The past,present, and future of computational models of cognitive development

Does modeling matter? We address this question by providing a broad survey of the computational models of cognitive development that have been proposed and studied over the last three decades. We begin by noting the advantages and limitations of computational models. We then describe four key dimensions across which models of development can be organized and classified. With this taxonomy in hand, we focus on how the modeling enterprise has evolved over time. In particular, we separate the timeline into three overlapping historical waves and highlight how each wave of models has not only been shaped by developmental theory and behavioral research, but in return also provided valuable insights and innovations to the study of cognitive development.     

Computer simulations of developmental change: the contributions of working memory capacity and long-term knowledge

Increasing working memory (WM) capacity is often cited as a major influence on children's development and yet WM capacity is difficult to examine independently of long-term knowledge. A computational model of children's nonword repetition (NWR) performance is presented that independently manipulates long-term knowledge and WM capacity to determine the relative contributions of each in explaining the developmental data. The simulations show that (a) both mechanisms independently cause the same overall developmental changes in NWR performance, (b) increase in long-term knowledge provides the better fit to the child data, and (c) varying both long-term knowledge and WM capacity adds no significant gains over varying long-term knowledge alone. Given that increases in long-term knowledge must occur during development, the results indicate that increases in WM capacity may not be required to explain developmental differences. An increase in WM capacity should only be cited as a mechanism of developmental change when there are clear empirical reasons for doing so.     

The role of embodied intention in early lexical acquisition

We examine the influence of inferring interlocutors' referential intentions from their body movements at the early stage of lexical acquisition. By testing human participants and comparing their performances in different learning conditions, we find that those embodied intentions facilitate both word discovery and word-meaning association. In light of empirical findings, the main part of this article presents a computational model that can identify the sound patterns of individual words from continuous speech, using nonlinguistic contextual information, and employ body movements as deictic references to discover word-meaning associations. To our knowledge, this work is the first model of word learning that not only learns lexical items from raw multisensory signals to closely resemble infant language development from natural environments, but also explores the computational role of social cognitive skills in lexical acquisition.     

An assessment of age differences in theory of mind using the virtual assessment of mentalizing ability

ABSTRACT

Prior research has shown that older adults perform more poorly than young on tasks that assess theory of mind (ToM). However, these studies have used tasks that are performed “offline” (i.e., requiring a second-person perspective) as opposed to “online” (i.e., requiring a first-person perspective). Therefore, the present study was designed to establish whether age-related ToM difficulties are also evident when an “online” measure of ToM is used. Forty younger and 40 older adults completed the Virtual Assessment of Mentalizing Ability (VAMA) along with two conventional ToM tasks. No age differences were evident on the conventional measures, but older adults had lower accuracy on the VAMA relative to their younger counterparts. The overall pattern of errors did not differ between the groups. These data provide no evidence that age effects are reduced when stimuli are used that are more likely to engage the mentalizing processes elicited in real life social interactions.     

Understanding intentions from actions: Direct perception,inference, and the roles of mirror and mentalizing systems

This review asks whether observers can obtain information about others’ intentions from observation of their actions; and if so, whether this process is performed using direct perceptual or inferential processes (prominent examples of each being the intention understanding theory of mirror neuron function, and mentalizing accounts of intention understanding, respectively). I propose four conditions that should be fulfilled in order to support a direct perception account, and suggest that only two of these conditions are supported by the existing data. I then propose and review three further sources of evidence which have the potential to inform this debate, concluding that the data do not support the direct perception account. In particular, mirror neurons may be involved in lower-level processes of action perception, but there is no evidence to support their involvement in the type of higher-level intention understanding that is proposed by the direct perception account.     

Aging and the use of context in ambiguity resolution: complex changes from simple slowing

Older and younger adults' abilities to use context information rapidly during ambiguity resolution were investigated. In Experiments 1 and 2, younger and older adults heard ambiguous words (e.g., fires) in sentences where the preceding context supported either the less frequent or more frequent meaning of the word. Both age groups showed good context use in offline tasks, but only young adults demonstrated rapid use of context in cross-modal naming. A 3rd experiment demonstrated that younger and older adults had similar knowledge about the contexts used in Experiments 1 and 2. The experiment results were simulated in 2 computational models in which different patterns of context use were shown to emerge from varying a single speed parameter. These results suggest that age-related changes in processing efficiency can modulate context use during language comprehension.     

Rule transition on the balance scale task: a case study in belief change

For various domains in proportional reasoning cognitive development is characterized as a progression through a series of increasingly complex rules. A multiplicative relationship between two task features, such as weight and distance information of blocks placed at both sides of the fulcrum of a balance scale, appears difficult to discover. During development, children change their beliefs about the balance scale several times: from a focus on the weight dimension (Rule I) to occasionally considering the distance dimension (Rule II), guessing (Rule III), and applying multiplication (Rule IV; Siegler, 1981). Because of the detailed empirical findings the balance scale task has become a benchmark task for computational models of proportional reasoning. In this article, we present a large empirical study (N = 420) of which the findings provide a challenge for computational models. The effect of feedback and the effect of individually adapted training items on rule transition were tested for children using Rule I or Rule II. Presenting adapted training items initiates belief revision for Rule I but not for Rule II. The experience of making mistakes (by providing feedback) induces a change for both Rule I and Rule II. However, a delayed posttest shows that these changes are preserved after 2 weeks only for children using Rule I. We conclude that the transition from Rule I to Rule II differs from the transition from Rule II to a more complex rule. Concerning these empirical findings, we will review performance of computational models and the implications for a future belief revision model.

It is one Thing, to show a Man that he is in an Error, and another, to put him in possession of Truth. John Locke

     

Variability in the early development of visual self-recognition

A study was conducted to evaluate the (1) developmental course and (2) the temporal sequencing of visual (mirror, photo) and verbal (personal pronoun use) measures of self-recognition as well as the ability to locate a toy from its mirror image in relation to the child's own mirror image. A microgenetic approach was adopted to assess 10 toddlers biweekly between 15 and 23 months of age and for comparison, a cross-section of children tested once across the same age range. Longitudinal data indicated that visual self-recognition emerged gradually and showed wide variability in expression prior to becoming stable, a finding masked in the cross-sectional data where performance appeared to improve abruptly. Both cross-sectional and longitudinal data confirmed that mirror self-recognition was the earliest precursor of the indices of self-recognition to emerge followed by the use of personal pronouns and photo identification. Implications for the emergence and integration of the self are discussed.     

The role of experience in children's developing folk epistemology: review and analysis from the theory–theory perspective

The role of experience in children's developing folk epistemology, specifically in their acquisition of an understanding of belief, is explored from the theory–theory perspective (i.e., the perspective claiming that children's conceptual development can be viewed as analogous to scientific theory development) through a selective review and analysis of extant empirical literature. Three types of investigations are identified: examinations of observed relationships between social experiences and belief understanding, attempts to facilitate the reasoning of children on the verge of belief understanding through experiences intended to compel children to face reasoning inconsistencies; and training studies demonstrating that children can be trained to understand belief over time through specific feedback experiences. A comparison of these studies reveals that experience is defined differently across them, resulting in varying and sometimes ambiguous implications for a theory–theory perspective on development. Implications for future research are discussed.     

A computational model of early argument structure acquisition

How children go about learning the general regularities that govern language, as well as keeping track of the exceptions to them, remains one of the challenging open questions in the cognitive science of language. Computational modeling is an important methodology in research aimed at addressing this issue. We must determine appropriate learning mechanisms that can grasp generalizations from examples of specific usages, and that exhibit patterns of behavior over the course of learning similar to those in children. Early learning of verb argument structure is an area of language acquisition that provides an interesting testbed for such approaches due to the complexity of verb usages. A range of linguistic factors interact in determining the felicitous use of a verb in various constructions —associations between syntactic forms and properties of meaning that form the basis for a number of linguistic and psycholinguistic theories of language. This article presents a computational model for the representation, acquisition, and use of verbs and constructions. The Bayesian framework is founded on a novel view of constructions as a probabilistic association between syntactic and semantic features. The computational experiments reported here demonstrate the feasibility of learning general constructions, and their exceptions, from individual usages of verbs. The behavior of the model over the timecourse of acquisition mimics, in relevant aspects, the stages of learning exhibited by children. Therefore, this proposal sheds light on the possible mechanisms at work in forming linguistic generalizations and maintaining knowledge of exceptions.     

A computational cognitive framework of spatial memory in brains and robots

Computational cognitive models of spatial memory often neglect difficulties posed by the real world, such as sensory noise, uncertainty, and high spatial complexity. On the other hand, robotics is unconcerned with understanding biological cognition. Here, we describe a computational framework for robotic architectures aiming to function in realistic environments, as well as to be cognitively plausible.We motivate and describe several mechanisms towards achieving this despite the sensory noise and spatial complexity inherent in the physical world. We tackle error accumulation during path integration by means of Bayesian localization, and loop closing with sequential gradient descent. Finally, we outline a method for structuring spatial representations using metric learning and clustering. Crucially, unlike the algorithms of traditional robotics, we show that these mechanisms can be implemented in neuronal or cognitive models.We briefly outline a concrete implementation of the proposed framework as part of the LIDA cognitive architecture, and argue that this kind of probabilistic framework is well-suited for use in cognitive robotic architectures aiming to combine spatial functionality and psychological plausibility.     

Parameters, Predictions, and Evidence in Computational Modeling: A Statistical View Informed by ACT-R

Model validation in computational cognitive psychology often relies on methods drawn from the testing of theories in experimental physics. However, applications of these methods to computational models in typical cognitive experiments can hide multiple, plausible sources of variation arising from human participants and from stochastic cognitive theories, encouraging a "model fixed, data variable" paradigm that makes it difficult to interpret model predictions and to account for individual differences. This article proposes a likelihood-based, "data fixed, model variable" paradigm in which models are treated as stochastic processes in experiments with participant-to-participant variation that can be applied to a broad range of mechanistic cognitive architectures. This article discusses the implementation and implications of this view in model validation, with a concrete focus on a simple class of ACT-R models of cognition. This article is not intended as a recipe for broad application of these preliminary, proof-of-concept methods, but as a framework for communication between statisticians searching for interesting problems in the cognitive modeling sphere, and cognitive modelers interested in generalizing from deterministic to stochastic model validation, in the face of random variation in human experimental data.     

Controlled & automatic processing: behavior,theory, and biological mechanisms

This paper provides an overview of developments in a dual processing theory of automatic and controlled processing that began with the empirical and theoretical work described by Schneider and Shiffrin (1977) and Shiffrin and Schneider (1977) over a quarter century ago. A review of relevant empirical findings suggests that there is a set of core behavioral phenomena reflecting differences between controlled and automatic processing that must be addressed by a successful theory. These phenomena relate to: consistency in training, serial versus parallel processing, level of effort, robustness to stressors, degree of control, effects on long‐term memory, and priority encoding. We detail a computational model of controlled processing, CAP2, that accounts for these phenomena as emergent properties of an underlying hybrid computational architecture. The model employs a large network of distributed data modules that can categorize, buffer, associate, and prioritize information. Each module is a connectionist network with input and output layers, and each module communicates with a central Control System by outputting priority and activity report signals, and by receiving control signals. The Control System is composed of five processors including a Goal Processor, an Attention Controller, an Activity Monitor, an Episodic Store, and a Gating & Report Relay. The transition from controlled to automatic processing occurs in this model as the data modules become capable of transmitting their output without mediation by the Control System. We describe recent progress in mapping the components of this model onto specific neuroanatomical substrates, briefly discuss the potential for applying functional neuroimaging techniques to test the model's predictions, and its relation to other models.     

Learning about life and death in early childhood

have argued that young children initially understand biological phenomena in terms of vitalism, a mode of construal in which "life" or "life-force" is the central causal-explanatory concept. This study investigated the development of vitalistic reasoning in young children's concepts of life, the human body and death. Sixty preschool children between the ages of 3 years, 7 months and 5 years, 11 months participated. All children were initially given structured interviews to assess their knowledge of (1) human body function and (2) death. From this sample 40 children in the Training group were taught about the human body and how it functions to maintain life. The Control group (n=20) received no training. All 60 children were subsequently reassessed on their knowledge of human body function and death. Results from the initial interviews indicated that young children who spontaneously appealed to vitalistic concepts in reasoning about human body functioning were also more sophisticated in their understanding of death. Results from the posttraining interviews showed that children readily learned to adopt a vitalistic approach to human body functioning, and that this learning coincided with significant development in their understanding of human body function, and of death. The overall pattern of results supports the claim that the acquisition of a vitalistic causal-explanatory framework serves to structure children's concepts and facilitates learning in the domain of biology.     

From blooming,buzzing confusion to media literacy: The early development of television viewing

Television comprehension is a surprisingly demanding task for very young children. Based on a task analysis of television viewing and review of research, we suggest that by 6 months of age, infants can identify objects and people on screen. By 24 months they can comprehend and imitate simple actions contained in single shots and begin to integrate information across shots. Toddlers nevertheless suffer from the video deficit whereby their comprehension is less than from equivalent real life displays. During the preschool years they learn much of the grammar of filmic montage and can comprehend straightforward narrative productions. Essentially adult comprehension appears to be in place by 13 years of age.     

Children do not follow the rule "ignorance means getting it wrong"

Two experiments tested whether 4- and 5-year-olds follow the rule “ignorance means you get it wrong.” Following this rule should lead children to infer that a character who is ignorant about some situation will also have a false belief about it. This rule should sometimes lead children into error because ignorance does not imply false belief. In Experiment 1, children and adults were told about a girl who is looking for her dog but does not know which of two boxes it is under. Most children predicted that the girl would look in the box with the dog and not in the empty box; adults chose both boxes equally. Experiment 2 used a similar story but varied whether the girl wants to approach or avoid her dog. Again, most children predicted that the girl would succeed. These findings suggest that children do not follow the rule “ignorance means you get it wrong.”     

Evaluating scientific research: Belief,hindsight bias,ethics, and research evaluation

Students may exhibit two forms of cognitive biases, belief and hindsight bias, in evaluating a scientific experiment. Counter to disagreement, they may only believe an outcome that agrees with their belief to be more predictable in hindsight than foresight. The focus of this research is on the relationship between these biases. Students were queried about their dichotomous beliefs (learned vs. genetic) about behavior for an animal experiment and then assigned randomly to a no‐outcome or genetic outcome condition. With agreement between students' belief and outcome, the findings revealed hindsight bias (foreseeability) supported by the outcomes for surprise, disappointment, ethics, and research evaluation. With disagreement, hindsight bias was trumped along with perceiving the experiment as being less ethical and scientifically sound. Regardless of the outcome, students seem to adhere to their beliefs. Hence, students may believe that the outcome is inconsequential because it is obvious or contrary to their beliefs.     

Autonomous construction of ecologically and socially relevant semantics

This article presents a synthetic modeling approach to the problem of grounded construction of concepts. In many computational models of grounded language acquisition and evolution, meanings are created in the process of discrimination between a chosen object and other objects present on the scene of communication. We argue that categories constructed for the purpose of identification rather than discrimination are more suitable for the detached language use (talking about things not present here and now). We describe a semantics based on so-called identification criteria constructed by extracting cross-situational similarities among instances of a category, and present several computational models. In the model of individual category construction, the instances are grouped to categories by common motor programs (affordances), while in the model of social learning, focused on the influence of naming on category formation, entities are considered members of the same category, if they are labeled with the same word by an external teacher. By these two mechanisms, the learner can construct interactionally grounded representation of objects, properties, relations, changes, complex situations and events. We also report and analyze simulation results of an experiment focused on the dynamics of meanings in iterated intergenerational transmission.     

Selective attention,diffused attention,and the development of categorization

How do people learn categories and what changes with development? The current study attempts to address these questions by focusing on the role of attention in the development of categorization. In Experiment 1, participants (adults, 7-year-olds, and 4-year-olds) were trained with novel categories consisting of deterministic and probabilistic features, and their categorization and memory for features were tested. In Experiment 2, participants’ attention was directed to the deterministic feature, and in Experiment 3 it was directed to the probabilistic features. Attentional cueing affected categorization and memory in adults and 7-year-olds: these participants relied on the cued features in their categorization and exhibited better memory of cued than of non-cued features. In contrast, in 4-year-olds attentional cueing affected only categorization, but not memory: these participants exhibited equally good memory for both cued and non-cued features. Furthermore, across the experiments, 4-year-olds remembered non-cued features better than adults. These results coupled with computational simulations provide novel evidence (1) pointing to differences in category representation and mechanisms of categorization across development, (2) elucidating the role of attention in the development of categorization, and (3) suggesting an important distinction between representation and decision factors in categorization early in development. These issues are discussed with respect to theories of categorization and its development.