Universitätsbibliothek

Cognitive control is what makes goal-directed actions possible. Whenever the environment or our impulses strongly suggests a response that is incompatible with our goals, conflict arises. Such conflicts are believed to cause negative affect. Aversive consequences of conflict may be registered in a conflict monitoring module, which subsequently initiates attentional changes and action tendencies to reduce negative affect. This association suggests that behavioral adaptation might be a reflection of emotion regulation. The theoretical cornerstone of current research on emotion regulation is the process model of emotion regulation, which postulates the regulation strategies situation selection, situation modification, attentional deployment, cognitive change, and response modulation. Under the assumption that conflict adaptation and affect regulation share common mechanisms, I derived several predictions regarding cognitive control from the process model of emotion regulation and tested them in 11 experiments (N = 509). Participants engaged in situation selection towards conflict, but only when they were explicitly pointed to action-outcome contingencies (Experiments 1 to 3). I found support for a mechanism resembling situation modification, but no evidence for a role of affect (Experiments 4 to 10). Changing the evaluation of conflict had no impact on the extent of conflict adaptation (Experiment 11). Overall, there was evidence for an explicit aversiveness of cognitive conflict, but less evidence for implicit aversiveness, suggesting that conflict may trigger affect regulation processes, particularly when people explicitly have affect regulation goals in mind.

Learning with digital media has become a substantial part of formal and informal educational processes and is gaining more and more importance. Technological progress has brought overwhelming opportunities for learners, but challenges them at the same time. Learners have to regulate their learning process to a much greater extent than in traditional learning situations in which teachers support them through external regulation. This means that learners must plan their learning process themselves, apply appropriate learning strategies, monitor, control and evaluate it. These requirements are taken into account in various models of self-regulated learning (SRL). Although the roots of research on SRL go back to the 1980s, the measurement and adequate support of SRL in technology-enhanced learning environments is still not solved in a satisfactory way. An important obstacle are the data sources used to operationalize SRL processes. In order to support SRL in adaptive learning systems and to validate theoretical models, instruments are needed which meet the classical quality criteria and also fulfil additional requirements. Suitable data channels must be measurable "online", i.e., they must be available in real time during learning for analyses or the individual adaptation of interventions. Researchers no longer only have an interest in the final results of questionnaires or tasks, but also need to examine process data from interactions between learners and learning environments in order to advance the development of theories and interventions. In addition, data sources should not be obtrusive so that the learning process is not interrupted or disturbed. Measurements of physiological data, for example, require learners to wear measuring devices. Moreover, measurements should not be reactive. This means that other variables such as learning outcomes should not be influenced by the measurement. Different data sources that are already used to study and support SRL processes, such as protocols on thinking aloud, screen recording, eye tracking, log files, video observations or physiological sensors, meet these criteria to varying degrees. One data channel that has received little attention in research on educational psychology, but is non-obtrusive, non-reactive, objective and available online, is the detailed, timely high-resolution data on observable interactions of learners in online learning environments. This data channel is introduced in this thesis as "peripheral data". It records both the content of learning environments as context, and related actions of learners triggered by mouse and keyboard, as well as the reactions of learning environments, such as structural or content changes. Although the above criteria for the use of the data are met, it is unclear whether this data can be interpreted reliably and validly with regard to relevant variables and behavior. Therefore, the aim of this dissertation is to examine this data channel from the perspective of SRL and thus further close the existing research gap. One development project and four research projects were carried out and documented in this thesis.

In three studies, we investigated, if and how different modes of presentation - written, auditory, audiovisual (auditory combined with pictures) - affect comprehension of semantically identical materials. Children, beginning from the age of 7, and adults were included into the studies. A vast amount of studies have shown that pictures can facilitate text comprehension (e.g. Carney & Levin, 2002). Other than the majority of these previous studies, we assessed text comprehension with methods that we assume to allow more differentiated insights into the cognitive processes that - according to current theories - underlie text comprehension. Text comprehension involves at least three levels of mental representations (see Kintsch, 1998). Moreover, text comprehension means constructing a locally and globally coherent mental representation of the text content. Using a sentence recognition task (see Schmalhofer & Glavanov, 1986), we examined whether the memory of the text surface, the text base, and the situation model differs between written, auditory, and audiovisual text presentation in a sample of 103 8- and 10-year-olds and adults (Study I), and between auditory and audiovisual text presentation in a sample of 106 7-, 9-, and 11-year-olds (Study II). Furthermore, we examined with 155 9- and 11-year-olds, whether the ability to draw inferences to establish local and global coherence differs between written, auditory, and audiovisual text presentation. These inferences were indicated by reaction times to words associated with a protagonist's super- (global) or subordinate (local) goal. Overall, the results of these three studies taken together, indicate that children up to age 11 do not only have better memory of not only the text surface, but also of the situation model when pictures are added to an auditory text. This effect became apparent in comparison with both auditory and written texts. For the adults, in contrast, we did not find an effect of the presentation mode. Furthermore, both 9- and 11-year-olds were better at establishing global coherence at audiovisual compared to auditory text presentation. Written presentation turned out to be superior to auditory presentation in terms of both local and global coherence.

Generating information, compared to reading, improves learning and enhances long-term retention of the learned content. This so-called generation effect has been demonstrated repeatedly for recall and recognition of single words. However, before adopting generating as a learning strategy in educational contexts, conditions moderating the effect need to be identified. This study investigated the impact of positive and negative mood states on the generation effect with short expository texts. According to the dual-force framework (Fiedler, Nickel, Asbeck, & Pagel, 2003), positive mood should facilitate generation by enhancing creative knowledge-based top-down processing (assimilation). Negative mood, however, should facilitate learning in the read-condition by enhancing critical stimulus-driven bottom-up processing (accommodation). In contrast to our expectations, we found no general generation effect but an overall learning advantage of read compared to generated texts. However, a significant interaction of learning condition and mood indicates that learners in a better mood recall generated texts better than learners in a more negative mood, whereas no mood effect was found when the texts were read. The results of the present study partially support the predictions of the dual-force framework and are discussed in the context of recent theoretical approaches to the generation effect.

The current dissertation addresses the analysis of technology-enhanced learning processes by using Process Mining techniques. For this purpose, students’ coded think-aloud data served as the measurement of the learning process, in order to assess the potential of this analysis method for evaluating the impact of instructional support. The increasing use of digital media in higher education and further educational sectors enables new potentials. However, it also poses new challenges to students, especially regarding the self-regulation of their learning process. To help students with optimally making progress towards their learning goals, instructional support is provided during learning. Besides the use of questionnaires and tests for the assessment of learning, researchers make use increasingly of process data to evaluate the effects of provided support. The analysis of observed behavioral traces while learning (e.g., log files, eye movements, verbal reports) allows detailed insights into the student’s activities as well as the impact of interventions on the learning process. However, new analytical challenges emerge, especially when going beyond the analysis of pure frequencies of observed events. For example, the question how to deal with temporal dynamics and sequences of learning activities arises. Against this background, the current dissertation concentrates on the application of Process Mining techniques for the detailed analysis of learning processes. In particular, the focus is on the additional value of this approach in comparison to a frequency-based analysis, and therefore on the potential of Process Mining for the evaluation of instructional support. An extensive laboratory study with 70 university students, which was conducted to investigate the impact of a support measure, served as the basis for pursuing the research agenda of this dissertation. Metacognitive prompts supported students in the experimental group (n = 35) during a 40-minute hypermedia learning session; whereas the control group (n = 35) received no support. Approximately three weeks later, all students participated in another learning session; however, this time all students learned without any help. The participants were instructed to verbalize their learning activities concurrently while learning. In the three analyses of this dissertation, the coded think aloud data were examined in detail by using frequency-based methods as well as Process Mining techniques. The first analysis addressed the comparison of the learning activities between the experimental and control groups during the first learning session. This study concentrated on the research questions whether metacognitive prompting increases the number of metacognitive learning activities, whether a higher number of these learning activities corresponds with learning outcome (mediation), and which differences regarding the sequential structure of learning activities can be revealed. The second analysis investigated the impact of the individual prompts as well as the conditions of their effectiveness on the micro level. In addition to Process Mining, we used a data mining approach to compare the findings of both analysis methods. More specifically, we classified the prompts by their effectiveness, and we examined the learning activities preceding and following the presentation of instructional support. Finally, the third analysis considered the long-term effects of metacognitive prompting on the learning process during another learning session without support. It was the key objective of this study to examine which fostered learning activities and process patterns remained stable during the second learning session. Overall, all three analyses indicated the additional value of Process Mining in comparison to a frequency-based analysis. Especially when conceptualizing the learning process as a dynamic sequence of multiple activities, Process Mining allows identifying regulatory loops and crucial routing points of the process. These findings might contribute to optimizing intervention strategies. However, before drawing conclusions for the design of instructional support based on the revealed process patterns, additional analyses need to investigate the generalizability of results. Moreover, the application of Process Mining remains challenging because guidelines for analytical decisions and parameter settings in technology-enhanced learning context are currently missing. Therefore, future studies need to examine further the potential of Process Mining as well as related analysis methods to provide researchers with concrete recommendations for use. Nevertheless, the application of Process Mining techniques can already contribute to advance the understanding of the impact of instructional support through the use of fine-grained process data.