The Experts below are selected from a list of 183 Experts worldwide ranked by ideXlab platform

Scott P Johnson - One of the best experts on this subject based on the ideXlab platform.

  • anticipating an effect from predictive visual sequences development of infants causal Inference from 9 to 18 months
    Cognitive Science, 2014
    Co-Authors: Jeffrey K Bye, Bryan Nguyen, Scott P Johnson
    Abstract:

    Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months Jeffrey K. Bye 1 (jkbye@ucla.edu), Bryan D. Nguyen 1 (bnguyen07@ucla.edu), Hongjing Lu 1,2 (hongjing@ucla.edu), Scott P. Johnson 1,3 (scott.johnson@ucla.edu) Departments of Psychology 1 , Statistics 2 , and Psychiatry and Biobehavioral Sciences 3 University of California, Los Angeles; Los Angeles, CA, USA Abstract Causal Inference in Infants Teinonen et al., 2009) and motion-based causal perception given spatial and temporal contiguity (Leslie & Keeble, 1987). By 18-24 months, children begin to expect predictable effects from learned sequences of discrete causal events (Bonawitz et al., 2010), and by 24-48 months, they routinely use the covariation patterns between actions and state changes to categorize objects by causal power and predict the effects of causal actions (Gopnik & Sobel, 2000). How do infants progress from the perceptual mechanisms of early infancy (statistical and rule-based learning and motion-based causal perception) to the extraction of causal structure from temporal covariation between discrete events? Little is known about learning processes in infancy that support causal Inference. Research on causal Inference in children under 18 months of age has been limited because the tasks typically used, such as the “blicket” detector paradigm (Gopnik & Sobel, 2000), require vocal ability and/or motor skills. It has been difficult, therefore, to characterize infants’ development of causal learning, leaving open the issue of whether early motion-based causal perception is generalized to higher-order causal Inference (Michotte, 1946/1963), or whether the former is merely a special case of the latter (Cheng, 1993). Causal Inferences are necessarily constrained by limits in perception, attention, and memory. In particular, potential causal cues and their temporal ordering must be determined before statistical computation can proceed. The literature on animal conditioning is consistent with an inherent connection between detection of causal cues and causal Inference. For example, Balleine et al. (2005) showed that increased perceptual discriminability of cues enhances rats’ ability to acquire cause-effect associations, influencing sophisticated behavior such as retrospective revaluation of cues. We would expect, therefore, that development of causal Inference is constrained by infants’ perceptual processing maturity and working memory capacity, as well as the identification of conditional probabilities involved in predicting an effect. Our experiment and our model represent a first step in understanding developments in causal Inferences from probabilistic information during this important transitional age range, as well as the perceptual and cognitive constraints on the learning process. Previous Research has shown that infants make remarkable progress in acquiring sophisticated inferential abilities in the first two years after birth. Within the first few months, for example, infants demonstrate statistical learning in multiple sensory modalities (Frank et al., 2009; Kirkham, Slemmer, & Johnson, 2002; Saffran, Aslin, & Newport, 1996; An experiment by Buchsbaum et al. (2011) with 41- to 70- month-old children provides a paradigm that we adapted to create a causal Inference task that can be performed by younger infants. The children in Buchsbaum et al.’s study There has been little Research on infants’ development of causal Inference in the second year after birth. We report an experiment in which 9- to 18-month-old infants viewed visual sequences consisting of three looming shapes, one after another. Half of the sequences (causes) were predictive of an attention-getting reward (effect), and the other half were non- predictive. The statistical complexity of predictive sequences was varied between conditions. We analyzed latencies of infants’ eye movements toward the reward location. Older infants yielded more anticipatory eye movements in predictive than non-predictive sequences. Effects of both infant age and complexity of causal sequences were observed. To qualitatively account for these findings, we formulated a Bayesian model based on generic priors favoring simple causal events coupled with noisy shape identification. Keywords: causal Inference; modeling; perception; infants development; Bayesian Introduction Understanding cause-effect relations is vital to cognitive development. Imagine, for example, an infant attempting to disambiguate his mother’s actions as she starts up the car, adjusts the mirror, scans the surroundings, engages first gear, and turns the steering wheel, at which point the car moves. How do infants make sense of such causal action sequences? How do they determine which actions are necessary, and in which order, to produce an effect? How do they use this knowledge to guide their own actions? We address these questions with a combination of behavioral and computational evidence. We report an experiment in which infants observe causal action sequences for which cause-effect relations are specified by conditional probabilities of varying complexity, and we describe a Bayesian model that simulates the Inferences that support identification of causal structure. We reasoned that identifying causal structure incorporates multiple sources of information (e.g., spatiotemporal, statistical) and perceptual/cognitive mechanisms (e.g., visual attention, detection of serial order, working memory, inherent biases) that must operate in concert during the learning process. Paradigm

  • CogSci - Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months
    Cognitive Science, 2014
    Co-Authors: Jeffrey K Bye, Bryan Nguyen, Scott P Johnson
    Abstract:

    Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months Jeffrey K. Bye 1 (jkbye@ucla.edu), Bryan D. Nguyen 1 (bnguyen07@ucla.edu), Hongjing Lu 1,2 (hongjing@ucla.edu), Scott P. Johnson 1,3 (scott.johnson@ucla.edu) Departments of Psychology 1 , Statistics 2 , and Psychiatry and Biobehavioral Sciences 3 University of California, Los Angeles; Los Angeles, CA, USA Abstract Causal Inference in Infants Teinonen et al., 2009) and motion-based causal perception given spatial and temporal contiguity (Leslie & Keeble, 1987). By 18-24 months, children begin to expect predictable effects from learned sequences of discrete causal events (Bonawitz et al., 2010), and by 24-48 months, they routinely use the covariation patterns between actions and state changes to categorize objects by causal power and predict the effects of causal actions (Gopnik & Sobel, 2000). How do infants progress from the perceptual mechanisms of early infancy (statistical and rule-based learning and motion-based causal perception) to the extraction of causal structure from temporal covariation between discrete events? Little is known about learning processes in infancy that support causal Inference. Research on causal Inference in children under 18 months of age has been limited because the tasks typically used, such as the “blicket” detector paradigm (Gopnik & Sobel, 2000), require vocal ability and/or motor skills. It has been difficult, therefore, to characterize infants’ development of causal learning, leaving open the issue of whether early motion-based causal perception is generalized to higher-order causal Inference (Michotte, 1946/1963), or whether the former is merely a special case of the latter (Cheng, 1993). Causal Inferences are necessarily constrained by limits in perception, attention, and memory. In particular, potential causal cues and their temporal ordering must be determined before statistical computation can proceed. The literature on animal conditioning is consistent with an inherent connection between detection of causal cues and causal Inference. For example, Balleine et al. (2005) showed that increased perceptual discriminability of cues enhances rats’ ability to acquire cause-effect associations, influencing sophisticated behavior such as retrospective revaluation of cues. We would expect, therefore, that development of causal Inference is constrained by infants’ perceptual processing maturity and working memory capacity, as well as the identification of conditional probabilities involved in predicting an effect. Our experiment and our model represent a first step in understanding developments in causal Inferences from probabilistic information during this important transitional age range, as well as the perceptual and cognitive constraints on the learning process. Previous Research has shown that infants make remarkable progress in acquiring sophisticated inferential abilities in the first two years after birth. Within the first few months, for example, infants demonstrate statistical learning in multiple sensory modalities (Frank et al., 2009; Kirkham, Slemmer, & Johnson, 2002; Saffran, Aslin, & Newport, 1996; An experiment by Buchsbaum et al. (2011) with 41- to 70- month-old children provides a paradigm that we adapted to create a causal Inference task that can be performed by younger infants. The children in Buchsbaum et al.’s study There has been little Research on infants’ development of causal Inference in the second year after birth. We report an experiment in which 9- to 18-month-old infants viewed visual sequences consisting of three looming shapes, one after another. Half of the sequences (causes) were predictive of an attention-getting reward (effect), and the other half were non- predictive. The statistical complexity of predictive sequences was varied between conditions. We analyzed latencies of infants’ eye movements toward the reward location. Older infants yielded more anticipatory eye movements in predictive than non-predictive sequences. Effects of both infant age and complexity of causal sequences were observed. To qualitatively account for these findings, we formulated a Bayesian model based on generic priors favoring simple causal events coupled with noisy shape identification. Keywords: causal Inference; modeling; perception; infants development; Bayesian Introduction Understanding cause-effect relations is vital to cognitive development. Imagine, for example, an infant attempting to disambiguate his mother’s actions as she starts up the car, adjusts the mirror, scans the surroundings, engages first gear, and turns the steering wheel, at which point the car moves. How do infants make sense of such causal action sequences? How do they determine which actions are necessary, and in which order, to produce an effect? How do they use this knowledge to guide their own actions? We address these questions with a combination of behavioral and computational evidence. We report an experiment in which infants observe causal action sequences for which cause-effect relations are specified by conditional probabilities of varying complexity, and we describe a Bayesian model that simulates the Inferences that support identification of causal structure. We reasoned that identifying causal structure incorporates multiple sources of information (e.g., spatiotemporal, statistical) and perceptual/cognitive mechanisms (e.g., visual attention, detection of serial order, working memory, inherent biases) that must operate in concert during the learning process. Paradigm

Jeffrey K Bye - One of the best experts on this subject based on the ideXlab platform.

  • anticipating an effect from predictive visual sequences development of infants causal Inference from 9 to 18 months
    Cognitive Science, 2014
    Co-Authors: Jeffrey K Bye, Bryan Nguyen, Scott P Johnson
    Abstract:

    Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months Jeffrey K. Bye 1 (jkbye@ucla.edu), Bryan D. Nguyen 1 (bnguyen07@ucla.edu), Hongjing Lu 1,2 (hongjing@ucla.edu), Scott P. Johnson 1,3 (scott.johnson@ucla.edu) Departments of Psychology 1 , Statistics 2 , and Psychiatry and Biobehavioral Sciences 3 University of California, Los Angeles; Los Angeles, CA, USA Abstract Causal Inference in Infants Teinonen et al., 2009) and motion-based causal perception given spatial and temporal contiguity (Leslie & Keeble, 1987). By 18-24 months, children begin to expect predictable effects from learned sequences of discrete causal events (Bonawitz et al., 2010), and by 24-48 months, they routinely use the covariation patterns between actions and state changes to categorize objects by causal power and predict the effects of causal actions (Gopnik & Sobel, 2000). How do infants progress from the perceptual mechanisms of early infancy (statistical and rule-based learning and motion-based causal perception) to the extraction of causal structure from temporal covariation between discrete events? Little is known about learning processes in infancy that support causal Inference. Research on causal Inference in children under 18 months of age has been limited because the tasks typically used, such as the “blicket” detector paradigm (Gopnik & Sobel, 2000), require vocal ability and/or motor skills. It has been difficult, therefore, to characterize infants’ development of causal learning, leaving open the issue of whether early motion-based causal perception is generalized to higher-order causal Inference (Michotte, 1946/1963), or whether the former is merely a special case of the latter (Cheng, 1993). Causal Inferences are necessarily constrained by limits in perception, attention, and memory. In particular, potential causal cues and their temporal ordering must be determined before statistical computation can proceed. The literature on animal conditioning is consistent with an inherent connection between detection of causal cues and causal Inference. For example, Balleine et al. (2005) showed that increased perceptual discriminability of cues enhances rats’ ability to acquire cause-effect associations, influencing sophisticated behavior such as retrospective revaluation of cues. We would expect, therefore, that development of causal Inference is constrained by infants’ perceptual processing maturity and working memory capacity, as well as the identification of conditional probabilities involved in predicting an effect. Our experiment and our model represent a first step in understanding developments in causal Inferences from probabilistic information during this important transitional age range, as well as the perceptual and cognitive constraints on the learning process. Previous Research has shown that infants make remarkable progress in acquiring sophisticated inferential abilities in the first two years after birth. Within the first few months, for example, infants demonstrate statistical learning in multiple sensory modalities (Frank et al., 2009; Kirkham, Slemmer, & Johnson, 2002; Saffran, Aslin, & Newport, 1996; An experiment by Buchsbaum et al. (2011) with 41- to 70- month-old children provides a paradigm that we adapted to create a causal Inference task that can be performed by younger infants. The children in Buchsbaum et al.’s study There has been little Research on infants’ development of causal Inference in the second year after birth. We report an experiment in which 9- to 18-month-old infants viewed visual sequences consisting of three looming shapes, one after another. Half of the sequences (causes) were predictive of an attention-getting reward (effect), and the other half were non- predictive. The statistical complexity of predictive sequences was varied between conditions. We analyzed latencies of infants’ eye movements toward the reward location. Older infants yielded more anticipatory eye movements in predictive than non-predictive sequences. Effects of both infant age and complexity of causal sequences were observed. To qualitatively account for these findings, we formulated a Bayesian model based on generic priors favoring simple causal events coupled with noisy shape identification. Keywords: causal Inference; modeling; perception; infants development; Bayesian Introduction Understanding cause-effect relations is vital to cognitive development. Imagine, for example, an infant attempting to disambiguate his mother’s actions as she starts up the car, adjusts the mirror, scans the surroundings, engages first gear, and turns the steering wheel, at which point the car moves. How do infants make sense of such causal action sequences? How do they determine which actions are necessary, and in which order, to produce an effect? How do they use this knowledge to guide their own actions? We address these questions with a combination of behavioral and computational evidence. We report an experiment in which infants observe causal action sequences for which cause-effect relations are specified by conditional probabilities of varying complexity, and we describe a Bayesian model that simulates the Inferences that support identification of causal structure. We reasoned that identifying causal structure incorporates multiple sources of information (e.g., spatiotemporal, statistical) and perceptual/cognitive mechanisms (e.g., visual attention, detection of serial order, working memory, inherent biases) that must operate in concert during the learning process. Paradigm

  • CogSci - Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months
    Cognitive Science, 2014
    Co-Authors: Jeffrey K Bye, Bryan Nguyen, Scott P Johnson
    Abstract:

    Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months Jeffrey K. Bye 1 (jkbye@ucla.edu), Bryan D. Nguyen 1 (bnguyen07@ucla.edu), Hongjing Lu 1,2 (hongjing@ucla.edu), Scott P. Johnson 1,3 (scott.johnson@ucla.edu) Departments of Psychology 1 , Statistics 2 , and Psychiatry and Biobehavioral Sciences 3 University of California, Los Angeles; Los Angeles, CA, USA Abstract Causal Inference in Infants Teinonen et al., 2009) and motion-based causal perception given spatial and temporal contiguity (Leslie & Keeble, 1987). By 18-24 months, children begin to expect predictable effects from learned sequences of discrete causal events (Bonawitz et al., 2010), and by 24-48 months, they routinely use the covariation patterns between actions and state changes to categorize objects by causal power and predict the effects of causal actions (Gopnik & Sobel, 2000). How do infants progress from the perceptual mechanisms of early infancy (statistical and rule-based learning and motion-based causal perception) to the extraction of causal structure from temporal covariation between discrete events? Little is known about learning processes in infancy that support causal Inference. Research on causal Inference in children under 18 months of age has been limited because the tasks typically used, such as the “blicket” detector paradigm (Gopnik & Sobel, 2000), require vocal ability and/or motor skills. It has been difficult, therefore, to characterize infants’ development of causal learning, leaving open the issue of whether early motion-based causal perception is generalized to higher-order causal Inference (Michotte, 1946/1963), or whether the former is merely a special case of the latter (Cheng, 1993). Causal Inferences are necessarily constrained by limits in perception, attention, and memory. In particular, potential causal cues and their temporal ordering must be determined before statistical computation can proceed. The literature on animal conditioning is consistent with an inherent connection between detection of causal cues and causal Inference. For example, Balleine et al. (2005) showed that increased perceptual discriminability of cues enhances rats’ ability to acquire cause-effect associations, influencing sophisticated behavior such as retrospective revaluation of cues. We would expect, therefore, that development of causal Inference is constrained by infants’ perceptual processing maturity and working memory capacity, as well as the identification of conditional probabilities involved in predicting an effect. Our experiment and our model represent a first step in understanding developments in causal Inferences from probabilistic information during this important transitional age range, as well as the perceptual and cognitive constraints on the learning process. Previous Research has shown that infants make remarkable progress in acquiring sophisticated inferential abilities in the first two years after birth. Within the first few months, for example, infants demonstrate statistical learning in multiple sensory modalities (Frank et al., 2009; Kirkham, Slemmer, & Johnson, 2002; Saffran, Aslin, & Newport, 1996; An experiment by Buchsbaum et al. (2011) with 41- to 70- month-old children provides a paradigm that we adapted to create a causal Inference task that can be performed by younger infants. The children in Buchsbaum et al.’s study There has been little Research on infants’ development of causal Inference in the second year after birth. We report an experiment in which 9- to 18-month-old infants viewed visual sequences consisting of three looming shapes, one after another. Half of the sequences (causes) were predictive of an attention-getting reward (effect), and the other half were non- predictive. The statistical complexity of predictive sequences was varied between conditions. We analyzed latencies of infants’ eye movements toward the reward location. Older infants yielded more anticipatory eye movements in predictive than non-predictive sequences. Effects of both infant age and complexity of causal sequences were observed. To qualitatively account for these findings, we formulated a Bayesian model based on generic priors favoring simple causal events coupled with noisy shape identification. Keywords: causal Inference; modeling; perception; infants development; Bayesian Introduction Understanding cause-effect relations is vital to cognitive development. Imagine, for example, an infant attempting to disambiguate his mother’s actions as she starts up the car, adjusts the mirror, scans the surroundings, engages first gear, and turns the steering wheel, at which point the car moves. How do infants make sense of such causal action sequences? How do they determine which actions are necessary, and in which order, to produce an effect? How do they use this knowledge to guide their own actions? We address these questions with a combination of behavioral and computational evidence. We report an experiment in which infants observe causal action sequences for which cause-effect relations are specified by conditional probabilities of varying complexity, and we describe a Bayesian model that simulates the Inferences that support identification of causal structure. We reasoned that identifying causal structure incorporates multiple sources of information (e.g., spatiotemporal, statistical) and perceptual/cognitive mechanisms (e.g., visual attention, detection of serial order, working memory, inherent biases) that must operate in concert during the learning process. Paradigm

Antti Tenhiala - One of the best experts on this subject based on the ideXlab platform.

  • contingency theory of capacity planning the link between process types and planning methods
    Journal of Operations Management, 2011
    Co-Authors: Antti Tenhiala
    Abstract:

    Abstract Although the reliability of production plans is crucial for the performance of manufacturing organizations, most practitioners use considerably simpler planning methods than what is recommended in the operations management literature. This article employs the contingency theory of organizations to explain the gap between the practice and the academic models of production planning. Arguments on the contingency effects of process complexity lead to a hypothesis that expects simple capacity planning methods to be most effective in certain production processes. A strong Inference Research setting is used to test the contingency hypothesis against a conventional hypothesis that expects the most sophisticated planning techniques to always be most effective. Multisource data from the machinery manufacturing industry support the contingency hypothesis and reject the universalistic hypothesis. The findings are explained using the concepts of task interdependence and bounded rationality. The results have several managerial implications, and they elaborate how classic concepts in organization theory can bring practically relevant insights to operations management Research.

Carol S. Dweck - One of the best experts on this subject based on the ideXlab platform.

  • “Meaningful” social Inferences: Effects of implicit theories on inferential processes
    Journal of Experimental Social Psychology, 2006
    Co-Authors: Daniel C. Molden, Jason E. Plaks, Carol S. Dweck
    Abstract:

    Abstract Perceivers’ shared theories about the social world have long featured prominently in social Inference Research. Here, we investigate how fundamental differences in such theories influence basic inferential processes. Past work has typically shown that integrating multiple interpretations of behavior during social Inference requires cognitive resources. However, three studies that measured or manipulated people’s beliefs about the stable versus dynamic nature of human attributes (i.e., their entity vs. incremental theory, respectively) qualify these past findings. Results revealed that, when interpreting others’ actions, perceivers’ theories selectively facilitate the consideration of interpretations that are especially theory-relevant. While experiencing cognitive load, entity theorists continued to incorporate information about stable dispositions (but not about dynamic social situations) in their social Inferences, whereas incremental theorists continued to incorporate information about dynamic social situations (but not about stable traits). Implications of these results for how perceivers find meaning in behavior are discussed.

Bryan Nguyen - One of the best experts on this subject based on the ideXlab platform.

  • anticipating an effect from predictive visual sequences development of infants causal Inference from 9 to 18 months
    Cognitive Science, 2014
    Co-Authors: Jeffrey K Bye, Bryan Nguyen, Scott P Johnson
    Abstract:

    Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months Jeffrey K. Bye 1 (jkbye@ucla.edu), Bryan D. Nguyen 1 (bnguyen07@ucla.edu), Hongjing Lu 1,2 (hongjing@ucla.edu), Scott P. Johnson 1,3 (scott.johnson@ucla.edu) Departments of Psychology 1 , Statistics 2 , and Psychiatry and Biobehavioral Sciences 3 University of California, Los Angeles; Los Angeles, CA, USA Abstract Causal Inference in Infants Teinonen et al., 2009) and motion-based causal perception given spatial and temporal contiguity (Leslie & Keeble, 1987). By 18-24 months, children begin to expect predictable effects from learned sequences of discrete causal events (Bonawitz et al., 2010), and by 24-48 months, they routinely use the covariation patterns between actions and state changes to categorize objects by causal power and predict the effects of causal actions (Gopnik & Sobel, 2000). How do infants progress from the perceptual mechanisms of early infancy (statistical and rule-based learning and motion-based causal perception) to the extraction of causal structure from temporal covariation between discrete events? Little is known about learning processes in infancy that support causal Inference. Research on causal Inference in children under 18 months of age has been limited because the tasks typically used, such as the “blicket” detector paradigm (Gopnik & Sobel, 2000), require vocal ability and/or motor skills. It has been difficult, therefore, to characterize infants’ development of causal learning, leaving open the issue of whether early motion-based causal perception is generalized to higher-order causal Inference (Michotte, 1946/1963), or whether the former is merely a special case of the latter (Cheng, 1993). Causal Inferences are necessarily constrained by limits in perception, attention, and memory. In particular, potential causal cues and their temporal ordering must be determined before statistical computation can proceed. The literature on animal conditioning is consistent with an inherent connection between detection of causal cues and causal Inference. For example, Balleine et al. (2005) showed that increased perceptual discriminability of cues enhances rats’ ability to acquire cause-effect associations, influencing sophisticated behavior such as retrospective revaluation of cues. We would expect, therefore, that development of causal Inference is constrained by infants’ perceptual processing maturity and working memory capacity, as well as the identification of conditional probabilities involved in predicting an effect. Our experiment and our model represent a first step in understanding developments in causal Inferences from probabilistic information during this important transitional age range, as well as the perceptual and cognitive constraints on the learning process. Previous Research has shown that infants make remarkable progress in acquiring sophisticated inferential abilities in the first two years after birth. Within the first few months, for example, infants demonstrate statistical learning in multiple sensory modalities (Frank et al., 2009; Kirkham, Slemmer, & Johnson, 2002; Saffran, Aslin, & Newport, 1996; An experiment by Buchsbaum et al. (2011) with 41- to 70- month-old children provides a paradigm that we adapted to create a causal Inference task that can be performed by younger infants. The children in Buchsbaum et al.’s study There has been little Research on infants’ development of causal Inference in the second year after birth. We report an experiment in which 9- to 18-month-old infants viewed visual sequences consisting of three looming shapes, one after another. Half of the sequences (causes) were predictive of an attention-getting reward (effect), and the other half were non- predictive. The statistical complexity of predictive sequences was varied between conditions. We analyzed latencies of infants’ eye movements toward the reward location. Older infants yielded more anticipatory eye movements in predictive than non-predictive sequences. Effects of both infant age and complexity of causal sequences were observed. To qualitatively account for these findings, we formulated a Bayesian model based on generic priors favoring simple causal events coupled with noisy shape identification. Keywords: causal Inference; modeling; perception; infants development; Bayesian Introduction Understanding cause-effect relations is vital to cognitive development. Imagine, for example, an infant attempting to disambiguate his mother’s actions as she starts up the car, adjusts the mirror, scans the surroundings, engages first gear, and turns the steering wheel, at which point the car moves. How do infants make sense of such causal action sequences? How do they determine which actions are necessary, and in which order, to produce an effect? How do they use this knowledge to guide their own actions? We address these questions with a combination of behavioral and computational evidence. We report an experiment in which infants observe causal action sequences for which cause-effect relations are specified by conditional probabilities of varying complexity, and we describe a Bayesian model that simulates the Inferences that support identification of causal structure. We reasoned that identifying causal structure incorporates multiple sources of information (e.g., spatiotemporal, statistical) and perceptual/cognitive mechanisms (e.g., visual attention, detection of serial order, working memory, inherent biases) that must operate in concert during the learning process. Paradigm

  • CogSci - Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months
    Cognitive Science, 2014
    Co-Authors: Jeffrey K Bye, Bryan Nguyen, Scott P Johnson
    Abstract:

    Anticipating an Effect from Predictive Visual Sequences: Development of Infants’ Causal Inference from 9 to 18 Months Jeffrey K. Bye 1 (jkbye@ucla.edu), Bryan D. Nguyen 1 (bnguyen07@ucla.edu), Hongjing Lu 1,2 (hongjing@ucla.edu), Scott P. Johnson 1,3 (scott.johnson@ucla.edu) Departments of Psychology 1 , Statistics 2 , and Psychiatry and Biobehavioral Sciences 3 University of California, Los Angeles; Los Angeles, CA, USA Abstract Causal Inference in Infants Teinonen et al., 2009) and motion-based causal perception given spatial and temporal contiguity (Leslie & Keeble, 1987). By 18-24 months, children begin to expect predictable effects from learned sequences of discrete causal events (Bonawitz et al., 2010), and by 24-48 months, they routinely use the covariation patterns between actions and state changes to categorize objects by causal power and predict the effects of causal actions (Gopnik & Sobel, 2000). How do infants progress from the perceptual mechanisms of early infancy (statistical and rule-based learning and motion-based causal perception) to the extraction of causal structure from temporal covariation between discrete events? Little is known about learning processes in infancy that support causal Inference. Research on causal Inference in children under 18 months of age has been limited because the tasks typically used, such as the “blicket” detector paradigm (Gopnik & Sobel, 2000), require vocal ability and/or motor skills. It has been difficult, therefore, to characterize infants’ development of causal learning, leaving open the issue of whether early motion-based causal perception is generalized to higher-order causal Inference (Michotte, 1946/1963), or whether the former is merely a special case of the latter (Cheng, 1993). Causal Inferences are necessarily constrained by limits in perception, attention, and memory. In particular, potential causal cues and their temporal ordering must be determined before statistical computation can proceed. The literature on animal conditioning is consistent with an inherent connection between detection of causal cues and causal Inference. For example, Balleine et al. (2005) showed that increased perceptual discriminability of cues enhances rats’ ability to acquire cause-effect associations, influencing sophisticated behavior such as retrospective revaluation of cues. We would expect, therefore, that development of causal Inference is constrained by infants’ perceptual processing maturity and working memory capacity, as well as the identification of conditional probabilities involved in predicting an effect. Our experiment and our model represent a first step in understanding developments in causal Inferences from probabilistic information during this important transitional age range, as well as the perceptual and cognitive constraints on the learning process. Previous Research has shown that infants make remarkable progress in acquiring sophisticated inferential abilities in the first two years after birth. Within the first few months, for example, infants demonstrate statistical learning in multiple sensory modalities (Frank et al., 2009; Kirkham, Slemmer, & Johnson, 2002; Saffran, Aslin, & Newport, 1996; An experiment by Buchsbaum et al. (2011) with 41- to 70- month-old children provides a paradigm that we adapted to create a causal Inference task that can be performed by younger infants. The children in Buchsbaum et al.’s study There has been little Research on infants’ development of causal Inference in the second year after birth. We report an experiment in which 9- to 18-month-old infants viewed visual sequences consisting of three looming shapes, one after another. Half of the sequences (causes) were predictive of an attention-getting reward (effect), and the other half were non- predictive. The statistical complexity of predictive sequences was varied between conditions. We analyzed latencies of infants’ eye movements toward the reward location. Older infants yielded more anticipatory eye movements in predictive than non-predictive sequences. Effects of both infant age and complexity of causal sequences were observed. To qualitatively account for these findings, we formulated a Bayesian model based on generic priors favoring simple causal events coupled with noisy shape identification. Keywords: causal Inference; modeling; perception; infants development; Bayesian Introduction Understanding cause-effect relations is vital to cognitive development. Imagine, for example, an infant attempting to disambiguate his mother’s actions as she starts up the car, adjusts the mirror, scans the surroundings, engages first gear, and turns the steering wheel, at which point the car moves. How do infants make sense of such causal action sequences? How do they determine which actions are necessary, and in which order, to produce an effect? How do they use this knowledge to guide their own actions? We address these questions with a combination of behavioral and computational evidence. We report an experiment in which infants observe causal action sequences for which cause-effect relations are specified by conditional probabilities of varying complexity, and we describe a Bayesian model that simulates the Inferences that support identification of causal structure. We reasoned that identifying causal structure incorporates multiple sources of information (e.g., spatiotemporal, statistical) and perceptual/cognitive mechanisms (e.g., visual attention, detection of serial order, working memory, inherent biases) that must operate in concert during the learning process. Paradigm