Context Of Task Research Articles

Avoiding Catastrophe: Active Dendrites Enable Multi-Task Learning in Dynamic Environments.

A key challenge for AI is to build embodied systems that operate in dynamically changing environments. Such systems must adapt to changing task contexts and learn continuously. Although standard deep learning systems achieve state of the art results on static benchmarks, they often struggle in dynamic scenarios. In these settings, error signals from multiple contexts can interfere with one another, ultimately leading to a phenomenon known as catastrophic forgetting. In this article we investigate biologically inspired architectures as solutions to these problems. Specifically, we show that the biophysical properties of dendrites and local inhibitory systems enable networks to dynamically restrict and route information in a context-specific manner. Our key contributions are as follows: first, we propose a novel artificial neural network architecture that incorporates active dendrites and sparse representations into the standard deep learning framework. Next, we study the performance of this architecture on two separate benchmarks requiring task-based adaptation: Meta-World, a multi-task reinforcement learning environment where a robotic agent must learn to solve a variety of manipulation tasks simultaneously; and a continual learning benchmark in which the model's prediction task changes throughout training. Analysis on both benchmarks demonstrates the emergence of overlapping but distinct and sparse subnetworks, allowing the system to fluidly learn multiple tasks with minimal forgetting. Our neural implementation marks the first time a single architecture has achieved competitive results in both multi-task and continual learning settings. Our research sheds light on how biological properties of neurons can inform deep learning systems to address dynamic scenarios that are typically impossible for traditional ANNs to solve.

Peer Assessment and Video Feedback for Fostering Self, Co, and Shared Regulation of Learning in a Higher Education Language Classroom

Self-regulated learning includes the cognitive, metacognitive, behavioral, motivational, and affective aspects of learning. The conceptualization of self and socially regulated learning has recently received much attention and peer assessment has been found to increase the use of metacognitive activity. The present exploratory qualitative study aimed to identify self-, co-, and socially shared regulatory processes in an oral English as a Foreign Language task. The regulatory activity deployed by 10 learners was studied within the context of a peer assessment task using an assessment form paired with video feedback in the context of an English language classroom at a French university. These interactions were filmed and discussed in individual self-confrontation interviews which were analyzed through inductive coding. Specific findings from the classroom setting shed light on existing gaps in the literature. First, students can gain confidence in their own skills through assessing their peers and activating regulatory processes both individually and as a group. Second, appropriate tools can increase co-regulated and socially regulated learning through the structuring of cooperative regulatory behaviors. Third, psychological safety appeared to be a propitious social context for supporting regulated learning (SRL, CoRL, and SSRL). We also shed light on the fact that adaptive regulatory strategies are present in oral (as well as written) English as a Foreign Language tasks. These results indicate the potential for learning situations based on video feedback used in conjunction with peer assessment and collaborative learning in order to develop regulatory behaviors in language learners.

Age and Framing Effects in the Balloon Analogue Risk Task.

Prior research has documented age differences in risky decisions and indicates that they are susceptible to gain versus loss framing. However, previous studies focused on "decisions from description" that explicitly spell out the probabilities involved. The present study expands this literature by examining the effects of framing on age differences in the Balloon Analogue Risk Task (BART), a widely used and ecologically valid measure of experience-based risky decision making that involves pumping a virtual balloon. In a preregistered study, younger (aged 18-30, n = 129) and older adults (aged 60 and older, n = 125) were randomly assigned to either a gain version of the BART, where pumping the balloon added monetary gains, or a loss version, where pumping the balloon avoided monetary losses. We found a significant age by frame interaction on risk-taking: in the loss frame, older adults pumped more frequently and experienced more popped balloons than younger adults, whereas in the gain frame no significant age differences were found. Total performance on the BART did not vary by age or frame. Supplementary analyses indicated that age differences in pumping rates were most pronounced at the beginning of the BART and leveled off in subsequent trials. Controlling for age differences in motivation, personality, and cognition did not account for age differences in risk-taking. In combination, findings suggest that age differences in risk-taking on the BART are more pronounced when the task context emphasizes avoiding losses rather than achieving gains.

Context meta-reinforcement learning via neuromodulation

Meta-reinforcement learning (meta-RL) algorithms enable agents to adapt quickly to tasks from few samples in dynamic environments. Such a feat is achieved through dynamic representations in an agent’s policy network (obtained via reasoning about task context, model parameter updates, or both). However, obtaining rich dynamic representations for fast adaptation beyond simple benchmark problems is challenging due to the burden placed on the policy network to accommodate different policies. This paper addresses the challenge by introducing neuromodulation as a modular component to augment a standard policy network that regulates neuronal activities in order to produce efficient dynamic representations for task adaptation. The proposed extension to the policy network is evaluated across multiple discrete and continuous control environments of increasing complexity. To prove the generality and benefits of the extension in meta-RL, the neuromodulated network was applied to two state-of-the-art meta-RL algorithms (CAVIA and PEARL). The result demonstrates that meta-RL augmented with neuromodulation produces significantly better result and richer dynamic representations in comparison to the baselines.

Robust Coding of Eye Position in Posterior Parietal Cortex despite Context-Dependent Tuning.

Neurons in posterior parietal cortex (PPC) encode many aspects of the sensory world (e.g., scene structure), the posture of the body, and plans for action. For a downstream computation, however, only some of these dimensions are relevant; the rest are "nuisance variables" because their influence on neural activity changes with sensory and behavioral context, potentially corrupting the read-out of relevant information. Here we show that a key postural variable for vision (eye position) is represented robustly in male macaque PPC across a range of contexts, although the tuning of single neurons depended strongly on context. Contexts were defined by different stages of a visually guided reaching task, including (1) a visually sparse epoch, (2) a visually rich epoch, (3) a "go" epoch in which the reach was cued, and (4) during the reach itself. Eye position was constant within trials but varied across trials in a 3 × 3 grid spanning 24° × 24°. Using demixed principal component analysis of neural spike-counts, we found that the subspace of the population response encoding eye position is orthogonal to that encoding task context. Accordingly, a context-naive (fixed-parameter) decoder was nevertheless able to estimate eye position reliably across contexts. Errors were small given the sample size (∼1.78°) and would likely be even smaller with larger populations. Moreover, they were comparable to that of decoders that were optimized for each context. Our results suggest that population codes in PPC shield encoded signals from crosstalk to support robust sensorimotor transformations across contexts.SIGNIFICANCE STATEMENT Neurons in posterior parietal cortex (PPC) which are sensitive to gaze direction are thought to play a key role in spatial perception and behavior (e.g., reaching, navigation), and provide a potential substrate for brain-controlled prosthetics. Many, however, change their tuning under different sensory and behavioral contexts, raising the prospect that they provide unreliable representations of egocentric space. Here, we analyze the structure of encoding dimensions for gaze direction and context in PPC during different stages of a visually guided reaching task. We use demixed dimensionality reduction and decoding techniques to show that the coding of gaze direction in PPC is mostly invariant to context. This suggests that PPC can provide reliable spatial information across sensory and behavioral contexts.

Human hand motion prediction based on feature grouping and deep learning: Pipe skid maintenance example

Human-robot collaboration has gained popularity in various construction applications. The key to a successful human-robot collaboration at the same construction workplace is the delicate algorithm for predicting human motions to strengthen the robot's situational awareness, i.e., robot-human awareness. Most existing approaches have focused on predicting human motions based on repetitive patterns of human behaviors in well-defined task contexts, such as specific object picking tasks, for a relatively short period of time. These methods can hardly capture the “pattern inconsistency” of human actions, i.e., the differences across people in terms of motion features and even for the same person at different time points of the task. This paper proposes an analytical pipeline that segments and clusters the human inconsistent behaviors into different pattern groups and builds separate human motion pattern prediction models correspondingly. The proposed method, Human Motion Feature Grouping and Prediction (HMFGP), quantifies the spatiotemporal relationship between gaze focus and hand movement trajectories, segments the raw data based on the detected gaze-hand relationship pattern changes, and clusters the matched gaze-hand data segments into several pattern groups based on the pattern similarity of the gaze-hand relationships. Then a time series Deep Learning method is used to predict hand motions based on gaze focus trajectories for each of the pattern groups. The gaze and hand motion data of a human subject experiment (n = 120) for pipe skid maintenance was used to test the prediction performance of HMFGP. The result shows that HMFGP can significantly improve the accuracy of human hand motion prediction and help quantity different patterns of human motions for specific analyses.

Research on Furniture Design Integrating Ming-Style Furniture Modeling Elements and Image Sensor Data: Taking Suitable Old Furniture as an Example

Ming-style furniture, adhering to the excellent traditional Chinese history and culture, is the pinnacle of traditional Chinese-style furniture, which vividly embodies the cultural essence of Chinese-style furniture. However, with the current popularity of smart home products, there are many obstacles for elderly users to use smart home products. Voice interaction is one of the main interaction methods of smart home products. The design strategy of voice user interface for aging smart home products is studied to improve the experience of elderly users in operating smart home products. Based on image sensing technology, combining Ming-style furniture modeling elements, and based on context theory research, this paper proposes four types of smart home contexts for elderly users, including user context, task context, time context, and environmental context, so as to propose a specific smart home product voice user interface design strategy. According to the user’s situational characteristics, a multichannel interaction design of voice user interface is proposed, an emotional and personalized voice user interface is constructed, and a design strategy of context memory assisting the dialogue process is constructed. According to the task, time, and environmental situation of elderly users, it proposes design strategies such as active interactive voice user interface design and provides continuous behavioral service experience. Conclusion. The research results provide a method and strategy reference for the design and development of the voice user interface of smart home products and provide a more natural and comfortable experience of using smart home products for elderly users.

Design patterns for human-AI co-learning: A wizard-of-Oz evaluation in an urban-search-and-rescue task

The rapid advancement of technology empowered by artificial intelligence is believed to intensify the collaboration between humans and AI as team partners. Successful collaboration requires partners to learn about each other and about the task. This human-AI co-learning can be achieved by presenting situations that enable partners to share knowledge and experiences. In this paper we describe the development and implementation of a task context and procedures for studying co-learning. More specifically, we designed specific sequences of interactions that aim to initiate and facilitate the co-learning process. The effects of these interventions on learning were evaluated in an experiment, using a simplified virtual urban-search-and-rescue task for a human-robot team. The human participants performed a victim rescue- and evacuation mission in collaboration with a wizard-of-Oz (i.e., a confederate of the experimenter who executed the robot-behavior consistent with an ontology-based AI-model). The designed interaction sequences, formulated as Learning Design Patterns (LDPs), were intended to bring about co-learning. Results show that LDPs support the humans understanding and awareness of their robot partner and of the teamwork. No effects were found on collaboration fluency, nor on team performance. Results are used to discuss the importance of co-learning, the challenges of designing human-AI team tasks for research into this phenomenon, and the conditions under which co-learning is likely to be successful. The study contributes to our understanding of how humans learn with and from AI-partners, and our propositions for designing intentional learning (LDPs) provide directions for applications in future human-AI teams.

Different Neural Information Flows Affected by Activity Patterns for Action and Verb Generation.

Shared brain regions have been found for processing action and language, including the left inferior frontal gyrus (IFG), the premotor cortex (PMC), and the inferior parietal lobule (IPL). However, in the context of action and language generation that shares the same action semantics, it is unclear whether the activity patterns within the overlapping brain regions would be the same. The changes in effective connectivity affected by these activity patterns are also unclear. In this fMRI study, participants were asked to perform hand action and verb generation tasks toward object pictures. We identified shared and specific brain regions for the two tasks in the left PMC, IFG, and IPL. The mean activation level and multi-voxel pattern analysis revealed that the activity patterns in the shared sub-regions were distinct for the two tasks. The dynamic causal modeling results demonstrated that the information flows for the two tasks were different across the shared sub-regions. These results provided the first neuroimaging evidence that the action and verb generation were task context driven in the shared regions, and the distinct patterns of neural information flow across the PMC-IFG-IPL neural network were affected by the polymodal processing in the shared regions.

Institutionalized affect in organizations: Not an oxymoron

Can affective states – emotions, moods, and sentiments – become institutionalized in an organization such that they become “objective” factors that are exterior to any one person and resistant to change? We argue that the answer is yes, through intertwined top–down and bottom–up processes that shape an organization’s (or subunit’s) affective climate and affective culture, resulting in a dynamic equilibrium. The top–down processes include leadership, attraction–selection–attrition, and socialization, coupled with the physical, task, and social context, while the bottom–up process of emergence occurs via affective events, appraisal, affective sharing, and affect schemas. We also consider how identification with the organization (or subunit) enhances the likelihood of institutionalized affect. We conclude that institutionalized affect in organizations is far from an oxymoron.

The influence of Facebook intrusion and task context on cognitive control.

Social networking sites, especially Facebook, have become increasingly popular over the past decades. However, besides the benefits of using Facebook, negative effects in the form of Facebook intrusion are also increasingly pointed out. Much of the research focuses on personality, emotional and social factors related to Facebook intrusion. However, there has been limited research on the relationship between this type of behavioural addiction and cognitive functioning. Consequently, the current study aimed to verify the relationship between Facebook intrusion and cognitive control in light of the dual mechanism of cognitive control model. Additionally, the study aim was to verify the impact of the Facebook-related context on cognitive control (proactive and reactive modes) compared to neutral and positive contexts. The participants (N = 82 young adults, 57 female, M = 22.24years, SD = 2.67years, age range 18-35years) were divided into two groups based on their level of Facebook intrusion. The Facebook intrusion scale was used to assess the level of Facebook intrusion. The AX-CPT task was used to assess proactive and reactive control modes in three task contexts: Facebook-related, neutral and positive. The current study results show that the participants with high Facebook intrusion had greater reactive control than participants with low Facebook intrusion. The differences between Facebook-related, neutral context and positive context were not found. However, the present findings demonstrate the interaction between Facebook intrusion and task context in cognitive control. More specifically, participants with low Facebook intrusion had greater proactive control than participants with high Facebook intrusion in the Facebook-related and positive context.

Challenging presumed technological superiority when working with (artificial) colleagues

Technological advancements are ubiquitously supporting or even replacing humans in all areas of life, bringing the potential for human-technology symbiosis but also novel challenges. To address these challenges, we conducted three experiments in different task contexts ranging from loan assignment over X-Ray evaluation to process industry. Specifically, we investigated the impact of support agent (artificial intelligence, decision support system, or human) and failure experience (one vs. none) on trust-related aspects of human-agent interaction. This included not only the subjective evaluation of the respective agent in terms of trust, reliability, and responsibility, when working together, but also a change in perspective to the willingness to be assessed oneself by the agent. In contrast to a presumed technological superiority, we show a general advantage with regard to trust and responsibility of human support over both technical support systems (i.e., artificial intelligence and decision support system), regardless of task context from the collaborative perspective. This effect reversed to a preference for technical systems when switching the perspective to being assessed. These findings illustrate an imperfect automation schema from the perspective of the advice-taker and demonstrate the importance of perspective when working with or being assessed by machine intelligence.

Exploration of Domain Relevance by Legal Professionals in Information Retrieval Systems

AbstractThis paper, written by Gineke Wiggers, Suzan Verberne, Gerrit-Jan Zwenne and Wouter Van Loon, addresses the concept of ‘relevance’ in relation to legal information retrieval (IR). They investigate whether the conceptual framework of relevance in legal IR, as described by Van Opijnen and Santos in their paper published in 2017, can be confirmed in practice.1 The research is conducted with a user questionnaire in which users of a legal IR system had to choose which of two results they would like to see ranked higher for a query and were asked to provide a reason for their choice. To avoid questions with an obvious answer and extract as much information as possible about the reasoning process, the search results were chosen to differ on relevance factors from the literature, where one result scores high on one factor, and the other on another factor. The questionnaire had eleven pairs of search results. A total of 43 legal professionals participated consisting of 14 legal information specialists, 6 legal scholars and 23 legal practitioners. The results confirmed the existence of domain relevance as described in the theoretical framework by Van Opijnen and Santos as published in 2017.2 Based on the factors mentioned by the respondents, the authors of this paper concluded that document type, recency, level of depth, legal hierarchy, authority, usability and whether a document is annotated are factors of domain relevance that are largely independent of the task context. The authors also investigated whether different sub-groups of users of legal IR systems (legal information specialists who are searching for others, legal scholars and also for legal practitioners) differ in terms of the factors they consider in judging the relevance of legal documents outside of a task context. Using a PERMANOVA there was found to be no significant difference in the factors reported by these groups. At this moment there is no reason to treat these sub-groups differently in legal IR systems.

The Media Literacy of Teenagers

In this paper, we present the conceptual framework, objectives, and methodology of an interdisciplinary research program (2018–2022) on the media literacy competence of teenagers in French-speaking Belgium, Quebec, France, and Switzerland. The program is undertaken by researchers in four universities, one in each national context. Focusing on information search and multimodal production as core media literacy activities, this program develops a multi-level assessment method to measure media literacy competence at varying levels of complexity. Additionally, we relate task-based measures of competence with self-reported competence and practices and seek to document the influence of the students’ interpretation of the tasks contexts and purposes, their motivation and feeling of empowerment, and their collaborative practices on the exercise of their media literacy. Preliminary results reveal that students report substantially higher levels of competence in information search than in web production; however, this difference is reduced when students assess their ability to perform specific search-related or production-related actions. These results are discussed in light of the potential usefulness of the program for teachers and for curricular design.

Educational results in the training of the future teacher and means of assessing their achievement

Introduction. The article deals with the concepts of “educational result” and “assessment tools”, which are components of the system for independent evaluation of the educational results of future teachers. Approaches to the description of the concepts of learning outcomes, competencies and educational result are analyzed from the standpoint of assessing the quality of training students and graduates of the specialization “Pedagogical education”. The structure and content of the concept of educational result are determined, the description of the invariant form of presentation of evaluation tools for assessing educational results is given. Examples are given.Materials and Methods. The study is based on systematic, activity-based and competence-based approaches to the training of bachelors of pedagogical education, which made it possible to describe the way of presenting educational results in the form of identifiable and measurable actions of the student. The specificity of the presentation of educational results based on the reformulation of competencies in combination with the labor actions of the teacher is revealed. It is shown that the description of indicators of achievement of the educational result in the form of descriptors serves as the basis for constructing an appropriate assessment tool. The principles of objectivity, validity and openness of evaluation procedures, which form the basis for constructing an invariant form of presentation of evaluation tools, are singled out.Results. The main provisions that determine the structure and content of the educational results of future teachers are highlighted. The definition of the concept of "educational result" is formulated; groups of educational results of future teachers in the field of education, in psychological, pedagogical and methodological activities, in the field of subject training are identified. The features of the description of the educational result based on the coordination of competencies presented in the Federal State Educational Standard of Higher Education and labor actions from the Professional Standard of the Teacher, and a way of constructing indicators of achievement of educational results are disclosed. A description of the invariant form of presentation of evaluation tools and examples of the construction of the educational result and the corresponding assessment tools are given.Discussion and Conclusions. The article discusses a method for constructing an assessment tool in the form of a contextual task that has been tested at the stage of the final state exam for bachelor's graduates in the field of study “Pedagogical education”, conducted in a new format.The conclusion formulated in the article is presented in the form of recommendations for the coordinated construction of educational results and assessment tools in the system of independent evaluation of the educational results of future teachers.

Regulatory focus theory: Disentangling goals and strategies

AbstractThe purpose of this research was to develop and test the Regulatory Goals and Strategies Questionnaire (RGSQ) to promote conceptual and empirical clarity in the literature on regulatory focus theory (RFT). According to RFT, construing an end‐state as a maximal goal (ideal, gain, and advancement) elicits the adoption of an eager promotion‐focused strategy, whereas construing an end‐state as a minimal goal (ought, non‐loss, and security) elicits the adoption of a vigilant prevention‐focused strategy. This means that promotion‐focused eagerness and prevention‐focused vigilance can be elicited by various goals and that eager and vigilant strategies are in fact separate from goals. However, extant measurement instruments have not treated them as such. The RGSQ was developed to distinguish various maximal and minimal goals and to separate these goals from promotion‐ and prevention‐focused strategies of eagerness and vigilance. Across three studies that combine six samples (total N = 1848), we examined the factor structure and nomological network of the RGSQ (Study 1: four samples of students and adults) and its predictive validity in a task context (Study 2) and working sample (Study 3). Overall, we found support for the RGSQ's validity and demonstrated that the RGSQ operationalizes regulatory focus in accordance with the theory's core assumptions.

Invited Session V: GABAergic function and dysfunction in visual perception: Strong evidence against a common center-surround mechanism in visual processing

Events and objects in the world generally do not occur in isolation. This surrounding spatial and temporal information has major effects on visual processing, effects that are typically described as contextual modulations. Given qualitative similarities among various contextual surround effects (e.g., the effects tend to exaggerate differences and suppress uniform stimuli), it might seem intuitive that these processes are related. It is unknown, however, whether contextual modulation processes across visual sub-modalities are independent, or at least in part share a common underlying mechanism. This is especially relevant for clinical work where these contextual effects are often linked to inhibitory function. We addressed this question using an individual differences approach in three different populations (total N = 129): neurotypical adults, older adults and patients with schizophrenia. The context battery included six tasks that assessed surround modulations in luminance, contrast, orientation, motion (2X) and size domains. The results revealed robust contextual effects across all tasks. However, despite large sample sizes, we found no significant correlations among different contextual tasks. Across groups, we did find significant differences: (a) weaker surround contrast effect in both schizophrenia and older adults, (b) stronger repulsive surround tilt and motion effects in older adults and (c) weaker surround suppression for motion in both schizophrenia and older adults. We find that (1) strength of contextual modulation does not correlate across visual sub-modalities and (2) deficits in contextual modulation that go in different directions. Together, this indicates an absence of a common underlying mechanism for contextual modulation in vision.

Role of anterior midcingulate cortex in self-reward representation and reward allocation judgments within social context.

Evaluating rewards for the self and others is essential for social interactions. Previous research has probed the neural substrates signaling rewards in social decision‐making tasks as well as the differentiation between self‐ and other‐reward representations. However, studies with different designs have yielded mixed results. After analyzing and comparing previous designs, we differentiated three components in this study: task (reward representation vs. social judgment of reward allocation), agency (self vs. other), and social context (without vs. within). Participants were asked to imagine various share sizes as a proposer in a dictator game during fMRI, and then rated their willingness and preference for these offers in a post‐scan behavioral task. To differentiate the regions involved in processing rewards without and within context, we presented the reward to each agent in two sequential frames. Parametric analyses showed that, in the second frame (i.e., within social context), the anterior midcingulate cortex (aMCC) signaled self‐reward and preferences for the offer, whereas the right insula tracked the likelihood of proposing the offer. Belief in a just world is positively associated with aMCC responses to self‐reward. These results shed light on the role of the aMCC in coding self‐reward within the social context to guide social behaviors.

Global connectivity fingerprints predict the domain generality of multiple-demand regions.

A set of distributed cognitive control networks are known to contribute to diverse cognitive demands, yet it is unclear how these networks gain this domain-general capacity. We hypothesized that this capacity is largely due to the particular organization of the human brain's intrinsic network architecture. Specifically, we tested the possibility that each brain region's domain generality is reflected in its level of global (hub-like) intrinsic connectivity as well as its particular global connectivity pattern ("connectivity fingerprint"). Consistent with prior work, we found that cognitive control networks exhibited domain generality as they represented diverse task context information covering sensory, motor response, and logic rule domains. Supporting our hypothesis, we found that the level of global intrinsic connectivity (estimated with resting-state functional magnetic resonance imaging [fMRI]) was correlated with domain generality during tasks. Further, using a novel information fingerprint mapping approach, we found that each cognitive control region's unique rule response profile("information fingerprint") could be predicted based on its unique intrinsic connectivity fingerprint and the information content in regions outside cognitive control networks. Together, these results suggest that the human brain's intrinsic network architecture supports its ability to represent diverse cognitive task information largely via the location of multiple-demand regions within the brain's global network organization.

Harnessing the power of promising technologies to transform science education: prospects and challenges to promote adaptive epistemic beliefs in science learning

ABSTRACT Forming learners’ science concepts and conceptual change entails adaptive epistemic beliefs to support a high degree of interactivity within a coherent knowledge structure. Adaptive epistemic beliefs are characterized by beliefs that knowledge is uncertain and should be justified through experimentation or multiple sources dependent upon the task contexts. Thus, assessing and evaluating learners’ adaptive epistemic beliefs is a complex process that requires laborious analysis of learner artifacts based on reliable and valid coding schemes. This article aims to describe new ways of assessing and applying technologies that can measure and foster adaptive epistemic beliefs. We propose new strategies for a theoretically-based human-and-machine symbiotic Learning Analytics (LA) framework. The application of this LA framework may facilitate the development of real-time detecting and representation of the individual and collective epistemic belief networks as well as diagnosing and providing appropriate scaffolds to promote adaptive epistemic beliefs via the design of personalised pedagogical feedback with experts’ input. The heuristic application of technology infrastructure may propel a movement for more tangible and personalised learning in science education. The current gaps of using AI-based emerging technologies in science learning and implications for science education are discussed to advance science education in new directions.