Search

This paper presents a theoretical framework for modeling human visual attention. The framework's core claim is that three mechanisms drive attention: selection, which picks out an item for further processing; engagement, which tags a selected item as relevant or irrelevant to the current task; and enhancement, which increases sensitivity to task-relevant items and decreases sensitivity to task-irrelevant items. Building on these mechanisms, the framework is able to explain human performance on attentionally demanding tasks like visual search and multiple object tracking, and it supports a broad range of predictions about the interactions between such tasks.

Quantity can be expressed in a variety of ways and at dif-ferent levels of precision. One factor that influences numer-ical description of elements in a visual scene is how long thescene is observed. We extend a previous incremental modelof numerical perception to model quantified description undertime constraints. Our extended model predicts that as presen-tation duration decreases and as the quantity of items to beenumerated increases, the frequency of inexact quantifiers willincrease. We conducted two human subject elicitation stud-ies to test these predictions. Our findings were consistent withour model’s predictions. Additionally, we demonstrate that ournovel model of incremental numerical perception and quanti-fied description closely predicts the precise proportion of exactnumerical responses generated by in these experiments.

People often reason about omissions. One line of researchshows that people can distinguish between the semantics ofomissive causes and omissive enabling conditions: forinstance, not flunking out of college enabled you (but didn’tcause you) to graduate. Another line of work shows that peoplerely on the normative status of omissive events in inferringtheir causal role: if the outcome came about because theomission violated some norm, reasoners are more likely toselect that omission as a cause. We designed a novel paradigmthat tests how norms interact with the semantics of omissiveenabling conditions. The paradigm concerns the circuitry of amechanical device that plays music. Two experiments used theparadigm to stipulate norms and present a distinct set ofpossibilities to participants. Participants chose which causalverb best described the operations of the machine. The studiesrevealed that participants’ responses are best predicted by theirtendency to consider the semantics of omissive relations. Incontrast, norms had little to no effect in participants’ responses.We conclude by marshaling the evidence and considering whatrole norms may play in people’s understanding of omissions.

We present a computational model of attentional capture inhumans. The model distinguishes between automatic mecha-nisms that directly determine the focus of visual attention, anddeliberate mental actions an individual can perform to influ-ence these mechanisms. The automatic mechanisms select anobject as the focus of attention and enhance its location andfeatures, so that nearby or similar objects are likely to be se-lected in the future. The deliberate actions include engagingwith a selected object to further enhance its features, and re-trieving a previously selected object from memory. By per-forming these actions, the model is able to exert limited top-down control over capture, increasing the probability that task-relevant objects will be attended and irrelevant objects will beignored. To evaluate the model, we conduct a simulation of arecent visual search study, demonstrating that the model canaccount for three established factors that are known to influ-ence capture.

Is it reasonable to draw temporal conclusions from omissive causal assertions? For example, if you learn that not chargingyour phone caused it to die, is it sensible to infer that your failure to charge your phone occurred before it died? Theconclusion seems intuitive, but no theory of causal reasoning explains how reasoners make the inference other than a recentproposal by Khemlani and colleagues (2018a). We present that theory and describe its consequences. If people conceiveof omissions as non-events, i.e., events unmoored in space and time, they might refrain from drawing conclusions whenasked whether an omissive cause precedes its effect. Two experiments speak against these predictions of the non-eventview and in favor of a view that omissive causation imposes temporal constraints on events and their effects. We concludeby considering whether drawing a temporal conclusion from an omissive cause constitutes a reasoning error.

We provide a new computational model of working memory inthe complex span task implemented in the ARCADIA cogni-tive framework. While there exist implementations of workingmemory successful enough to account for many of the bench-mark findings in the working memory literature, we demon-strate that further progress requires the integration of thesemodels with a rich conception of attention. ARCADIA pro-vides this intersection, allowing for precise control of the focusof attention on a time scale fine enough to begin to disentan-gle the overlapping effects of interference, temporal decay, andattentional refreshing.

Herein we describe CRAMM, a framework for Causal Reason-ing via Attention and Mental Models. CRAMM develops andextends assumptions made by a previously developed coun-terfactual simulation model of human causal judgment. Weimplement CRAMM computationally and demonstrate how itrobustly captures human causal judgments about simple two-object interactions at the level of underlying cognitive and per-ceptual processes, including data on eye-movements that serveas direct evidence for the role of counterfactuals in causal judg-ment.

Enumeration of canonical patterns (e.g., faces of six-sideddice) has generally been characterized by researchers as aholistic process, in which all items are perceived collectively.In previous work, based on a holistic processing view of enu-meration by pattern recognition, we predicted that enumera-tion of canonical forms would not be significantly affected byattentional load. In this paper, we present the results from twoexperiments designed to test this prediction using a divided-attention paradigm. In contrast to our predictions, enumerationof canonical patterns was disrupted by attentional load. Fur-thermore, enumeration of patterns under high attentional loadshowed evidence of conflation between patterns with similarcontours, providing evidence against a holistic processing ac-count of canonical pattern recognition.

We present a computational model exploring goal-directeddeployment of attention during object tracking. Once selected,objects are tracked in parallel, but serial attention can bedirected to an object that is visually crowded and in danger ofbeing lost. An attended object’s future position can beextrapolated from its past motion trajectory, allowing theobject to be tracked even when it is briefly occluded. Usingthe model, we demonstrate that the difficulty of trackingthrough occlusions increases with the number of objectsbecause they compete for serial attention.

Omissions figure prominently in causal reasoning fromdiagnosis to ascriptions of negligence. One philosophicalproposal posits that omissions are accompanied by acontrasting alternative that describes a case of orthodox (non-omissive) causation (Schaffer, 2005; Bernstein, 2014). Apsychological hypothesis can be drawn from this contrastview of omissions: by default, humans should interpretomissive causations as representing at least two possibilities,i.e., a possibility representing the omission and a possibilityrepresenting a contrast. The theory of mental models supposesthat reasoners construct only one possibility (the omission) bydefault, and that they consider separate alternativepossibilities in sequential order. Two experiments test thecontrast hypothesis against the model theory, and findevidence in favor of the model-theoretic account.

Recent studies in the perception of numerosity have indicatedthat subitizing (the rapid and accurate enumeration of smallquantities) requires attention. We present a novel computa-tional model of enumeration in which attention unifies dis-tinct processes of numerosity approximation, subitizing, andexplicit counting. We demonstrate how this model accountsfor both the reaction time results from the subitizing literatureand the effects of attentional load on subitizing accuracy.

The rich literature on multiple object tracking (MOT)conclusively demonstrates that humans are able to visuallytrack a small number of objects. There is considerably lessagreement on what perceptual and cognitive processes areinvolved. While it is clear that MOT is attentionallydemanding, various accounts of MOT performance centrallyinvolve pre-attentional mechanisms as well. In this paper wepresent an account of object tracking in the ARCADIAcognitive system that treats MOT as dependent upon both pre-attentive and attention-bound processes. We show that withminimal addition this model replicates a variety of corephenomena in the MOT literature and provides an algorithmicexplanation of human performance limitations.

Attention is thought to be a part of a larger cluster of mecha-nisms that serve to orient a cognitive system, to filter contentswith respect to their task relevance, and to devote more com-putation to certain options than to others. All these activitiesproceed under the plausible assumption that not all informationcan be or ought to be processed for a system to satisfice in anever changing world. In this paper, we describe an attention-centric cognitive system called ARCADIA that demonstratesthe orienting, filtering, and resource-skewing functions men-tioned above. The demonstration involves maintaining focuson cognitive tasks in a dynamic environment. While ARCA-DIA carries out a task, limits on its attentional capacity resultin “inattentional blindness” under circumstances analogous tothose where people fail to perceive otherwise salient stimuli.

Current psychological accounts of causal representation andreasoning do not capture phenomena related to causation byomission (e.g., “The absence of breathing causes death”),with one exception (Wolff, Barbey, & Hausknecht, 2010). Wedescribe a novel theory of omissive causation that posits thatpeople build discrete mental simulations – mental models – ofcausal relations (Goldvarg & Johnson-Laird 2001). Thetheory states that causes by omission refer to a set oftemporally ordered models of possibilities. Reasoners tend tofocus on only one of those models, i.e., the possibility inwhich breathing does not occur and death subsequently does.Likewise, the theory posits that reasoners distinguish betweenomission in the context of causation, enabling conditions, andprevention. We describe some initial predictions made by themodel-based account, contrast it with an alternativepsychological theory based on the transmission of causalforces, and set out directions for further research.

With few exceptions, architectural approaches to modelingcognition have historically emphasized what happens in themind following the transduction of environmental signals intopercepts. To our knowledge, none of these architecturesimplements a sophisticated, general theory of human attention.In this paper we summarize progress to date on a new cognitivearchitecture called ARCADIA that gives a central role toattention in both perception and cognition. First, we give anoverview of the architecture, comparing it to other approacheswhen appropriate. Second, we present a model of incrementalobject construction and property binding in ARCADIA usingthe well known change blindness phenomena to illustrate thetime course of object perception and its dependence on attention.Finally, we discuss near-term challenges and future plans

The common view of the transition between subitizing and numerosity estimation regimes is that there is a hard bound on the subitizing range, and beyond this range, people estimate. However, this view does not adequately address the behavioral signatures of enumeration under conditions of attentional load or in the immediate post‐subitizing range. The possibility that there might exist a numerosity range where both processes of subitizing and estimation operate in conjunction has so far been ignored. Here, we investigate this new proposal, that people strategically combine the processes of subitizing and estimation to maximize accuracy and precision, given time or attentional constraints. We present a process‐level account of how subitizing and estimation can be combined through strategic deployment of attention to maximize the precision of perceived numerosity given time constraints. We then describe a computational model of this account and apply it in two experimental simulations to demonstrate how it can explain key findings in prior enumeration research. While recent modeling work has argued that the behavioral signatures of enumeration can best be explained through a single numerosity system with a single form of representation, we argue that our model demonstrates how the traditional two‐systems view of numerical representation accounts for behavioral data through coordination with a unified attentional mechanism, rather than a unified representation. [ABSTRACT FROM AUTHOR]

Computational models will play an important role in our understanding of human higher-order cognition. How can a model's contribution to this goal be evaluated? This article argues that three important aspects of a model of higher-order cognition to evaluate are (a) its ability to reason, solve problems, converse, and learn as well as people do; (b) the breadth of situations in which it can do so; and (c) the parsimony of the mechanisms it posits. This article argues that fits of models to quantitative experimental data, although valuable for other reasons, do not address these criteria. Further, using analogies with other sciences, the history of cognitive science, and examples from modern-day research programs, this article identifies five activities that have been demonstrated to play an important role in our understanding of human higher-order cognition. These include modeling within a cognitive architecture, conducting artificial intelligence research, measuring and expanding a model's ability, finding mappings between the structure of different domains, and attempting to explain multiple phenomena within a single model. (Contains 1 figure and 1 note.)