Markus Neubeck

• This dissertation project aims to examine the potential of network modelling, an increasingly popular methodology in emotion research (e.g., Fried et al., 2016), to better comprehend age-related differences in structural connections between cognitive processes such as fluid intelligence and executive control functions. Furthermore, it aims to identify the key variables that link self-regulation to executive control functions and age-related discrepancies. Lastly, it seeks to delve into the key variables and correlations between executive control functions, self-regulation, and affect utilizing a longitudinal design in combination with machine learning as a data-driven method. In study 1, differences between the cognitive performance networks of younger (M = 38.0 years of age, SD = 9.9) and older (M = 64.1 years of age, SD = 7.7) adults were explored. Network modelling showed that while speeded attention is essential throughout the life-span, connections between fluid intelligence and working memory were stronger, and intelligence was more central in the older group. Additionally, confirmatory factor modelling demonstrated that latent correlations were highest between working memory and intelligence, particularly in older adults, whereas inhibition had the lowest correlations with other abilities. This research suggests that the relations of cognitive abilities may differ between younger and older adults, indicating process-specific changes in the cognitive performance network. In study 2, we investigated the connections of self-regulation (SR) and executive control functions (EF), which are theoretical concepts encompassing various cognitive abilities supporting the regulation of behavior, thoughts, and emotions (Inzlicht et al., 2021; Wiebe & Karbach, 2017). Evidence, however, implies that correlations between self-report measures and performance-based tasks are often difficult to observe (e.g., Eisenberg et al., 2019). We investigated connections and overlap between different aspects of SR and EF in a life-span sample (14-82 years). Participants completed several self-report measures and behavioral tasks, such as sensation seeking, mindfulness, grit, or eating behavior questionnaires and working memory, inhibition, and shifting tasks. Network models for a youth, middle-aged, and older-aged group were estimated to identify key variables that are well connected in the SR and EF construct space. In general, stronger connections were observed within the clusters of SR and EF than between them, and older adults appeared to have more connections between SR and EF than younger individuals, probably because of declining cognitive resources. In study 3, we analyzed the intricate links between EF, SR and affect, as well as individual differences in these relations. Bridgett et al. (2013) proposed that EF and self-regulation SR are psychological constructs to support the regulation of cognition and affect. A total of 315 participants, aged 14 to 80, answered questionnaires and took part in behavioral tasks which evaluated EF, SR, and both positive and negative affect two times (one-month apart). Combined X-means and deep learning algorithms aided in the separation of two distinct groups who featured different EF performances, SR tendencies, and affective experiences. Network model analysis was then utilized to confirm the connections between the EF, SR, and affect variables in each of the two groups. The two groups displayed a maximal centrality for variables linked to SR and positive affect. Group membership remained mostly consistent (85%) across both measurement occasions. Logistic regression indicated that age and personality (conscientiousness, neuroticism, and agreeableness) predicted group membership. This sheds light on stable individual differences in the complex relations of EF, SR, and affect. This dissertation project utilized a combination of standard approaches (such as confirmatory factor analysis; CFA) and advanced approaches (such as network models, machine learning algorithms, and deep learning) to explore the connections between cognitive abilities, EF, SR, and affect. Our findings are in line with the theory of process specific changes in age-dedifferentiation. Findings suggested that connections between SR and EF were stronger within clusters, and positive affect was better connected to SR than EF measures. Lastly, age and personality traits were found to predict the clusters. These findings suggest that computational modelling is an effective exploratory tool in understanding how cognitive abilities and other psychological constructs may interact. Further research is necessary to gain further insights on the mechanisms behind differences in network structures.
• Ziel dieses Dissertationsprojektes ist es, mit Netzwerkmodellierung altersbedingte Unterschiede in den strukturellen Verbindungen zwischen fluider Intelligenz und exekutiven Kontrollfunktionen sowie Selbstregulation und Affekt besser zu verstehen. Dabei werden Schlüsselvariablen identifiziert und altersbedingte Diskrepanzen in den Netzwerkstrukturen aufgedeckt. Weiterhin werden anhand von maschinellem Lernen und Deep Learning im Längsschnittdesign interindividuelle Unterschiede in diesen Zusammenhängen untersucht. In Studie 1 wurden die Unterschiede zwischen den Netzwerken der kognitiven Leistung von jüngeren und älteren Erwachsenen untersucht. Dabei zeigte sich, dass die Zusammenhänge zwischen den Schlüsselvariablen fluide Intelligenz und Arbeitsgedächtnis bei älteren Erwachsenen stärker waren, was auf prozessspezifische Veränderungen im Netzwerk kognitiver Leistungen hindeutet. Diese Ergebnisse decken sich mit denen einer konfirmatorischen Faktorenanalyse. In Studie 2 wurden Zusammenhänge zwischen Selbstregulation, exekutiven Kontrollfunktionen und verschiedenen kognitiven Fähigkeiten untersucht. Ältere Erwachsene zeigten stärkere Verbindungen zwischen Selbstregulation und exekutiven Kontrollfunktionen als jüngere Personen. In Studie 3 wurden die komplexen Beziehungen zwischen exekutiven Kontrollfunktionen, Selbstregulation und Affekt in verschiedenen Altersgruppen im Längsschnitt mit Netzwerkmodellen und Machine Learning untersucht. Dabei wurde deutlich, dass ältere Erwachsene eine stärkere Verbindung zwischen positivem Affekt und Selbstregulation aufweisen. Alter und Persönlichkeitsmerkmale können die verschiedenen Cluster vorhersagen. Diese Ergebnisse betonen die Bedeutung von computergestützter Modellierung bei der Untersuchung der Zusammenhänge zwischen kognitiven Fähigkeiten und anderen psychologischen Konstrukten. Weitere Forschung ist erforderlich, um die Mechanismen hinter altersbedingten Unterschieden in den Netzwerkstrukturen besser zu verstehen.