Graph Co-Attentive Session-based Recommendation

Session-based recommendation aims to generate recommendations merely based on the ongoing session, which is a challenging task. Previous methods mainly focus on modeling the sequential signals or the transition relations between items in the current session using RNNs or GNNs to identify user’s intent for recommendation. Such models generally ignore the dynamic connections between the local and global item transition patterns, although the global information is taken into consideration by exploiting the global-level pair-wise item transitions. Moreover, existing methods that mainly adopt the cross-entropy loss with softmax generally face a serious over-fitting problem, harming the recommendation accuracy. Thus, in this article, we propose a Graph Co-Attentive Recommendation Machine (GCARM) for session-based recommendation. In detail, we first design a Graph Co-Attention Network (GCAT) to consider the dynamic correlations between the local and global neighbors of each node during the information propagation. Then, the item-level dynamic connections between the output of the local and global graphs are modeled to generate the final item representations. After that, we produce the prediction scores and design a Max Cross-Entropy (MCE) loss to prevent over-fitting. Extensive experiments are conducted on three benchmark datasets, i.e., Diginetica, Gowalla, and Yoochoose. The experimental results show that GCARM can achieve the state-of-the-art performance in terms of Recall and MRR, especially on boosting the ranking of the target item.

Serial Recall Order of Category Fluency Words: Exploring Its Neural Underpinnings

Background: Although performance on the category fluency test (CFT) is influenced by many cognitive functions (i.e., including language, executive functioning and speed of processing), item-level scoring methods of CFT performance might be a promising way to capture aspects of semantic memory that are less influenced by intervenient abilities. One such approach is based on the calculation of correlation coefficients that quantify the association between item-level features and the serial order with which words are recalled (SRO).Methods: We explored the neural underpinnings of 10 of these correlational indices in a sample of 40 healthy adults who completed a classic 1-min CFT and an MRI protocol inclusive of T1-weighted (analysed with voxel-based morphometry) and resting-state fMRI sequences for the evaluation of the default-mode network (DMN). Two sets of linear models were defined to test the association between neural maps and each correlational index: a first set in which major demographic and clinical descriptors were controlled for and a second set in which, additionally, all other 9 correlational indices were regressed out.Results: In the analysis of the DMN, ‘SRO-frequency’, ‘SRO-dominance’ and ‘SRO-body-object interaction’ correlational indices were all negatively associated with the anterior portion of the right temporoparietal junction. The ‘SRO-frequency’ correlational index was also negatively associated with the right dorsal anterior cingulate and the ‘SRO-dominance’ correlational index with the right lateral prefrontal cortex. From the second set of models, the ‘SRO-typicality’ correlational index was positively associated with the left entorhinal cortex. No association was found in relation to grey matter maps.Conclusion: The ability to retrieve more difficult words during CFT performance as measured by the correlational indices between SRO and item-level descriptors is associated with DMN expression in regions deputed to attentional reorienting and processing of salience of infrequent stimuli and dominance status. Of all item-level features, typicality appears to be that most closely linked with entorhinal functioning and may thus play a relevant role in assessing its value in testing procedures for early detection of subtle cognitive difficulties in people with suspected Alzheimer’s degeneration. Although exploratory, these findings warrant further investigations in larger cohorts.

Scale length does matter: Recommendations for measurement invariance testing with categorical factor analysis and item response theory approaches

In social sciences, the study of group differences concerning latent constructs is ubiquitous. These constructs are generally measured by means of scales composed of ordinal items. In order to compare these constructs across groups, one crucial requirement is that they are measured equivalently or, in technical jargon, that measurement invariance (MI) holds across the groups. This study compared the performance of scale- and item-level approaches based on multiple group categorical confirmatory factor analysis (MG-CCFA) and multiple group item response theory (MG-IRT) in testing MI with ordinal data. In general, the results of the simulation studies showed that MG-CCFA-based approaches outperformed MG-IRT-based approaches when testing MI at the scale level, whereas, at the item level, the best performing approach depends on the tested parameter (i.e., loadings or thresholds). That is, when testing loadings equivalence, the likelihood ratio test provided the best trade-off between true-positive rate and false-positive rate, whereas, when testing thresholds equivalence, the χ2 test outperformed the other testing strategies. In addition, the performance of MG-CCFA’s fit measures, such as RMSEA and CFI, seemed to depend largely on the length of the scale, especially when MI was tested at the item level. General caution is recommended when using these measures, especially when MI is tested for each item individually.

Empowering Teachers with AI: Co-Designing a Learning Analytics Tool for Personalized Instruction in the Science Classroom

AI-based educational technology that is designed to support teachers in providing personalized instruction can enhance their ability to address the needs of individual students, hopefully leading to better learning gains. This paper presents results from participatory research aimed at co-designing with science teachers a learning analytics tool that will assist them in implementing a personalized pedagogy in blended learning contexts. The development process included three stages. In the first, we interviewed a group of teachers to identify where and how personalized instruction may be integrated into their teaching practices. This yielded a clustering-based personalization strategy. Next, we designed a mock-up of an AI-based tool that supports this strategy and worked with another group of teachers to define an `explainable learning analytics' scheme that explains each cluster in a way that is both pedagogically meaningful and can be generated automatically. Third, we developed an AI algorithm that supports this `explainable clusters' pedagogy and conducted a controlled experiment that evaluated its contribution to teachers' ability to plan personalized learning sequences. The planned sequences were evaluated in a blinded fashion by an expert, and the results demonstrated that the experimental group -- teachers who received the clusters with the explanations -- designed sequences that addressed the difficulties exhibited by different groups of students better than those designed by teachers who received the clusters without explanations. The main contribution of this study is twofold. First, it presents an effective personalization approach that fits blended learning in the science classroom, which combines a real-time clustering algorithm with an explainable-AI scheme that can automatically build pedagogically meaningful explanations from item-level meta-data (Q Matrix). Second, it demonstrates how such an end-to-end learning analytics solution can be built with teachers through a co-design process and highlights the types of knowledge that teachers add to system-provided analytics in order to apply them to their local context. As a practical contribution, this process informed the design of a new learning analytics tool that was integrated into a free online learning platform that is being used by more than 1000 science teachers.

The Tangled Webs We Wreak: Examining the Structure of Aggressive Personality Using Psychometric Networks

Trait aggression is a prominent construct in the psychological literature, yet little work has sought to situate trait aggression among broader frameworks of personality. Initial evidence suggests that trait aggression may be best couched within the nomological network of the Five Factor Model (FFM). The current work sought to locate the most appropriate home for trait aggression among the FFM. We applied a preregistered regimen of psychometric network analyses to three datasets (combined N = 2,927) that contained self-reports of trait aggression and the FFM traits. Trait aggression was highly central in the factor-level networks, which contained associations consistent with the conceptualization of this construct as a lower-order component of low agreeableness. The facet-level networks revealed that the behavioral facets of trait aggression reflected low agreeableness, but that the anger and hostility facets reflected high neuroticism. The item-level network suggested that the intent to initiate aggressive encounters was the primary bridge that empirically linked trait aggression to agreeableness. Our results indicate that trait aggression is primarily a lower-order facet of agreeableness, advance our understanding of trait aggression, integrate it with broader frameworks of personality, and suggest future directions to refine this complex dispositional tendency.

A Response-Time-Based Latent Response Mixture Model for Identifying and Modeling Careless and Insufficient Effort Responding in Survey Data

Careless and insufficient effort responding (C/IER) can pose a major threat to data quality and, as such, to validity of inferences drawn from questionnaire data. A rich body of methods aiming at its detection has been developed. Most of these methods can detect only specific types of C/IER patterns. However, typically different types of C/IER patterns occur within one data set and need to be accounted for. We present a model-based approach for detecting manifold manifestations of C/IER at once. This is achieved by leveraging response time (RT) information available from computer-administered questionnaires and integrating theoretical considerations on C/IER with recent psychometric modeling approaches. The approach a) takes the specifics of attentive response behavior on questionnaires into account by incorporating the distance–difficulty hypothesis, b) allows for attentiveness to vary on the screen-by-respondent level, c) allows for respondents with different trait and speed levels to differ in their attentiveness, and d) at once deals with various response patterns arising from C/IER. The approach makes use of item-level RTs. An adapted version for aggregated RTs is presented that supports screening for C/IER behavior on the respondent level. Parameter recovery is investigated in a simulation study. The approach is illustrated in an empirical example, comparing different RT measures and contrasting the proposed model-based procedure against indicator-based multiple-hurdle approaches.