A learning system for adjustment processes based on human sensory perceptions

Creating, designing and adjusting products are essential decision processes underlying creative industries, such as painting, perfume, food and beverage industries. These processes require the participation and continuous supervision of professionals with highly-developed expert sensory abilities. Training of these experts is very complex due to the difficulty of transmitting intuitive knowledge obtained from perception. A new methodology for capturing this sensory expert knowledge that relies on a machine learning tool, previously trained with ‘state-action’ type patterns, jointly with an actions generator module, is proposed in this work. The method is based on a closed loop architecture together with the decomposition of complex sensory knowledge into basic elements capable of being handled by standard machine learning systems. A real case application to color-adjustment in the automotive paint manufacturing industry is presented showing the potential benefits of the method.     

Event knowledge,elaborative propensity,and the development of learning proficiency

An experimental test was made of two hypotheses formulated to account for age differences across adolescence in the learning of arbitrary associations. One hypothesis ascribes such differences to two factors: the propensity to elaborate coherent relationships among initially disparate items, and accessibility to event knowledge that can form the bases of such relationships. The other hypothesis assumes that propensity remains constant across age, and that development stems entirely from increases in the accessibility of relevant event knowledge. These hypotheses are evaluated with reference to the performance of 11- and 17-year-olds in learning relationships among paired nouns. The results discredited both hypotheses, instigating the formulation of a revised conception of the relationship between knowledge and propensity as developmental determinants.     

Curiosity-driven recommendation strategy for adaptive learning via deep reinforcement learning

The design of recommendation strategies in the adaptive learning systems focuses on utilizing currently available information to provide learners with individual-specific learning instructions. As a critical motivate for human behaviours, curiosity is essentially the drive to explore knowledge and seek information. In a psychologically inspired view, we propose a curiosity-driven recommendation policy within the reinforcement learning framework, allowing for an efficient and enjoyable personalized learning path. Specifically, a curiosity reward from a well-designed predictive model is generated to model one's familiarity with the knowledge space. Given such curiosity rewards, we apply the actor–critic method to approximate the policy directly through neural networks. Numerical analyses with a large continuous knowledge state space and concrete learning scenarios are provided to further demonstrate the efficiency of the proposed method.     

Cognitive biases in human causal learning

The main aim of this work was to look for cognitive biases in human inference of causal relationships in order to emphasize the psychological processes that modulate causal learning. From the effect of the judgment frequency, this work presents subsequent research on cue competition (overshadowing, blocking, and super-conditioning effects) showing that the strength of prior beliefs and new evidence based upon covariation computation contributes additively to predict causal judgments, whereas the balance between the reliability of both, beliefs and covariation knowledge, modulates their relative weight. New findings also showed "inattentional blindness" for negative or preventative causal relationships but not for positive or generative ones, due to failure in codifying and retrieving the necessary information for its computation. Overall results unveil the need of three hierarchical levels of a whole architecture for human causal learning: the lower one, responsible for codifying the events during the task; the second one, computing the retrieved information; finally, the higher level, integrating this evidence with previous causal knowledge. In summary, whereas current theoretical frameworks on causal inference and decision-making usually focused either on causal beliefs or covariation information, the present work shows how both are required to be able to explain the complexity and flexibility involved in human causal learning.     

Evolution of learning strategies in changing environments

Learning is an important aspect of cognition that is crucial for the success of many species, and has been a factor involved in the evolution of distinct patterns of life history that depend on the environments in question. The extent to which different degrees of social and individual learning emerge follows from various species-dependent factors, such as the fidelity of information transmission between individuals, and that has previously been modelled in agent-based simulations with meme-based representations of learned knowledge and behaviours. A limitation of that previous work is that it was based on fixed environments, and it is known that different learning strategies will emerge depending on the variability of the environment. This paper will address that limitation by extending the existing modelling framework to allow the simulation of life history evolution and the emergence of appropriate learning strategies in changing environments.     

Metacognitive monitoring during category learning: how success affects future behaviour

The purpose of this study was to see how people perceive their own learning during a category learning task, and whether their perceptions matched their performance. In two experiments, participants were asked to learn natural categories, of both high and low variability, and make category learning judgements (CLJs). Variability was manipulated by varying the number of exemplars and the number of times each exemplar was presented within each category. Experiment 1 showed that participants were generally overconfident in their knowledge of low variability families, suggesting that they considered repetition to be more useful for learning than it actually was. Also, a correct trial, for a particular category, was more likely to occur if the previous trial was correct. CLJs had the largest increase when a trial was correct following an incorrect trial and the largest decrease when an incorrect trial followed a correct trial. Experiment 2 replicated these results, but also demonstrated that global CLJ ratings showed the same bias towards repetition. These results indicate that we generally identify success as being the biggest determinant of learning, but do not always recognise cues, such as variability, that enhance learning.     

Adaptive sampling for active learning with genetic programming

Active learning is a machine learning paradigm allowing to decide which inputs to use for training. It is introduced to Genetic Programming (GP) essentially thanks to the dynamic data sampling, used to address some known issues such as the computational cost, the over-fitting problem and the imbalanced databases. The traditional dynamic sampling for GP gives to the algorithm a new sample periodically, often each generation, without considering the state of the evolution. In so doing, individuals do not have enough time to extract the hidden knowledge. An alternative approach is to use some information about the learning state to adapt the periodicity of the training data change. In this work, we propose an adaptive sampling strategy for classification tasks based on the state of solved fitness cases throughout learning. It is a flexible approach that could be applied with any dynamic sampling. We implemented some sampling algorithms extended with dynamic and adaptive controlling re-sampling frequency. We experimented them to solve the KDD intrusion detection and the Adult incomes prediction problems with GP. The experimental study demonstrates how the sampling frequency control preserves the power of dynamic sampling with possible improvements in learning time and quality. We also demonstrate that adaptive sampling can be an alternative to multi-level sampling. This work opens many new relevant extension paths.     

Visual feature learning in artificial grammar classification

The Artificial Grammar Learning task has been used extensively to assess individuals' implicit learning capabilities. Previous work suggests that participants implicitly acquire rule-based knowledge as well as exemplar-specific knowledge in this task. This study investigated whether exemplar-specific knowledge acquired in this task is based on the visual features of the exemplars. When a change in the font and case occurred between study and test, there was no effect on sensitivity to grammatical rules in classification judgments. However, such a change did virtually eliminate sensitivity to training frequencies of letter bigrams and trigrams (chunk strength) in classification judgments. Performance of a secondary task during study eliminated this font sensitivity and generally reduced the contribution of chunk strength knowledge. The results are consistent with the idea that perceptual fluency makes a contribution to artificial grammar judgments.     

Selective social learning in infancy: looking for mechanisms

Although there is mounting evidence that selective social learning begins in infancy, the psychological mechanisms underlying this ability are currently a controversial issue. The purpose of this study is to investigate whether theory of mind abilities and statistical learning skills are related to infants’ selective social learning. Seventy‐seven 18‐month‐olds were first exposed to a reliable or an unreliable speaker and then completed a word learning task, two theory of mind tasks, and a statistical learning task. If domain‐general abilities are linked to selective social learning, then infants who demonstrate superior performance on the statistical learning task should perform better on the selective learning task, that is, should be less likely to learn words from an unreliable speaker. Alternatively, if domain‐specific abilities are involved, then superior performance on theory of mind tasks should be related to selective learning performance. Findings revealed that, as expected, infants were more likely to learn a novel word from a reliable speaker. Importantly, infants who passed a theory of mind task assessing knowledge attribution were significantly less likely to learn a novel word from an unreliable speaker compared to infants who failed this task. No such effect was observed for the other tasks. These results suggest that infants who possess superior social‐cognitive abilities are more apt to reject an unreliable speaker as informant. A video abstract of this article can be viewed at: https://youtu.be/zuuCniHYzqo     

Self‐perceived competence,learning conceptions and preferred learning situations in the domain of communication

Background: An important purpose of education in the field of social work is the development of social‐communicative competence and students' individual learning theories (ILTs) concerning this domain. Aims: Our first aim was to develop diagnostic instruments for ILT assessment and to understand the relationships between ILT variables. Our second purpose was to study the differences in ILT variables between students of three study years. Samples: A total of 396 full‐time social work students participated in this study: 176 first‐year, 147 second‐year and 73 fourth‐year students (92% women and 8% men). Method: Based on a theoretical framework, three questionnaires have been constructed, covering three ILT variables: self‐perceived competence, learning conceptions and preferred learning situations. For scale construction, principal component analyses and reliability analyses were conducted. ANOVAs and post hoc comparisons of means were used to investigate cross‐sectional differences regarding ILT variables. Pearson correlations and regression analyses were performed to gain more insight into the relationships between ILT variables. Results: Five aspects of self‐perceived competence, four learning conceptions and five preferred learning situations were found. Learning conceptions and self‐perceived competencies were found to be predictors of students' preferred learning situations. Many differences were found between the three groups of students, especially between the first‐year students and the others. Conclusions: When studying the acquisition of social‐communicative competence, it is important to take students' individual learning theories into account. Increased insight into the role ILTs play can be of help in improving social work education.     

Reflecting talks: a pedagogical model in the learning organization

The aim of the article is to deepen the understanding of how a pedagogical model for reflecting talks can be used in order to make sustainable learning part of the daily work in the learning organization. From an interactive research approach, we have together with a project management group in a European Social Fund project worked with sustainable learning and knowledge development. Empirical data has been collected at the implementation of ten reflecting talks about sustainable equality. The results of the study lead to a strategy for how sustainable learning can become part of the daily work at a workplace. The strategy is constituted by a pedagogical model for reflecting talks, which clearly shows how sustainable learning in an organization can be structured. The core of the pedagogical model for the reflecting talks where both practically applied and theoretically anchored knowledge are important components. The learning process is based on observation, reflection, analysis and discussion of concrete situations/events. The models rests on four basic conditions; pedagogical competence, a delimited problem area, the learning group and timeframes. The model can be used in the daily work at short dialogues or at more penetrating discussions.     

Only correlated sequences that are actively processed contribute to implicit sequence learning

The purpose of the study was to investigate how implicit sequence learning is affected by the presence of secondary information that is correlated with the primary sequence but not necessarily relevant to performance. In a previous work, we have shown that correlation plays an important role but other prerequisites may also be involved. In Experiments 1 and 2, using a task sequence learning paradigm, we found that primary sequence learning was not affected by secondary information that was sequenced but irrelevant to performance, even though the two streams of information were correlated. In contrast, in Experiment 3, we found that sensitivity to the main sequence was greater with the provision of extra sequenced information that was relevant to performance in addition to being correlated. This suggests that sequence learning was enhanced through the integration of information. We conclude that information in secondary as well as primary sequences must be actively processed if it is to have a beneficial impact. By actively processed we mean information that is selectively attended and necessary for carrying out the tasks.     

A reinforcement learning approach to personalized learning recommendation systems

Personalized learning refers to instruction in which the pace of learning and the instructional approach are optimized for the needs of each learner. With the latest advances in information technology and data science, personalized learning is becoming possible for anyone with a personal computer, supported by a data-driven recommendation system that automatically schedules the learning sequence. The engine of such a recommendation system is a recommendation strategy that, based on data from other learners and the performance of the current learner, recommends suitable learning materials to optimize certain learning outcomes. A powerful engine achieves a balance between making the best possible recommendations based on the current knowledge and exploring new learning trajectories that may potentially pay off. Building such an engine is a challenging task. We formulate this problem within the Markov decision framework and propose a reinforcement learning approach to solving the problem.     

Evolution and learning: An epistemological perspective

The deep formal and conceptual link existing between artificial life and artificial intelligence can be highlighted using conceptual tools derived by Karl Popper's evolutionary epistemology. Starting from the observation that the structure itself of an organism embodies knowledge about the environment which it is adapted to, it is possible to regard evolution as a learning process. This process is subject to the same rules indicated by Popper for the growth of scientific knowledge: causal conjectures (mutations) and successive refutations (extinction). In the field of machine learning such a paradigm is represented by genetic algorithms that, simulating biological processes, emulate cognitive processes. From a practical viewpoint, that perspective allows to identify the two different kinds of learning considered by artificial intelligence, knowledge acquisition and skill improvement, and to get a different view of the problem of heuristic knowledge in learning systems. From a theoretical point of view, these considerations can shade a new light on an old epistemological problem: why do we live in a learnable world?     

Incidental sequence learning across the lifespan

The purpose of the present study was to investigate incidental sequence learning across the lifespan. We tested 50 children (aged 7-16), 50 young adults (aged 20-30), and 50 older adults (aged >65) with a sequence learning paradigm that involved both a task and a response sequence. After several blocks of practice, all age groups slowed down when the training sequences were removed, providing indirect evidence for sequence learning. This performance slowing was comparable between groups, indicating no age-related differences. However, when explicit sequence knowledge was considered, age effects were found. For both children and older adults with no or only little explicit knowledge, incidental sequence learning was largely reduced and statistically not significant. In contrast, young adults showed sequence learning irrespective of the amount of explicit knowledge. These results indicate that different learning processes are involved in incidental sequence learning depending on age.     

Implicit visual learning and the expression of learning

Although the existence of implicit motor learning is now widely accepted, the findings concerning perceptual implicit learning are ambiguous. Some researchers have observed perceptual learning whereas other authors have not. The review of the literature provides different reasons to explain this ambiguous picture, such as differences in the underlying learning processes, selective attention, or differences in the difficulty to express this knowledge. In three experiments, we investigated implicit visual learning within the original serial reaction time task. We used different response devices (keyboard vs. mouse) in order to manipulate selective attention towards response dimensions. Results showed that visual and motor sequence learning differed in terms of RT-benefits, but not in terms of the amount of knowledge assessed after training. Furthermore, visual sequence learning was modulated by selective attention. However, the findings of all three experiments suggest that selective attention did not alter implicit but rather explicit learning processes.     

Probabilities in implicit learning

The visual system possesses a remarkable ability in learning regularities from the environment. In the case of contextual cuing, predictive visual contexts such as spatial configurations are implicitly learned, retained, and used to facilitate visual search-all without one's subjective awareness and conscious effort. Here we investigated whether implicit learning and its facilitatory effects are sensitive to the statistical property of such implicit knowledge. In other words, are highly probable events learned better than less probable ones even when such learning is implicit? We systematically varied the frequencies of context repetition to alter the degrees of learning. Our results showed that search efficiency increased consistently as contextual probabilities increased. Thus, the visual contexts, along with their probability of occurrences, were both picked up by the visual system. Furthermore, even when the total number of exposures was held constant between each probability, the highest probability still enjoyed a greater cuing effect, suggesting that the temporal aspect of implicit learning is also an important factor to consider in addition to the effect of mere frequency. Together, these findings suggest that implicit learning, although bypassing observers' conscious encoding and retrieval effort, behaves much like explicit learning in the sense that its facilitatory effect also varies as a function of its associative strengths.     

Role of prior knowledge in implicit and explicit learning of artificial grammars

Artificial grammar learning (AGL) performance reflects both implicit and explicit processes and has typically been modeled without incorporating any influence from general world knowledge. Our research provides a systematic investigation of the implicit vs. explicit nature of general knowledge and its interaction with knowledge types investigated by past AGL research (i.e., rule- and similarity-based knowledge). In an AGL experiment, a general knowledge manipulation involved expectations being either congruent or incongruent with training stimulus structure. Inconsistent observations paradoxically led to an advantage in structural knowledge and in the use of general world knowledge in both explicit (conscious) and implicit (unconscious) cases (as assessed by subjective measures). The above findings were obtained under conditions of reduced processing time and impaired executive resources. Key findings from our work are that implicit AGL can clearly be affected by general knowledge, and implicit learning can be enhanced by the violation of expectations.     

Social learning and the etiology of autism

A developmental theory of autism is presented as an alternative to current nativist theories. The traits of autism are seen as results of failure of the process of social learning. The distinction between social learning and ontogenic discovery is discussed and a model of normal social learning is presented, showing its cyclical nature and significance. Recognized traits of autism are briefly described and classified, and it is shown how the main categories of traits could result from defective social learning, while defective social learning itself could be the result of a variety of causes, some of which might also affect other aspects of intelligence. The theory provides a framework into which current knowledge of autism can be integrated.     

The generation of conscious awareness in an incidental learning situation

The purpose of the present experiments was to investigate the generation of conscious awareness (i.e., of verbal report) in an incidental learning situation. While the single-system account assumes that all markers of learning, verbal or nonverbal, index the same underlying knowledge representation, multiple-systems accounts grant verbal report a special status as a marker of learning because they assume that the nonverbal and verbal effects of learning rely on different memory representations. We tested these two accounts in two experiments in which we held the amount of learning in the nonverbal memory system constant while manipulating independent variables aimed at affecting learning in the declarative system. The results of both experiments revealed significant differences in verbal report between experimental conditions, but no significant differences in response times. Overall, these results provide clear evidence in favor of the multiple-systems account.