545298 research outputs found
Sort by
Evaluating testimony from multiple witnesses: exploring qualitative intuitions
Get PDFThis study further explored a novel reasoning error. When faced with evidence from multiple sources, a substantial number of lay reasoners inaccurately integrate cues of reliability and report number. Particularly when further reports are less reliable than initial (highly reliable) reports. When evaluating the added value of supplementary corroborative reports, we find that, in most instances, participants are equally likely to provide correct or incorrect qualitative judgements. When using a sequential presentation and explicitly prompting participants to consider the impact of additional credible evidence, 36.7%-45% indicate that their beliefs should remain the same and 10% or less indicate that their beliefs should decrease. Only a third correctly believed that in each instance of corroborating evidence the likelihood of the target hypothesis should increase. Qualitative judgements also significantly impacted the accuracy of belief estimates; deviations from normative, Bayesian, predictions at the group level are explained by sub-groups with incorrect qualitative intuitions
Step-by-step analogical reasoning in humans and neural networks
Get PDFBoth humans and large language models (LLMs) perform better on some reasoning tasks when they are encouraged to think step by step. However, it is unclear whether these performance gains are based on similar principles. In this work, we investigate two hypotheses: (1) that these benefits arise due to the presence of local statistical structure in the training data, where intermediate steps of reasoning may be common but any specific reasoning trajectory is rare, and (2) that sequential processing improves reasoning by mitigating interference. Using LLMs and transformers trained on a synthetic dataset, we show how analogical distance effects previously observed in humans and LLMs may be explained by the presence of local statistical structure. Testing both humans and LLMs on a novel word analogy task, we find that interference caused by semantic similarity can hurt performance and drives humans to engage in a sequential reasoning process. Our findings show that both locality structure and interference may be key principles underlying the benefits of step-by-step thinking
Epistemic Monocultures and the Effect of AI Personalization
Get PDFIt has been argued that when scientists employ algorithmic tools to assist in problem-solving, epistemic monocultures may emerge in which research tools, topics, findings, etc. are homogenized. As a result, fertile areas of research might be left unexplored, impeding scientific progress. To explore the nature of these epistemic monocultures, we develop an agent-based model where agents have the ability to query an AI system to assist in their search of an epistemic (NK) landscape. In general, we find that AI use negatively affects the community of researchers by reducing heterogeneity, but both the rate of AI queries and how AI is used impact the ultimate success of the community. We then implement a potential solution suggested in the literature, AI personalization, and find somewhat mixed results on its potential for mitigating homogenization in research communities
Naturalistic action sampling as foraging in the option space
Get PDFHuman decision-making involves navigating unbounded spaces of possible goals, subgoals, and action sequences. Yet, computational models typically assume pre-defined option sets. This creates a critical gap between the algorithms developed in cognitive science research on decision-making and the open nature of real-world decisions. We propose that option generation in open-ended settings operates as a search through structured decision space. Drawing on foraging theory, we hypothesized that option generation follows Lévy flight distributions, a pattern observed in both spatial foraging and memory retrieval. We found that the inter-generation time between consecutive responses in open-ended option generation problems approximated a Lévy distribution, while semantic distances demonstrated properties of heavy-tailed distributions. These findings reveal connections between action planning, information search, and memory retrieval, suggesting shared computational principles in how humans explore unbounded decision spaces
Adults hold two parallel causal frameworks for reasoning about people's minds, actions and bodies
Get PDFUnderstanding other people involves making sense of their physical actions, mental states, and physiological experiences, yet little is known about the causal beliefs we hold across these domains. Across two exploratory studies, we measured these beliefs and their use in social cognition. In Study 1 (N = 50, M age = 39.44y), US adults (1) freely sorted and (2) reported causal beliefs about events of the mind, body, and actions.
Representational similarity analysis (RSA) revealed two causal frameworks: one representing the 3 distinct latent categories, and another expressing causal relationships across them. Study 2 (N = 100, M age = 39.95y) demonstrated that adults flexibly apply either framework depending on the task, using the latent causes for trait inference, and causal beliefs to plan interventions on other agents. These findings suggest that intuitive theories of other people include both a sense of which capacities"go together" and their causal connections within and across domains
How constraints on editing affects cultural evolution
Get PDFWhen is it beneficial to constrain creativity? Creativity thrives with freedom, but when people collaborate to create artifacts, there is tension between giving individuals freedom to revise, and protecting prior achievements. To test how imposing constraints may affect collective creativity, we performed cultural evolution experiments where participants collaborated to create melodies and images in chains. With melodies, we found that limiting step size (number of musical notes that can be changed) improved pleasantness ratings. Similar results were observed in cohorts of musicians, and with different selection regimes. This outcome was due to the tendency to overcrowd melodies. Interestingly, limiting step size in creating images consistently reduced pleasantness. These conflicting findings suggest that in domains such as music, where artifacts can be easily damaged, collective creativity may benefit from imposing small step sizes or limiting overcrowding. We discuss parallels with search algorithms and the evolution of conservative birdsong cultures
Spatial Terms in English Plus Twelve Languages: Evidence for Functional and Geometric Classes
Get PDFA long-standing open question concerns the universal properties of the representations of spatial expressions, specifically, the question of whether they fall into two separate classes, functional/ force-dynamic vs. geometric (Landau, 2025). Recently Viechnicki et al. (2024) proposed a new method for examining spatial expressions across languages by filtering massive parallel text corpora for basic locative constructions (BLCs); that study showed promising but tentative and limited evidence for the two classes of spatial expressions across languages. The current study replicates and extends those overall findings using a larger corpus and a more effective filtration technique. Experiment 1 analyzes cross-linguistic variational patterns from a corpus of parallel BLCs from 12 languages, finding distinct patterns for functional and geometric spatial terms. Experiment 2 examines the semantics of ground objects from a large corpus of English BLCs and reveals additional evidence for two underlying classes of spatial relations. The two experiments strengthen the evidence from web-scale corpus linguistics supporting distinct universal cognitive representations of functional vs. geometric spatial terms
Sense-Making, Cultural Scripts, and the Inferential Basis of Meaningful Experience
Get PDFCognitive science has made great progress in understanding how we explain and make sense of a complex world. We lack, however, an account of a deeper notion: how an experience that makes sense can become one that is meaningful. We present an account of how explanation and sense-making can lead to meaning-making, by the use—and, crucially, re-use—of a small set of cultural scripts: explanatory complexes that can be shared across domains by members of a social group. We explain meaning-making as a process of inference in which an individual leverages these cultural scripts to segment their full, unstructured set of experiences into a form that can be understood and endorsed as significant. Our account suggests how cultural artifacts (particularly stories in the form of novels, plays, and movies) are crucial for the transmission of these scripts. We present a mathematical model of this inferential process that can account for a range of phenomena which typically resist formalization. This includes the importance of narratives in meaning-making, the difficulty of articulating meaning separately from experiences that encapsulate it, and the ways in which the standard interpretation of a stable cultural artifact can change radically over time
The Role of Task-Unrelated Thinking Characteristics and Function in Affect Regulation During Online and On-site Classes
Get PDFTask-unrelated thinking (TUT) can impact both performance and well-being, yet its role in affect regulation remains underexplored, especially in an educational context. This study examined TUT level, characteristics, and functions in 173 on-site and 143 online students, assessing their affect and class experiences in an ecological setting. While overall TUT levels did not differ between groups, distinctions emerged in characteristics (e.g., inner speech) and functions (e.g., stimulation or avoidance). Valence was the only characteristic predicting prospective sadness or anxiety. Using TUT for problem-solving or avoidance was linked to increased sadness, whereas using it for stimulation was linked to reduced anxiety. These findings highlight that TUT's effects depend more on its nature and purpose than its frequency. The observed link between avoidance-related TUT and negative affect has significant implications for clinical psychology and educational settings, particularly in understanding emotion regulation in online and on-site learning
Path encoding and manner salience in motion event descriptions: the case of Bulgarian and English
Get PDFExamining motion event descriptions allows us to evaluate what information individuals deem to be salient when communicating about events. Systematic variations between languages regarding how they encode details about motion events have given rise to theories of typological classifications of languages. In this study we evaluate the classification of Bulgarian, a South Slavic language, along Talmy's typology of verb-framed and satellite-framed languages, and relate this to Slobin's concept of manner salience. Based on behavioural evidence from an experiment using free-form descriptions in Bulgarian and English, we show that path encoding in Bulgarian is distinct from that of English (a satellite-framed language), but the use of complex path expressions in Bulgarian means it cannot be easily captured by Talmy's two-way classification system. We show that Bulgarian displays a lower rate of manner salience than English and patterns similarly to verb-framed languages in its treatment of the default manner of motion (walking)