on the symbols on the cars than anything else, and they fixated longer when the rearrangement was more difficult. They also tended to fixate regions of the tracks where a car's movement began or ended, as if they were imagining a car moving along the tracks. The results suggest that humans rely on a kinematic mental simulation when creating informal algorithms.

Social media is often used as a platform where individuals engage in debate regarding topics that are important to them. Not all arguments are equally convincing, and whilst a given argument may be persuasive to some people, it is often seen as inadequate by others. We are interested in both the individual and argument level differences that make ‘everyday' arguments such as those on social media persuasive. In a replication of our Everyday Belief Bias Task (Deans-Browne & Singmann, 2024), we investigate this question using a paradigm that consists of two parts. In the first part, we measure participant's individual beliefs about eight claims each referring to a political topic (e.g., Abortion should be legal). In the second part, participants rated an argument for each of these claims that was deemed as either good, inconsistent (containing internal inconsistencies), or authority-based (being centered around appeals to authority). We replicated the belief consistency effect – participants preferred arguments that were also in line with their beliefs. We also found that authority-based arguments were rated as worse than inconsistent arguments, and that both types of arguments were rated as worse than good arguments. The implications are first that people do not evaluate arguments independently of the background beliefs held about them. Secondly, people are willing to ignore inconsistencies in arguments more than they are willing to accept the endorsement of authority figures as adequate evidence for arguments.

Decisions about extinction risks are ubiquitous in everyday life and for our continued existence as a species. We introduce a new risky-choice task that can be used to study this topic: The Extinction Gambling Task. Here, we investigate two versions of this task: a Keep variant, where participants cannot accumulate any more earnings after the extinction event, and a Lose variant, where extinction also wipes out all previous earnings. We derive optimal solutions for both variants and compare them to behavioural data. Our findings suggest that people understand the difference between the two variants and their behaviour is qualitatively in line with the optimal solution. Further, we find evidence for risk-aversion in the Keep condition but not in the Lose condition. We hope that this task can facilitate further research on this vital topic.

In term series problems where multiple mental models can be constructed, partial-order models can be created as mental representations, which make it easier to perceive the symmetry of the terms. To test these hypotheses, we categorized multi-model (indeterminate) term series problems according to the patterns of partial-order models that could be constructed, and analyzed the reasoning performance for each pattern. These results suggest that reasoners tend to use the symmetry of terms to reduce the cognitive load of reasoning. Analysis of the patterns of incorrect answers also suggests that attempts to exploit the symmetry of the term may be biased, leading to errors in reasoning.

Prospect Theory has been highly influential; however its experimental paradigm lacks higher orders of uncertainty. To introduce this, participants are asked to imagine themselves facing a choice between two bags containing 100,000 blue or red balls in unknown proportions. A red ball wins £500. Participants are shown samples from each bag; e.g., 5 balls from Bag 1 (3 red) and 100 balls from Bag 2 (55 red). The bags can be represented by distributions with Bag 1 having a higher mean probability estimate (60% vs 55%), but more variance (second order uncertainty) in that estimate. By varying observed frequencies and gain vs loss formats, we seek to determine if classic findings remain when higher order uncertainties are present. Results consistent with the four-fold pattern are seen for gains (uncertainty seeking at low probability values, uncertainty aversion at high probability values) but for losses, uncertainty aversion is seen at all values.

Independent pieces of corroborating evidence should provide stronger support to a hypothesis than dependent pieces of evidence. Overlooking the inferiority of dependent relative to independent items of evidence can lead to a chain reaction of double-counting evidence, over-estimating the probability that the fact under consideration is true, and making wrongful decisions. Within fictitious scenarios, we investigate people's sensitivity to the independency advantage. We assess their ability to integrate multiple items of evidence that come from (in)dependent sources who differ in reliability. We find that participants properly perceive dependencies when explicitly asked but fail to distinguish the probative value of dependent versus independent evidence in their belief updating. Still, individuals who perceive a strong dependence between sources treat the evidence as being more redundant. We find no dependency-related effects on participants' individual Bayesian network model predictions. Potential reasons why participants perceive (in)dependencies and yet (mostly) fail to discount for them are discussed.

When a human and an AI agent collaborate to complete a task and something goes wrong, who is responsible? Prior work has developed theories to describe how people assign responsibility to individuals in teams. However, there has been little work studying the cognitive processes that underlie responsibility judgments in human-AI collaborations, especially for tasks comprising a sequence of interdependent actions. In this work, we take a step towards filling this gap. Using semi-autonomous driving as a paradigm, we develop an environment that simulates stylized cases of human-AI collaboration using a generative model of agent behavior. We propose a model of responsibility that considers how unexpected an agent's action was, and what would have happened had they acted differently. We test the model's predictions empirically and find that in addition to action expectations and counterfactual considerations, participants' responsibility judgments are also affected by how much each agent actually contributed to the outcome.

Question asking is a key tool for learning, especially in childhood. However, formulating good questions is challenging. In any given situation, many questions are possible but only few are informative. In the present work, we investigate two ways 5- to 10-year-olds and adults simplify the challenge of formulating questions: by reusing previous questions, and by remixing components of previous questions to form new questions. Our experimental results suggest that children and adults reuse and remix questions and adaptively modulate reuse depending on how informative a question will be in a particular situation. This work shows that task-relevant experience asking questions provides fodder for future questions, simplifying the challenge of inquiry and enabling effective learning.

We provide a generalized, normative model of visual detection that accounts for key asymmetries between decisions about presence and about absence. In our model, decisions about presence are made based on the visibility of presented stimuli, but decisions about absence are made based on counterfactual visibility: beliefs about the degree to which a stimulus would have been visible if present. Behavioral patterns in visual detection experiments under different levels of partial occlusion validate key model predictions. Specifically, we find that unlike decisions about presence, the confidence and speed of decisions about absence are largely independent of perceptual evidence, but are sensitive to the counterfactual visibility of absent stimuli. Finally, we reveal robust individual differences in counterfactual perception, with some participants systematically incorporating counterfactual visibility into perceptual decisions in a different fashion from others. We discuss implications for the varied and inferential nature of visual perception more broadly.

Building on counterfactual theories of causal-selection, according to which humans intuitively evaluate the causal responsibility of events, we developed an experimental paradigm to examine the effect of causal-selection explanations on abductive causal inference. In our experiment, participants attempted to infer the rule responsible for winning outcomes of random draws from urns with varying sampling probabilities.Participants who were provided with causal-selection judgments as explanations for the outcomes made significantly closer inferences to the rule than those relying on observations alone, or on other explanations of causal relevance.We mirror these empirical results with a computational model of inference from explanation leveraging the theories of causal selection.

We propose that kinds relate to particular things either constitutively or contingently. Taxonomic categories of animals and artifacts constitutively relate their members: DOG and CAR group things by aspects of the forms of their matter; the forms that make them things instead of stuff. Categories of things in roles or with diseases contingently relate to their members: LAWYER and DIABETIC group things by forms other than the forms that make them things. We confirm this distinction in five experiments with American adults.

When children use online apps, they often share personal information, such as their name, address, and birthday. In the present study, we investigated the mental models children use to reason about what apps are allowed to do with personal data after it has been willingly shared with an app. 57 children ages 8- to 11-years-old were read a story in which they were asked to judge whether an online game (app) was allowed or not allowed to perform four different actions: looking, saving, selling, and showing. We compared these judgments to a comparison condition where we asked children what users themselves should be allowed to do with their data. We found that children viewed the app as less permitted to act on the data than users as well as some further differences by action-type. Our findings suggest that children use something akin to a “lending” model to conceptualize data transfers, in which apps have less rights than users despite the data being willingly transferred to the app. Our findings also suggest that children differentiate among the uses of information as children think certain actions by the app are less permissible than others (e.g., looking is more permissible than selling).

Recent work shows that participants make asymmetric causal inferences from apparently symmetric correlational statements (e.g., “A is associated with B”). Can we make sense of this behavior in terms of rational language use? Experiment 1 investigates these interpretive preferences—what we call “PACE effects”—in light of theoretical and experimental pragmatics and psycholinguistics. We uncover several linguistic factors that influence them, suggesting that a pragmatic explanation is possible. Yet, since PACE effects do not show that correlational language leads to causal implicatures strong enough to influence action choice in practical decision contexts, Experiment 2 offers new evidence from an experiment that explicitly compares the effects of causal vs. correlational claims on decision-making. Our results suggest that causal inferences from correlation language are an intricate, but possibly

The present research evaluates a coherence-based network approach to moral judgement. Under this view, judgement is an outcome of achieving coherence between a network of causally interacting beliefs. Consistent with this, despite similar initial views, participants re-evaluated their beliefs and attitudes in support of their judgement, driving polarisation between individuals reporting competing judgements. Different properties of the dynamic network structure determined metacognitive properties of judgement such as confidence and perceived task difficulty. Whilst the judgement formation process involves revising beliefs and values to achieve a coherent arrangement, the nature of the judgement reached depends on the aggregate weight of these beliefs once the revision process is completed.

Social norms are a hallmark of human social intelligence, yet the reasoning processes involved in norm formation have been difficult to capture with traditional modeling frameworks. We developed a computational model of norm formation as joint planning via theory-of-mind. The model is designed to capture the distinctively human ability to flexibly develop more complex norms in more complex situations, via simulation of joint decision-making with other agents over an extended time horizon. We evaluated the predictions of the model against participant interactions in a 2-player iterated decision-making task. Across 3 conditions our model captured the way participants balanced joint reward, fairness, and complexity when forming norms.

Recent research has provided experimental support for a new ``Inferentialist'' theory of conditionals, challenging the Equation P(If A, C) = P(C | A) and theories that support it. The key evidence comes from probability judgments involving conditionals whose antecedent and consequent are relevant vs. irrelevant to each other. Expanding on recent experimental work, we argue that Inferentialism has difficulty explaining the data. However, theories that support The Equation theory are well-placed to account for the results once we recognize an independent phenomenon of pragmatic intrusion on probability judgment - in this case, participants' tendency to assign lower probability to conditionals that are pragmatically incoherent.

According to a common view in epistemology, a set of propositions is justified if it is coherent. Similarly, a new proposition should be accepted if it is coherent with the accepted body of beliefs. But what is coherence? And what, in turn, justifies the above claims? To answer these questions, various Bayesian measures of epistemic coherence have been proposed. Most of these measures are based on the prior probability distribution over the corresponding propositional variables. We criticize this ``static'' conceptualization of coherence and propose instead that the coherence of an information set is related to how well the information set responds when each of the propositions it contains is confirmed by an independent and partially reliable information source. The elaboration of this idea will show that the proposed ``dynamic'' perspective has several advantages and solves some open problems of coherentist epistemology. It also has implications for our understanding of reasoning and argumentation in science and beyond.

It is uncontroversial that humans can represent possibilities, but it is debated what this claim amounts to. Under broad views of modal cognition, many representational and reasoning systems represent possibilities at multiple levels of cognitive architecture. Under narrow views of modal cognition, there exists a special kind of higher-level modal thought, that can be measured with purpose built non-verbal modal cognition tasks. Here we ask whether object tracking mechanisms that are assumed to lack the higher-level narrow modal capacity, show behavioral signatures that are assumed to require it. We find signature of modal representation in one task, but not another. The finding suggests that there is no clear difference between tasks that tap broad and narrow modal cognition, and invites a reassessment of the evidence for the latter.

When do children understand that number words (such as ‘five') refer to exact quantities and that the same number word can be used to label two sets whose items correspond 1-to-1 (e.g., if each bunny has a hat, and there are five hats, then there are five bunnies)? Two studies with English-speaking 2- to 5-year-olds revealed that children who could accurately count large sets (CP knowers) were able to infer that sets exhibiting 1-to-1 correspondence share the same number word, but not children who could not accurately count large sets (subset knowers). However, not all CP knowers made this inference, suggesting that learning to construct and label large sets is a critical but insufficient step in discovering that numbers represent exact quantities. CP knowers also failed to identify 1-to-1 corresponding sets when faced with sets that had an off-by-one difference, suggesting that children who could accurately count large sets used approximate magnitude to establish set equality, rather than 1-to-1 correspondence. These results suggest that children's initial intuitions about numerical and set equality are based on approximation, not 1-to-1 correspondence, and that this occurs well after they have learned to count and construct large sets.

Understanding referring expression generation has long been of interest to psycholinguistics, pragmatics, and sociolinguistics. Experimental data in the former two has shown that referring expression generation is modulated by both pragmatic and cognitive considerations, and the latter suggests that referring expressions have social meaning beyond their literal referential utility. This project integrates these three accounts by extending Burnett (2017)'s socially-enriched implementation of the Rational Speech Act (RSA) framework to account for variation in referring expressions used to denote transgender women in two politically opposed media corpora. Our findings highlight the utility of the RSA framework in explaining socially-modulated variation while also accounting for pragmatic and cognitive considerations. Finally, this paper contributes to growing literatures that address the relationship between (alt-)right ideologies about gender and language by highlighting the use of bioessentialist language such as 'biological male' in the propagation of anti-trans rhetoric in the United States.

We can evaluate features of problems and their potential solutions well before we can effectively solve them. When considering a game we have never played, for instance, we might infer whether it is likely to be challenging, fair, or fun simply from hearing the game rules, prior to deciding whether to invest time in learning the game or trying to play it well. Many studies of game play have focused on optimality and expertise, characterizing how people and computational models play based on moderate to extensive search and after playing a game dozens (if not thousands or millions) of times. Here, we study how people reason about a range of simple but novel connect-n style board games. We ask people to judge how fair and how fun the games are from very little experience: just thinking about the game for a minute or so, before they have ever actually played with anyone else, and we propose a resource-limited model that captures their judgments using only a small number of partial game simulations and almost no lookahead search. For more information about this project, see https://sites.google.com/view/intuitive-game-theory

Most current research on commonsense question answering (CQA) has focused on proposing different techniques in natural language processing and text information retrieval. However, for human cognition, retrieving and organizing desired answers from text knowledge related to commonsense questions is far less intuitive and comprehensive than it is when using multi-modal knowledge, such as related images and videos. Motivated by this, we propose a framework for trying the acquisition of diverse modal information, and embedding and integrating it into CQA tasks, further improving the performance and user experience. Specifically, this paper proposes the integration of multi-modal knowledge, including images, image description statements, image scene graphs, and knowledge sub-graphs, into a CQA system. It introduces a parallel embedding technique for this multi-modal knowledge and employs an alignment-interaction-fusion mechanism to facilitate the seamless integration of this multi-modal knowledge. Through extensive experiments, the effectiveness and superiority of our proposed method are demonstrated.

One of the core properties of human language is compositionality: the meaning of a sentence can be understood by the meaning of individual words and the rules for combining them (Szabó, 2020). We investigate the development of conceptual compositionality (the combination of concepts). In our study, 6- to 9-year-old children (N = 40) were shown a card with two objects (e.g., a car and a star). Participants were introduced to two characters (a robot and a wizard) that used their powers to change the objects in different ways (e.g., turning one object pink). In the test trials, participants were asked to predict what a card would look like after both characters used their powers on the same card. All participants successfully learned the characters' powers, but only participants 7.5 years and older succeeded in the compositionality test trials. Our findings suggest that by age 7.5 children can successfully compose functions.

This paper seeks to understand designers' abstraction in ill- structured problem-solving. We utilize a protocol study with expert designers to empirically analyze the abstraction process in the latent need problem setting. A logic-based abstraction schema is found to model the process the designers employed. The study reveals how designers utilize this schema, detailing, developing, and evaluating solutions for ill-structured problems. It highlights the recursive nature of abstraction and raises questions about the termination of the process in ill- structured domains. We conclude by proposing a computational model to further evaluate abstraction in complex problem-solving scenarios.

How do people perform general-purpose physical reasoning across a variety of scenarios in everyday life? Across two studies with seven different physical scenarios, we asked participants to predict whether or where two objects will make contact. People achieved high accuracy and were highly consistent with each other in their predictions. We hypothesize that this robust generalization is a consequence of mental simulations of noisy physics. We designed an "intuitive physics engine'' model to capture this generalizable simulation. We find that this model generalized in human-like ways to unseen stimuli and to a different query of predictions. We evaluated several state-of-the-art deep learning and scene feature models on the same task and found that they could not explain human predictions as well. This study provides evidence that human's robust generalization in physics predictions are supported by a probabilistic simulation model, and suggests the need for structure in learned dynamics models.

Autonomous agents often need to decide between choosing actions that are familiar and have previously yielded positive results (exploitation) and seeking new information that could help uncover more effective actions (exploration). We present an “observe or bet” task that separates “pure exploration” from “pure exploitation”: 75 five-to-seven-year-old children, 60 adults and computational agents have to decide either to observe an outcome without reward, or to bet on an action without immediate feedback at varying probability levels. Their performances were measured against solutions from the partially observable Markov decision process and meta-RL models. Children and adults tended to choose observation more than both algorithm classes would suggest. Children also modulated their betting policy based on the probability structure and amount of evidence, exhibiting “hedging behavior” a strategy not evident in standard bandit tasks. The results provide a benchmark for reasoning about reward and information in humans and neural network models.

There is wide agreement that the mind has different mechanisms it can use to make moral judgments. But how does it decide which one to use when? Recent theoretical work has suggested that people select mechanisms of moral judgment in a way that is resource-rational --- that is, by rationally trading off effort against utility. For instance, people may follow general rules in low-stakes situations, but engage more computationally intensive mechanisms such as consequentialist or contractualist reasoning when the stakes are high. Here, we evaluate whether humans and large language models (LLMs) exhibit resource-rational moral reasoning in two moral dilemmas by manipulating the stakes of each scenario. As predicted, we found that the higher the stakes, the more people employed a more effortful mechanism over following a general rule. However, there was mixed evidence for similar resource-rational moral reasoning in the LLMs. Our results provide evidence that people's moral judgments reflect resource-rational cognitive constraints, and they highlight the opportunities for developing AI systems better aligned with human moral values.

Why do beliefs that gender differences are innate (i.e., gender essentialism) sometimes lead to normative judgments about how individual people ought to be? In the current study, we propose that a missing premise linking gender essentialism and normativity rests on the common folk-biological assumption that biological features serve a biological function. When participants (N = 289) learned that a novel feature of the gender category “mothers” was common and innate, they overwhelmingly assumed that it must have served some function across human history. When they learned that it served a historical function, they assumed that it must still be beneficial in today's environment. When participants learned that the feature was beneficial, they judged that contemporary mothers ought to have it, and they were more willing to intervene to ensure that they would by constraining the choices of individual mothers. Thus, we suggest that essentialist assumptions can shape normative social judgments via the explanations people tend to generate about why certain features of natural kind categories become common to begin with. This finding articulates one manifestation of the naturalistic fallacy, with implications for policy debates about bodily autonomy and choice.

What does a social learner learn? Research has explored imitation-based social learning strategies as well as inverse reinforcement algorithms that estimate others' true reward function. In the current study, we propose that social learning may be more elaborate and develop a model of social learning using Bayesian inference that seeks to understand both the task an observed demonstrator is performing and the demonstrator itself. Using simulations, we show that the model is able to learn about the demonstrator when provided with full and partial information. We strengthen this point by asking the model to make inferences about missing choice and reward information. Last, we show that the model is able to represent one set of beliefs about the environment while attributing a distinct set of beliefs to the demonstrator. Thus, we move away from simple models of social learning, investigating inference-making as a core mechanism of social learning.

In syllogistic reasoning research, humans are predominantly evaluated on their capabilities to judge whether a conclusion necessarily follows from a set of premises. To tackle this limitation, we build on work by Evans, Handley, Harper, and Johnson-Laird (1999), and present two studies where we asked participants for possible and likely conclusions. Combined with previous data (containing necessary), we present a comprehensive dataset with responses for all syllogisms, offering individual patterns for all three argument types - a first of its kind. We discovered that likely serves as a middle ground between possible and necessary, paving the way to further investigate biases and preferences. Generally, individuals were able to handle the different notions, yet tended to interpret quantifiers in a pragmatic way, overlooking logical implicatures. Finally, we tested mReasoner, an implementation of the Mental Model Theory, and concluded that it was not able to capture the patterns observed in our data.

Humans possess a remarkable capacity to recognize and manipulate abstract structure, which is especially apparent in the domain of geometry. Recent research in cognitive science suggests neural networks do not share this capacity, concluding that human geometric abilities come from discrete symbolic structure in human mental representations. However, progress in artificial intelligence (AI) suggests that neural networks begin to demonstrate more human-like reasoning after scaling up standard architectures in both model size and amount of training data. In this study, we revisit empirical results in cognitive science on geometric visual processing and identify three key biases in geometric visual processing: a sensitivity towards complexity, regularity, and the perception of parts and relations. We test tasks from the literature that probe these biases in humans and find that large pre-trained neural network models used in AI demonstrate more human-like abstract geometric processing.

The discrete entities and explicit relations of symbolic systems make them transparent and easy to communicate. This contrasts with distributed systems, which tend to be opaque. This can lead us to pursue symbolic characterizations of human cognition. Symbolic interpretations can, however, oversimplify distributed systems. This is demonstrated in the developmental number cognition literature, where recent findings suggest a gradience of counting ability in children's learning. We take inspiration from these findings to explore the meaning of symbols in Recurrent Neural Networks (RNNs). We align recurrent neural representations with number symbols by causally intervening on the neural representations. We find that symbol-like representations of numbers do emerge in RNNs. We use this to inform the discussion on how neural systems represent quantity. We also show that the symbol-like representations evolve with learning, and continue to vary after the RNNs solve the task, demonstrating the graded nature of symbols in distributed systems.

How, and how strongly, do default comprehension inferences shape verbal reasoning? When do they lead to fallacies? We address these questions for reasoning with polysemous verbs (verbs with distinct, but related senses) and ask when their use leads to fallacies of equivocation. The ‘linguistic salience bias hypothesis' specifies conditions where subordinate uses of unbalanced polysemes trigger defeasible default inferences that are supported only by the dominant sense but influence further cognition, regardless. But does this happen even where the verb is preceded by disambiguating context that invites subordinate interpretations from the start? We present three experimental-philosophy studies that address this question: We use the psycholinguistic cancellation paradigm and fixation time measurements to examine inferences from polysemous appearance verbs. We find that default inferences can beat even preceding contextual information. Beyond their psycho-linguistic interest, findings have important philosophical consequences.