The competition-compensation account of developmental language disorder.

Children with developmental language disorder (DLD) regularly use the bare form of verbs (e.g., dance) instead of inflected forms (e.g., danced). We propose an account of this behavior in which processing difficulties of children with DLD disproportionally affect processing novel inflected verbs in their input. Limited experience with inflection in novel contexts leads the inflection to face stronger competition from alternatives. Competition is resolved through a compensatory behavior that involves producing a more accessible alternative: in English, the bare form. We formalize this hypothesis within a probabilistic model that trades off context-dependent versus independent processing. Results show an over-reliance on preceding stem contexts when retrieving the inflection in a model that has difficulty with processing novel inflected forms. We further show that following the introduction of a bias to store and retrieve forms with preceding contexts, generalization in the typically developing (TD) models remains more or less stable, while the same bias in the DLD models exaggerates difficulties with generalization. Together, the results suggest that inconsistent use of inflectional morphemes by children with DLD could stem from inferences they make on the basis of data containing fewer novel inflected forms. Our account extends these findings to suggest that problems with detecting a form in novel contexts combined with a bias to rely on familiar contexts when retrieving a form could explain sequential planning difficulties in children with DLD. RESEARCH HIGHLIGHTS: Generalization difficulties with inflectional morphemes in children with Developmental Language Disorder arise from these children's limited experience with novel inflected forms. Limited experience with a form in novel contexts could lead to a storage bias where retrieving a form often requires relying on familiar preceding stems. While generalization in typically developing models remains stable across a range of model parameters, certain parameter values in the impaired models exaggerate difficulties with generalization. Children with DLD compensate for these retrieval difficulties through accessibility-driven language production: they produce the most accessible form among the alternatives.

Efficient Discretization of Optimal Transport.

Obtaining solutions to optimal transportation (OT) problems is typically intractable when marginal spaces are continuous. Recent research has focused on approximating continuous solutions with discretization methods based on i.i.d. sampling, and this has shown convergence as the sample size increases. However, obtaining OT solutions with large sample sizes requires intensive computation effort, which can be prohibitive in practice. In this paper, we propose an algorithm for calculating discretizations with a given number of weighted points for marginal distributions by minimizing the (entropy-regularized) Wasserstein distance and providing bounds on the performance. The results suggest that our plans are comparable to those obtained with much larger numbers of i.i.d. samples and are more efficient than existing alternatives. Moreover, we propose a local, parallelizable version of such discretizations for applications, which we demonstrate by approximating adorable images.

Conditional Deep Gaussian Processes: Empirical Bayes Hyperdata Learning.

It is desirable to combine the expressive power of deep learning with Gaussian Process (GP) in one expressive Bayesian learning model. Deep kernel learning showed success as a deep network used for feature extraction. Then, a GP was used as the function model. Recently, it was suggested that, albeit training with marginal likelihood, the deterministic nature of a feature extractor might lead to overfitting, and replacement with a Bayesian network seemed to cure it. Here, we propose the conditional deep Gaussian process (DGP) in which the intermediate GPs in hierarchical composition are supported by the hyperdata and the exposed GP remains zero mean. Motivated by the inducing points in sparse GP, the hyperdata also play the role of function supports, but are hyperparameters rather than random variables. It follows our previous moment matching approach to approximate the marginal prior for conditional DGP with a GP carrying an effective kernel. Thus, as in empirical Bayes, the hyperdata are learned by optimizing the approximate marginal likelihood which implicitly depends on the hyperdata via the kernel. We show the equivalence with the deep kernel learning in the limit of dense hyperdata in latent space. However, the conditional DGP and the corresponding approximate inference enjoy the benefit of being more Bayesian than deep kernel learning. Preliminary extrapolation results demonstrate expressive power from the depth of hierarchy by exploiting the exact covariance and hyperdata learning, in comparison with GP kernel composition, DGP variational inference and deep kernel learning. We also address the non-Gaussian aspect of our model as well as way of upgrading to a full Bayes inference.

Conditional Deep Gaussian Processes: Multi-Fidelity Kernel Learning.

Deep Gaussian Processes (DGPs) were proposed as an expressive Bayesian model capable of a mathematically grounded estimation of uncertainty. The expressivity of DPGs results from not only the compositional character but the distribution propagation within the hierarchy. Recently, it was pointed out that the hierarchical structure of DGP well suited modeling the multi-fidelity regression, in which one is provided sparse observations with high precision and plenty of low fidelity observations. We propose the conditional DGP model in which the latent GPs are directly supported by the fixed lower fidelity data. Then the moment matching method is applied to approximate the marginal prior of conditional DGP with a GP. The obtained effective kernels are implicit functions of the lower-fidelity data, manifesting the expressivity contributed by distribution propagation within the hierarchy. The hyperparameters are learned via optimizing the approximate marginal likelihood. Experiments with synthetic and high dimensional data show comparable performance against other multi-fidelity regression methods, variational inference, and multi-output GP. We conclude that, with the low fidelity data and the hierarchical DGP structure, the effective kernel encodes the inductive bias for true function allowing the compositional freedom.

Pedagogical Questions in Parent-Child Conversations.

Questioning is a core component of formal pedagogy. Parents commonly question children, but do they use questions to teach? This article defines "pedagogical questions" as questions for which the questioner already knows the answer and intended to help the questionee learn. Transcripts of parent-child conversations were collected from the CHILDES database to examine the frequency and distribution of pedagogical questions. Analysis of 2,166 questions from 166 mother-child dyads and 64 father-child dyads (child's age between 2 and 6 years) showed that pedagogical questions are commonplace during day-to-day parent-child conversations and vary based on child's age, family environment, and historical era. The results serve as a first step toward understanding the role of parent-child questions in facilitating children's learning.

Spatial summation of broadband contrast.

Spatial summation of luminance contrast signals has historically been psychophysically measured with stimuli isolated in spatial frequency (i.e., narrowband). Here, we revisit the study of spatial summation with noise patterns that contain the naturalistic 1/fα distribution of contrast across spatial frequency. We measured amplitude spectrum slope (α) discrimination thresholds and verified if sensitivity to α improved according to stimulus size. Discrimination thresholds did decrease with an increase in stimulus size. These data were modeled with a summation model originally designed for narrowband stimuli (i.e., single detecting channel; Baker & Meese, 2011; Meese & Baker, 2011) that we modified to include summation across multiple-differently tuned-spatial frequency channels. To fit our data, contrast gain control weights had to be inversely related to spatial frequency (1/f); thus low spatial frequencies received significantly more divisive inhibition than higher spatial frequencies, which is a similar finding to previous models of broadband contrast perception (Haun & Essock, 2010; Haun & Peli, 2013). We found summation across spatial frequency channels to occur prior to summation across space, channel summation was near linear and summation across space was nonlinear. Our analysis demonstrates that classical psychophysical models can be adapted to computationally define visual mechanisms under broadband visual input, with the adapted models offering novel insight on the integration of signals across channels and space.

Questioning supports effective transmission of knowledge and increased exploratory learning in pre-kindergarten children.

How can education optimize transmission of knowledge while also fostering further learning? Focusing on children at the cusp of formal schooling (N = 180, age = 4.0-6.0 y), we investigate learning after direct instruction by a knowledgeable teacher, after questioning by a knowledgeable teacher, and after questioning by a naïve informant. Consistent with previous findings, instruction by a knowledgeable teacher allows effective information transmission but at the cost of exploration and further learning. Critically, we find a dual benefit for questioning by a knowledgeable teacher: Such pedagogical questioning both effectively transmits knowledge and fosters exploration and further learning, regardless of whether the question was directed to the child or directed to a third party and overheard by the child. These effects are not observed when the same question is asked by a naïve informant. We conclude that a teacher's choice of pedagogical method may differentially influence learning through their choices of how, and how not, to present evidence, with implications for transmission of knowledge and self-directed discovery. A video abstract of this article can be viewed at: https://www.youtube.com/watch?v=FJXH2b65wL8.

Young children's attributions of causal power to novel invisible entities.

In two studies, we investigated the development of children's reasoning about potent invisible entities. In Study 1, children aged 2.2-5.5years (N=48) were briefly told about a novel invisible substance that could produce a novel outcome-make a novel box turn green. During this introduction, children watched as one container was inverted over a box and the box lit up green, and then another identical container was inverted over the box and the box did not light up. On test trials, the experimenter inserted a spoon in novel (actually empty) containers and inverted the spoon over the box, which turned green in one trial and did not light up in the other trial. For both trials, children were asked whether there was anything in each container. Children across this age range appropriately reported that an invisible substance was present only when the box lit up. In Study 2, children aged 2.4-4.5years (N=48) watched similar demonstrations but were not explicitly provided information about the invisible substance. Children as young as 3years spontaneously inferred that an invisible substance was present when the box lit up and was absent when the box did not light up. A final task tested children's ability to use their causal knowledge of invisible substances to produce an effect-making the box light up. The youngest children had difficulty with this task, but many children aged 3.5-4.5years performed capably. These results indicate an early-emerging understanding of potent invisible entities that develops rapidly during early childhood.

Distribution of content in recently-viewed scenes whitens perception.

Anisotropies in visual perception have often been presumed to reflect an evolutionary adaptation to an environment with a particular anisotropy. Here, we adapt observers to globally-atypical environments presented in virtual reality to assess the malleability of this well-known perceptual anisotropy. Results showed that the typical bias in orientation perception was in fact altered as a result of recent experience. Application of Bayesian modeling indicates that these global changes of the recently-viewed environment implicate a Bayesian prior matched to the recently experienced environment. These results suggest that biases in orientation perception are fluid and predictable, and that humans adapt to orientation biases in their visual environment "on the fly" to optimize perceptual encoding of content in the recently-viewed visual world.

Epistemic Trust and Education: Effects of Informant Reliability on Student Learning of Decimal Concepts.

The epistemic trust literature emphasizes that children's evaluations of informants' trustworthiness affects learning, but there is no evidence that epistemic trust affects learning in academic domains. The current study investigated how reliability affects decimal learning. Fourth and fifth graders (N = 122; Mage = 10.1 years) compared examples from consistently accurate and inaccurate informants (consistent) or informants who were each sometimes accurate and inaccurate (inconsistent). Fourth graders had higher conceptual knowledge and fewer misconceptions in the consistent condition than the inconsistent condition, and vice versa for fifth graders due to differences in prior exposure to decimals. Given the same examples, learning differed depending on informant reliability. Thus, epistemic trust is a malleable factor that affects learning in an academic domain.

A rational account of pedagogical reasoning: teaching by, and learning from, examples.

Much of learning and reasoning occurs in pedagogical situations--situations in which a person who knows a concept chooses examples for the purpose of helping a learner acquire the concept. We introduce a model of teaching and learning in pedagogical settings that predicts which examples teachers should choose and what learners should infer given a teacher's examples. We present three experiments testing the model predictions for rule-based, prototype, and causally structured concepts. The model shows good quantitative and qualitative fits to the data across all three experiments, predicting novel qualitative phenomena in each case. We conclude by discussing implications for understanding concept learning and implications for theoretical claims about the role of pedagogy in human learning.

Talk of the Town mobile app platform: New method for engaging family in STEM learning and research in homes and communities.

Children do not just learn in the classroom. They engage in "informal learning" every day just by spending time with their family and peers. However, while researchers know this occurs, less is known about the science of this learning-how this learning works. This is so because investigators lack access to those moments of informal learning. In this mini-review we present a technical solution: a mobile-based research platform called "Talk of the Town" that will provide a window into children's informal learning. The tool will be open to all researchers and educators and is flexibly adaptable to these needs. It allows access to data that have never been studied before, providing a means for developing and testing vast educational interventions, and providing access to much more diverse samples than are typically studied in laboratories, homes, and science museums. The review details the promise and challenges associated with these new methods of data collection and family engagement in STEM learning sciences.

The Inner Loop of Collective Human-Machine Intelligence.

With the rise of artificial intelligence (AI) and the desire to ensure that such machines work well with humans, it is essential for AI systems to actively model their human teammates, a capability referred to as Machine Theory of Mind (MToM). In this paper, we introduce the inner loop of human-machine teaming expressed as communication with MToM capability. We present three different approaches to MToM: (1) constructing models of human inference with well-validated psychological theories and empirical measurements; (2) modeling human as a copy of the AI; and (3) incorporating well-documented domain knowledge about human behavior into the above two approaches. We offer a formal language for machine communication and MToM, where each term has a clear mechanistic interpretation. We exemplify the overarching formalism and the specific approaches in two concrete example scenarios. Related work that demonstrates these approaches is highlighted along the way. The formalism, examples, and empirical support provide a holistic picture of the inner loop of human-machine teaming as a foundational building block of collective human-machine intelligence.

When Children's Production Deviates From Observed Input: Modeling the Variable Production of the English Past Tense.

As children gradually master grammatical rules, they often go through a period of producing form-meaning associations that were not observed in the input. For example, 2- to 3-year-old English-learning children use the bare form of verbs in settings that require obligatory past tense meaning while already starting to produce the grammatical -ed inflection. While many studies have focused on overgeneralization errors, fewer studies have attempted to explain the root of this earlier stage of rule acquisition. In this work, we use computational modeling to replicate children's production behavior prior to the generalization of past tense production in English. We illustrate how seemingly erroneous productions emerge in a model, without being licensed in the grammar and despite the model aiming at conforming to grammatical forms. Our results show that bare form productions stem from a tension between two factors: (1) trying to produce a less frequent meaning (the past tense) and (2) being unable to restrict the production of frequent forms (the bare form) as learning progresses. Like children, our model goes through a stage of bare form production and then converges on adult-like production of the regular past tense, showing that these different stages can be accounted for through a single learning mechanism.

Human Variability and the Explore-Exploit Trade-Off in Recommendation.

The enormous scale of the available information and products on the Internet has necessitated the development of algorithms that intermediate between options and human users. These algorithms attempt to provide the user with relevant information. In doing so, the algorithms may incur potential negative consequences stemming from the need to select items about which it is uncertain to obtain information about users versus the need to select items about which it is certain to secure high ratings. This tension is an instance of the exploration-exploitation trade-off in the context of recommender systems. Because humans are in this interaction loop, the long-term trade-off behavior depends on human variability. Our goal is to characterize the trade-off behavior as a function of human variability fundamental to such human-algorithm interaction. To tackle the characterization, we first introduce a unifying model that smoothly transitions between active learning and recommending relevant information. The unifying model gives us access to a continuum of algorithms along the exploration-exploitation trade-off. We then present two experiments to measure the trade-off behavior under two very different levels of human variability. The experimental results inform a thorough simulation study in which we modeled and varied human variability systematically over a wide rage. The main result is that exploration-exploitation trade-off grows in severity as human variability increases, but there exists a regime of low variability where algorithms balanced in exploration and exploitation can largely overcome the trade-off.

An integrated account of generalization across objects and features.

Humans routinely make inductive generalizations about unobserved features of objects. Previous accounts of inductive reasoning often focus on inferences about a single object or feature: accounts of causal reasoning often focus on a single object with one or more unobserved features, and accounts of property induction often focus on a single feature that is unobserved for one or more objects. We explore problems where people must make inferences about multiple objects and features, and propose that people solve these problems by integrating knowledge about features with knowledge about objects. We evaluate three computational methods for integrating multiple systems of knowledge: the output combination approach combines the outputs produced by these systems, the distribution combination approach combines the probability distributions captured by these systems, and the structure combination approach combines a graph structure over features with a graph structure over objects. Three experiments explore problems where participants make inferences that draw on causal relationships between features and taxonomic relationships between animals, and we find that the structure combination approach provides the best account of our data.

Epistemic trust: modeling children's reasoning about others' knowledge and intent.

A core assumption of many theories of development is that children can learn indirectly from other people. However, indirect experience (or testimony) is not constrained to provide veridical information. As a result, if children are to capitalize on this source of knowledge, they must be able to infer who is trustworthy and who is not. How might a learner make such inferences while at the same time learning about the world? What biases, if any, might children bring to this problem? We address these questions with a computational model of epistemic trust in which learners reason about the helpfulness and knowledgeability of an informant. We show that the model captures the competencies shown by young children in four areas: (1) using informants' accuracy to infer how much to trust them; (2) using informants' recent accuracy to overcome effects of familiarity; (3) inferring trust based on consensus among informants; and (4) using information about mal-intent to decide not to trust. The model also explains developmental changes in performance between 3 and 4 years of age as a result of changing default assumptions about the helpfulness of other people.

Sensitivity to the slope of the amplitude spectrum is dependent on the spectral slopes of recently viewed environments: A visual adaptation study in modified reality.

Scenes contain many statistical regularities that could benefit visual processing if accounted for by the visual system. One such statistic is the orientation-averaged slope (α) of the amplitude spectrum of natural scenes. Human observers show different discrimination sensitivity to α: sensitivity is highest for α values between 1.0 and 1.2 and decreases as α is steepened or shallowed. The range of α for peak discrimination sensitivity is concordant with the average α of natural scenes, which may indicate that visual mechanisms are optimized to process information at α values commonly encountered in the environment. Here we explore the association between peak discrimination sensitivity and the most viewed αs in natural environments. Specifically, we verified whether discrimination sensitivity depends on the recently viewed environments. Observers were immersed, using a Head-Mounted Display, in an environment that was either unaltered or had its average α steepened or shallowed by 0.4. Discrimination thresholds were affected by the average shift in α, but this effect was most prominent following adaptation to a shallowed environment. We modeled these data with a Bayesian observer and explored whether a change in the prior or a change in the likelihood best explained the psychophysical effects. Change in discrimination thresholds following adaptation could be explained by a shift in the central tendency of the prior concordant with the shift of the environment, in addition to a change in the likelihood. Our findings suggest that expectations on the occurrence of α that result from a lifetime of exposure remain plastic and able to accommodate for the statistical structure of recently viewed environments.

Machine learning modeling practices to support the principles of AI and ethics in nutrition research.

BACKGROUND: Nutrition research is relying more on artificial intelligence and machine learning models to understand, diagnose, predict, and explain data. While artificial intelligence and machine learning models provide powerful modeling tools, failure to use careful and well-thought-out modeling processes can lead to misleading conclusions and concerns surrounding ethics and bias. METHODS: Based on our experience as reviewers and journal editors in nutrition and obesity, we identified the most frequently omitted best practices from statistical modeling and how these same practices extend to machine learning models. We next addressed areas required for implementation of machine learning that are not included in commercial software packages. RESULTS: Here, we provide a tutorial on best artificial intelligence and machine learning modeling practices that can reduce potential ethical problems with a checklist and guiding principles to aid nutrition researchers in developing, evaluating, and implementing artificial intelligence and machine learning models in nutrition research. CONCLUSION: The quality of AI/ML modeling in nutrition research requires iterative and tailored processes to mitigate against potential ethical problems or to predict conclusions that are free of bias.

"There's something inside": Children's intuitions about animate agents.

From infancy, humans have the ability to distinguish animate agents from inert objects, and preschoolers map biological and mechanical insides to their appropriate kinds. However, less is known about how identifying something as an animate agent shapes specific inferences about its internal properties. Here, we test whether preschool children (N = 92; North American population) have specifically biological expectations about animate agents, or if they have more general expectations that animate agents should have an internal source of motion. We presented preschoolers with videos of two puppets: a "self-propelled" fur-covered puppet, and a fur-covered puppet that is seen to be moved by a human actor. In addition, we presented preschoolers with images of a familiar artifact (motorcycle) and familiar animal (sheep). For each item, we asked them to choose what they thought was inside each of these entities: nothing, biological insides, or mechanical insides. Preschoolers were less likely to say that a self-propelled fur-covered object was empty, compared to a fur-covered object that was moved by a human actor, which converges with past work with infants. However, preschoolers showed no specifically biological expectations about these objects, despite being able to accurately match biological insides to familiar animals and mechanical insides to familiar artifacts on the follow-up measure. These results suggest that preschoolers do not have specifically biological expectations about animate agents as a category, but rather general expectations that such agents should not be empty inside.