Explainable AI has become a common term in the literature, scrutinized by computer scientists and statisticians and highlighted by psychological or philosophical researchers. One major effort many researchers tackle is constructing general guidelines for XAI schemes, which we derived from our study. While some areas of XAI are well studied, we focus on uncertainty explanations and consider global explanations, which are often left out. We chose an algorithm that covers various concepts simultaneously, such as uncertainty, robustness, and global XAI, and tested its ability to calibrate trust. We then checked whether an algorithm that aims to provide more of an intuitive visual understanding, despite being complicated to understand, can provide higher user satisfaction and human interpretability.

Over-trusting AI systems can lead to complacency and decision errors.However, human and system variables may affect complacency and it is important to understand their interplay for HCI.In our experiment, 90 participants were confronted with traffic route problems guided by AI recommendations and thereby assigned to either a transparent system providing reasons for recommendations or a non-transparent system.We found transparent systems to lower the potential to alleviate workload (albeit not to neglect monitoring), but to simultaneously foster actual complacent behavior.On the contrary, we found performance expectancy to foster the potential to alleviate workload, but not complacent behavior.Interaction analyses showed that effects of performance expectancy depend on system transparency.This contributes to our understanding how system-and person-related variables interact in affecting complacency and stresses the differences between dimensions of complacency and the need for carefully considering transparency and performance expectancy in AI research and design.

Accurately assessing learning partners’ knowledge profiles improves collaborative learning. Group awareness tools facilitate constructing such social context knowledge during learning and the formation of learning partner models retrievable afterwards. In this experimental study (N = 70), we investigated potential schema effects in partner modeling: Participants were first provided with descriptions of two learning partners (expert vs. novice of an area) and with their knowledge profiles consisting of knowledge levels (high vs. low) regarding certain topics of this area. In a memory test, participants had to remember the specific knowledge levels of both partners. Partner modeling was analyzed with multinomial processing tree models, as these models disentangle memory and guessing, which are often confounded when schema effects on memory are examined. High and low knowledge levels were equally well remembered for both partners. However, participants showed metacognitive biases, expecting their memory to be better for high knowledge levels. Additionally, knowledge profile estimations revealed that while novices’ knowledge was estimated accurately, experts’ knowledge was overestimated. We discuss the results and potential benefits of using multinomial processing tree models in the learning sciences for and beyond analyzing schema effects in partner modeling.

In order to effectively use augmented reality (AR) for learning support, it is important to take into account learners’ subjective AR-specific learning experience. In this work-in-progress paper, a first version of the ARcis questionnaire intended to measure learners’ perception of contextuality, interactivity, and spatiality in AR-based learning environments is presented. Results of a first implementation of this questionnaire in four studies (total N = 456) with different AR-based learning materials are described. Analyses of reliability and factor structure of the questionnaire show promising first outcomes on which suggestions for adaptations and future validation strategies are made.

Learning about three-dimensional (3D) objects is the focus of many augmented reality (AR) applications. Although this underlines the importance of spatiality in AR experiences, learners' spatial abilities should be considered in this context. While spatial abilities may compensate for the lack of spatial information in 2D representations (ability-as-compensator hypothesis), they may also be necessary for learning with 3D representations in the first place (ability-as-enhancer hypothesis). In the current study, we examine the role 3D spatial visualization abilities and 2D spatial memory abilities may play when learning with 3D AR objects. Both variables were measured in an exploratory pilot study in which N = 33 participants learned about the spatial structure of the modules of the International Space Station (ISS) with either an AR or a non-AR mobile application. We found that spatial abilities indeed had moderating effects on achievement in the learning task and knowledge test, although the results are inconclusive concerning an ability-as-enhancer or -compensator hypothesis. We discuss the results and conclude that it is necessary that researchers take a closer look at the role of learners' spatial abilities when learning with 3D representations in AR. With additional insights, practitioners can then make informed decisions on using AR applications.

In this paper, a study with the focus on interactivity in augmented reality (AR) applications concerning the influence of different forms of interaction with AR learning material is presented. While research on multimedia learning often distinguishes between mental and physical interaction with learning material, other research fields state that physical interaction is necessary to interact mentally. To look at how this distinction may play a role in AR-based learning material, an experimental study with a 2x2 design manipulating mental and physical interaction was conducted, including learning material on the topic of power plants. The data (N = 128) were collected and analyzed, showing that, although not expected, learning was better in groups in which either more physical or more mental interaction was applied, but not in groups in which both were high. The results are discussed under the potential idea of cognitive overload.

Digitization in vocational education and training (VET) offers new opportunities for integrating work processes and background knowledge through interactive media. This involves the design, testing and evaluation of innovative means of didactic and technical support. In the project “Digitized support for competence development in the workplace”, this task is approached from a combined psychological and computer science perspective. This paper reports on one usage scenario from the first year of the project examining how to embed learning videos in VET. The results include new findings on video-based learning, and inform the design and implementation of a prototypical online learning platform.

A common ground of knowledge is the foundation for collaborative learning. However, learners often do not know how to engage in beneficial grounding activities during collaboration, which highlights the need to support such activities and the development of grounding skills. Two prominent forms of support are collaboration scripts and cognitive group awareness tools that pursue different approaches to promoting grounding. While collaboration scripts explicitly ask learners to perform related activities to trigger the exchange of relevant information, cognitive group awareness tools directly provide relevant information about learning partners to implicitly guide learners during learning processes. The present paper examines how the two approaches differ in supporting grounding activities and to what extent they support the development of grounding skills that represent a key qualification of collaborative learning. In the next step of this research project on grounding, the question of how to support the long-term development of grounding skills will be investigated.

In this paper, three technological characteristics of augmented reality (AR) are reframed from a perceptual, user’s perspective and discussed concerning their potential for education and in the context of research on technology-supported learning. The first characteristic, contextuality, describes that users of AR can experience the real world and virtual elements simultaneously. The second characteristic, interactivity, includes the possibilities to interact with AR through the manipulation of both real objects and virtual properties, which offers novel possibilities for interaction. The third characteristic, spatiality, focusses on the linking of virtual objects to specific points in space and the more realistic three-dimensionality that AR visualizations offer. It is proposed that these three characteristics can provide a way to structure the broad research landscape of AR in education and form a basis for future research projects. Two studies are presented and linked to the three characteristics. In the first study, the comparison of a desktop simulation and an AR simulation in an individual learning setting is linked to the characteristics of interactivity and spatiality. In the second study, the contextuality of AR is systematically varied and exploited to present group awareness information about other learners next to these learners instead of separated from them. The results of the studies are discussed in the context of the three characteristics and the paper concludes that there are a lot of different educational settings in which AR could be beneficial. The classification of and systematic variation in research based on the three characteristics may form a basis to systematize educational AR research. Furthermore, the results of this research and the three characteristics themselves can inform the design of AR applications to support learning.

In computer-supported collaborative learning, group awareness tools have been shown to be helpful regarding learning processes and outcomes. Previous research has focused on the support via cognitive and behavioural group awareness information, largely neglecting emotional group awareness information and often investigating these three aspects separately. To support large social media groups such as wiki communities, integrating different types of group awareness (GA) information may yield benefits, since these communities encounter several challenges. Although jointly presenting different GA information is assumed to be advantageous for collaborative learning, GA interaction effects including personality traits are still largely unexplored. In order to close this research gap, an integrated framework is proposed, which enables the systematic empirical investigation of (interaction) effects of different types of GA information on behavioural, cognitive, and emotional challenges in computer-supported collaborative learning, with a focus on wikis.

Students are expected to deal with complex learning material, that regularly challenges their learning processes especially in the first semester. When students ask peers for explanations, they apply a functional learning strategy, known as academic help-seeking. But finding peers to discuss questions on learning materials is difficult for those that are new to university. Finding adequate peers for help-seeking requires awareness about their knowledge and competencies in and beyond the learners’ own study group; i.e. group awareness. This paper investigates how group awareness information supports academic help-seeking among first-semester students on a Moodle-based learning platform that has been adapted with a newly developed extension to collect, transform and visualize group awareness information about peers. In a field study accompanying an introductory statistics lecture (N = 107), we examined (1) motivational, attitudinal and social characteristics of helpees that asked for help on the platform, (2) the effect of provided group awareness information on the selection of helpers, and (3) the cognitive characteristics of selected helpers and their difference from the helpees. The participating students were invited to seek help from their peers within Moodle and had access to a sortable list of all course members in order to find an appropriate helper. The control group (n = 55) only saw the full names of the peers while the treatment group (n = 52) was additionally provided with self-assessed group awareness information (competence and availability of a helper). Unfortunately, participants rarely used the platform and only 22 help-seeking episodes were recorded. Results indicate that helpees did not differ regarding motivational, intentional or social factors from the rest of the course. The overall occurrences of sorting the list was higher for the condition provided with group awareness information, but both groups did not differ with respect to the competence of selected helpers. The results provide useful, practical insights bringing online help-seeking into an educational context.

Collaborative practices cover a vast variety of contexts and educational goals. Despite these differences, most means of support draw on between- and within-learner discrepancies as driving forces of individual and collaborative learning. These discrepancies are a focus of cognitive Group Awareness Tools, that process knowledge-related learner data and feed it back to the group to ease the interpretation of the learning situation thereby guiding collaborative learning activities. In this paper, we examine the features and different characteristics of these tools. Based on three exemplary cognitive Group Awareness Tools focusing on different types of knowledge discrepancies, we explore how data processing is adjusted to different settings and discuss whether cognitive Group Awareness Tools are suitable devices to be deployed throughout various educational contexts.

Cognitive group awareness tools support learners in dealing with specific challenges of computer-supported collaborative learning scenarios. This paper suggests three supporting functions underlying cognitive group awareness tools and proposes to disentangle them in three consecutive experimental studies with regard to multimedia learning. It focuses on one of those studies that systematically investigates whether providing knowledge-related partner information can improve collaborative learning with multiple external representations and with dynamic and interactive visualizations. Learning dyads (N = 120) were compared in four groups that differed with regard to the presentation of knowledge-related partner information during two subsequent collaboration phases: (1) learning with multiple static representations and (2) learning with dynamic and interactive visualizations. Results indicate that providing partner information during collaboration can facilitate meaningful learning discourses and improve learning outcomes.

User generated content in Wikis is mainly distributed on the article view and its corresponding talk page. Potentials of analysing and supporting discussants' knowledge construction processes on the level of talk pages have still been rarely researched. The presented experimental study addresses this issue by providing external representations of content-related controversies which were led by contradictory evidence between discussants to foster awareness on socio-cognitive conflicts which can be beneficial for learning. Its aim is to investigate how increased salience of controversies can guide participants' (N = 81) navigation and learning processes. Three conditions differing in their degree of awareness support were implemented in this study. Results indicate that the implementation of awareness representations helped students to focus on meaningful discussion threads. Findings suggest that Wiki talk page users can benefit from additional structuring aids.