No description available.

No description available.

People have always been fascinated by the extent to which belief or will may influence<p>behavior. Proverbs, like “we tend to get what we expect,” and concepts, such as optimistic<p>thinking or self-fulfilling prophecy, reflect this intuition of an important link between one’s<p>dispositions and subsequent behavior. In other words, one’s predictions directly or<p>indirectly cause them to become true. In a similar manner, every culture, country or<p>religion has their own words for ‘expectation,’ ‘belief,’ ‘disappointment,’ ‘surprise,’ and<p>generally all have the same meaning: under uncertainty, what one expects or believes is the<p>most likely to happen. This relation between what caused a reaction in the past will<p>probably cause it again in the future might not be realistic. If the expected outcome is not<p>confirmed, it may result in a personal ‘disappointment’, and if the outcome fits no<p>expectations, it will be a ‘surprise’. Our brain is hardwired with this heuristic capacity of<p>learning the cause-effect relationship and to project its probability as the basis for much of<p>our behavior, as well as cognitions. This experience-based expectation is a form of<p>learning that helps the brain to bypass an exhaustive search in finding a satisfactory<p>solution. Expectations may thus be considered an innate theory of causality; that is, a set of<p>factors (causes) generating a given phenomenon (effects) influence the way we treat<p>incoming information but also the way we retrieve the stored information. These<p>expectancy templates may well represent one of the basic rules of how the brain processes<p>information, affecting the way we perceive the world, direct our attention and deal with<p>conflicting information. In fact, expectations have been shown to influence our judgments<p>and social interactions, along with our volition to individually decide and commit to a<p>particular course of action. However, people’s expectations may elicit the anticipation of<p>their own automatic reactions to various situations and behaviors cues, and can explain that<p>expecting to feel an increase in alertness after coffee consumption leads to experiencing<p>the consequent physiologic and behavioral states. We call this behavior-response<p>expectancy. This non-volitional form of expectation has been shown to influence<p>cognitions such as memory, pain, visual awareness, implicit learning and attention, through<p>the mediation of phenomena like placebo effects and hypnotic behaviors. Importantly,when talking about expectations, placebo and hypnosis, it is important to note that we are<p>also talking about suggestion and its modulating capability. In other words, suggestion has<p>the power to create response expectancies that activate automatic responses, which will, in<p>turn, influence cognition and behavior so as to shape them congruently with the expected<p>outcome. Accordingly, hypnotic inductions are a systematic manipulation of expectancy,<p>similar to placebo, and therefore they both work in a similar way. Considering such<p>assumptions, the major question we address in this PhD thesis is to know if these<p>expectancy-based mechanisms are capable of modulating more high-level information<p>processing such as cognitive conflict resolution, as is present in the well-known Stroop<p>task. In fact, in a recent series of studies, reduction or elimination of Stroop congruency<p>effects was obtained through suggestion and hypnotic induction. In this PhD thesis, it is<p>asked whether a suggestion reinforced by placebos, operating through response-expectancy<p>mechanisms, is able to induce a top-down cognitive modulation to overcome cognitive<p>conflict in the Stroop task, similar to those results found using suggestion and hypnosis<p>manipulation. / Doctorat en Sciences Psychologiques et de l'éducation / info:eu-repo/semantics/nonPublished

L'exclusion sociale est considérée comme l'une des situations les plus douloureuses pour les êtres humains. Les travaux dans ce domaine montrent que même de brefs épisodes d’ostracisme (paradigme du Cyberball) ont des effets importants aux niveaux neurophysiologique, émotionnel et comportemental, l’impact de cet ostracisme au niveau cognitif néanmoins n'a pas reçu beaucoup d'attention. Des résultats récents mettent en évidence une influence négative de l'ostracisme sur les marqueurs électrophysiologiques du contrôle cognitif, il n'y a cependant à ce jour aucune preuve directe d’une réduction de contrôle cognitif sous l’effet d’une exclusion sociale. Dans nos travaux nous avons étudié l'impact de l'ostracisme (Cyberball) sur le contrôle cognitif avec la tâche standard de Simon couplée à des analyses distributionnelles des temps de réaction auprès de populations caractérisées ou non par des expériences chroniques d’ostracisme (i.e., des étudiants ordinaires dans l’Étude 1, des personnes illettrées dans l'Étude 2 et des chômeurs de longue durée dans l'Étude 3). Dans les trois études, de brefs épisodes d'exclusion sociale suffisent à dégrader le niveau de satisfaction exprimé par les participants à l’égard des besoins fondamentaux (appartenance sociale, existence significative, estime de soi, contrôle des événements). Ces effets, cependant, s’avèrent réduits dans les populations chroniquement frappées d'ostracisme, suggérant leur moindre sensibilité à l'exclusion sociale en jeu dans le Cyberball. Plus important encore, cet ostracisme provoque chez les participants non stigmatisés une diminution du contrôle cognitif (Étude 1), mise en évidence dans nos travaux par un effet Simon stable (plutôt que réduit) sur les temps de réaction les plus longs pourtant les plus sensibles à l’expression d’un processus d'inhibition. Cependant, nos résultats ne montrent aucune différence de sensibilité entre les participants chroniquement ostracisés et leurs groupes contrôle (les Études 2 et 3), suggérant une certaine faiblesse du paradigme Cyberball auprès des personnes en situation d'exclusion sociale dans leur vie quotidienne. Nos résultats remettent donc en question la prédominance de ce paradigme pour la compréhension des effets cognitifs de l’exclusion sociale, au moins chez les individus caractérisés par un ostracisme chronique. / Impact on cognitive control, emotions and fundamental needsRésumé : Social exclusion is considered as one of the most painful situations for human beings. Past research showed that even brief episodes of ostracism (the Cyberball paradigm) have strong effects at the neurophysiological, emotional, and behavioral levels, its impact at the cognitive level however did not receive much attention. Recent findings revealed a negative influence of ostracism on electrophysiological markers of cognitive control, yet there is no direct evidence that being socially excluded reduces cognitive control. Here, we investigated the impact of ostracism (using the Cyberball) on cognitive control using a standard Simon task and distributional reaction time analyses with non-chronically-ostracized and chronically-ostracized populations (regular students in Study 1, illiterate people in Study 2, and long-term unemployed people in Study 3). In the three studies, brief episodes of social exclusion had negative effects on participants’ self-reports of fundamental needs' satisfaction (belonging, meaningful existence, self-esteem, and control). These effects, however, were substantially reduced in chronically-ostracized populations, suggesting that ostracism based on the Cyberball is a bit less meaningful for those populations. More importantly, this ostracism caused a transitory reduction in cognitive control in the non-chronically-ostracized participants (Study 1), as indicated by a stable (rather than decreased) Simon effect on longer reaction times where inhibition yet is more likely. However, we found no evidence of a differential sensitivity between the chronically-ostracized participants and their control groups (Study 2 and Study 3), suggesting that the Cyberball paradigm is not powerful enough with people experiencing social exclusion in their ordinary life. Our findings therefore call into question the predominance of the Cyberball paradigm for our understanding of the cognitive effects of ostracism, at least in chronically ostracized-individuals.

A number of studies have concluded that cognitive control is not fully established until late adolescence. The precise differences in brain function between adults and adolescents with respect to cognitive control, however, remain unclear. To address this issue, we conducted a study in which 185 adolescents (mean age (SD) 14.6 (0.3) years) and 28 adults (mean age (SD) 25.2 (6.3) years) performed a single task that included both a stimulus-response (S-R) interference component and a task-switching component. Behavioural responses (i.e. reaction time, RT; error rate, ER) and brain activity during correct, error and post-error trials, detected by functional magnetic resonance imaging (fMRI), were measured. Behaviourally, RT and ER were significantly higher in incongruent than in congruent trials and in switch than in repeat trials. The two groups did not differ in RT during correct trials, but adolescents had a significantly higher ER than adults. In line with similar RTs, brain responses during correct trials did not differ between groups, indicating that adolescents and adults engage the same cognitive control network to successfully overcome S-R interference or task switches. Interestingly, adolescents with stronger brain activation in the bilateral insulae during error trials and in fronto-parietal regions of the cognitive control network during post-error trials did have lower ERs. This indicates that those mid-adolescents who commit fewer errors are better at monitoring their performance, and after detecting errors are more capable of flexibly allocating further cognitive control resources. Although we did not detect a convincing neural correlate of the observed behavioural differences between adolescents and adults, the revealed interindividual differences in adolescents might at least in part be due to brain development.

Schizophrenia is a severe neuropsychiatric disorder characterized by positive, negative and cognitive symptoms with poor functional outcomes, placing an enormous burden on the individual, caregivers and society. Although deficits in cognition are an integral part of the disease and the best predictor of functional outcomes, there is as yet no established treatment addressing them. Avoidance of a hidden place on a continuously rotating arena (Carousel) requires cognitive control and is a rodent model of cognitive coordination of information from dissociated spatial frames, which is impaired in acute pharmacological and neurodevelopmental model of schizophrenia. Cognitive training on the Carousel during adolescence alleviates adult cognitive deficit in a neurodevelopmental model of schizophrenia and improves neural coordination (oscilations in the beta and gamma band), which is thought to be necessary for cognition. We examined if cognitive training during adolescence eliminates the deficit in neuronal coordination observed in adult rats after acute systemic NMDA receptor antagonist MK-801 (0.15 mg/kg). During adolescence, rats were either trained in spatial avoidance on Carousel or merely handled. As adults, rats received two 5-min exploration sessions in the same (A/A) or in two distinct...

Restless legs syndrome (RLS) is a sensory-motor disorder characterized by abnormal circadian rhythm with an increase in the severity of sensory and motor symptoms at night. Even though many neurological diseases have shown a strong nexus between motor and cognitive symptoms, to date, cognitive functions especially cognitive control in RLS has been poorly understood. Given that cognitive control is a key to leading a self-serving and successful life, including many aspects of employment, social life, and attaining long-term goals, this thesis aimed to examine cognitive control and the underlying mechanisms in RLS. Thalamic gamma aminobutyric acid (GABA), which has been linked to RLS sensory-motor symptoms, also plays an important role in cognitive control. Therefore, the potential relationship between thalamic GABA level and cognitive control in RLS was examined (Study I). RLS patients displayed reduced working memory-based control performances as compared to healthy controls. Elevated thalamic GABA was found to attenuate the observed control deficits in RLS, even though changes in thalamic GABA levels might not be the ultimate causes of these deficits. According to the modulatory effect of thalamic GABA on thalamic activity and thalamo-cortical connectivity, relatively higher GABA levels may have helped RLS patients compensate for their pathological changes such as thalamic hyperactivity and hypoconnectivity, which may underpin the observed control deficits. The critical feature of RLS, abnormal circadian rhythm is thought to be related to nocturnal striatal dopamine deficiency. Concerning the dopaminergic modulation of cognitive control, the circadian variation of cognitive control processes has been investigated (Study II & III). RLS patients displayed reduced attentional control (Study II) and automatic response activation (Study III) at night, which resulted from decreased activation within the extra-striate visual cortex, the superior parietal cortex, and the premotor cortex. As there were no activity changes within the prefrontal cortex, it is likely that cortico-basal ganglia cognitive loops were less prone to RLS. Instead, striatal dopamine deficiency at night may have influenced the cortico-cortical functional connectivity and cortico-basal ganglia motor loops in RLS. These findings not only shed light on the underlying mechanisms of cognitive control, but also advance early clinical treatment possibilities for cognitive changes in RLS patients. Furthermore, recent insights into daytime-related cognition may help patients develop a suitable daytime schedule to minimize the detrimental effects induced by cognitive deficits.

Making decisions is a dynamic process. Especially when we face a decision between conflicting options, different forces seem to drag our mind from one option to the other one (James, 1890), again and again. This process may last for a long time, sometimes only coming to a decision when we are finally forced to choose, e.g. by an important deadline. Psychology and many other disciplines were interested in how humans make decisions from their beginnings on. Many different influences on decisions were discovered (e.g. Kahneman & Tversky, 1979; Todd & Gigerenzer, 2000). In the face of these advances, it seems odd, that knowledge about the ongoing process of reaching a decision is rare and much of the investigation has focused on the final outcome of choice situations (Townsend & Busemeyer, 1995). A very recent approach, called neuroeconomics, started out to investigate what happens behind the scenes of a final decision. Using modern neuroimaging methods, many neuroeconomists explain decision making in the brain in terms of a hierarchy of different neural modules that work together like a big corporation to finally make the best possible decision (Sanfey, Loewenstein, McClure, & Cohen, 2006). However, the focus on neural modules also limits this approach to a quite static view of decision making and many questions, related to the dynamic aspects of decision making, still remain open: How do we continuously control impulsive or habitual tendencies in our decisions when we pursue long-term goals? How do we shift attention back and forth between (goal) relevant properties of choice options? How do we adjust and readjust our focus of attention to relevant information in order to avoid distraction by irrelevant or misleading information? And how are we influenced by the environmental context when we make decisions? The present work aims to show how an approach based on the concepts of dynamic systems theory could complement the module oriented approach and enhance our knowledge of the processes of decision making. Chapter 2 elaborates the limits of the module oriented approach, with a special focus on decisions under conflict, when we are faced with conflicting information, and introduces the principles of a complementary dynamic approach. Chapter 3 deduces the dynamic hypothesis of this work: ongoing processes interactions at different time scales can explain specific cognitive functions without postulating specialized modules for this function. To approach this hypothesis, chapter 4 will develop a theoretical and empirical framework to study decision making dynamically. The empirical part, building on the empirical framework, starts with chapter 5 presenting an EEG experiment. Chapter 6 presents two mouse tracking experiments, and chapter 7 presents a modelling study, reproducing the empirical data of chapters 5 and 6. The general discussion in chapter 8 summarizes the theoretical and empirical results and discusses possible limitations. Finally, chapter 9 discusses the implications of the dynamic approach to decision making, presents an outlook on future research projects, and closes the work by offering a dynamic picture of the processes behind the stage of a final decision.:Statement I Brief Contents III Contents V Figures IX Chapter 1 Introduction 1 Chapter 2 Decision making under conflict 3 Chapter 3 Investigating decision making under conflict dynamically 14 Chapter 4 Making decisions with a continuous mind 17 Chapter 5 The dynamics of cognitive control: evidence for within trial conflict adaptation from frequency tagged EEG 56 Chapter 6 How decisions evolve: the temporal dynamics of action selection 77 Chapter 7 Dynamic goal states: adapting cognitive control at different time scales without conflict monitoring 97 Chapter 8 General discussion 115 Chapter 9 Conclusion and outlook 123 References 130 Deutsche Zusammenfassung 153 Appendix I Supplementary material for chapter 5 159 Appendix II Model formulas for chapter 7 163 / „Man kann nicht beides haben: Den Rahm und die Butter.“ - „Wer die Wahl hat, hat die Qual.“ Mit diesen Sprichwörtern beklagt der Volksmund, womit das Leben uns immer wieder konfrontiert: wir müssen entscheiden, und oftmals führt uns das in Entscheidungskonflikte. Im Dilemma solcher Konflikte mag es begründet sein, dass das Thema der vorliegenden Arbeit, die Entscheidungsforschung, nicht nur in der Psychologie schon immer eine wichtige Rolle spielte, sondern auch in anderen Disziplinen, wie der Ökonomie, der angewandten Mathematik und der Philosophie. Die langjährigen Bestrebungen, diese unterschiedlichen Fachbereiche zu integrieren (z.B. Kahneman & Tversky, 1979; von Neumann & Morgenstern, 1944; Savage, 1972), münden aktuell in das Forschungsgebiet der Neuroökonomie (Camerer, Loewenstein, & Prelec, 2005; Loewenstein, Rick, & Cohen, 2008; Sanfey, Loewenstein, McClure, & Cohen, 2006). Neuroökonomen nutzen vielfach die Methoden der bildgebenden Hirnforschung, um durch die Lokalisierung der neuronalen Basis hierarchisch gegliederter Module Entscheidungsprozesse zu erklären (z.B. Sanfey et al., 2006; Fellows, 2004). Während die Anwendung bildgebender Methoden Potential birgt (z.B. Harrison, 2008), ist es vor allem der modulorientierte Ansatz, der das Risiko einer zu eingeschränkten Sichtweise auf Entscheidungsprozesse trägt (z.B. Ortmann, 2008; Oullier & Kelso, 2006). Dies zeigt sich zum Beispiel im von der kognitiven Psychologie intensiv erforschten Bereich von Entscheidungen unter Konflikt. Eine zentrale Rolle bei dieser Art von Entscheidungen spielen kognitive Kontrollprozesse, die der Umsetzung zielorientierten Verhaltens (Norman & Shallice, 2000) durch Konfliktlösung und -anpassung dienen. Als Bindeglied dieser beiden Prozesse gilt die Detektion von Entscheidungskonflikten, welche die vorherrschende Conflict Monitoring Theory (Botvinick, Braver, Barch, Carter, & Cohen, 2001) entsprechend dem modulorientiertem Ansatz einem speziellen neuronalen Modul zuordnet, das im anterioren cingulären Cortex lokalisiert ist (Botvinick, Cohen, & Carter, 2004). Die Probleme eines einseitigen modulorientierten Ansatzes verdeutlichen hier unter anderem die widersprüchliche Befundlage (z.B. Mansouri, Tanaka, & Buckley, 2009) und die letztlich weiterhin ungeklärte Frage nach den zugrundeliegenden Prozessen. Die Arbeit hat deshalb zum Ziel, den modulorientierten Ansatz um einen komplementären Ansatz auf Basis der Theorie dynamischer Systeme (Dynamical Systems Theory, DST) zu ergänzen. Aus dem grundlegenden DST-Prinzip der kontinuierlichen (z.B. Spivey, 2007) Interaktion rückgekoppelter Komponenten (z.B. Kelso, 1995; Van Orden, Holden, & Turvey, 2003) wird zunächst die dynamische Hypothese abgeleitet, dass sich Effekte auf verschiedenen Zeitskalen gegenseitig bedingen und einander hervorbringen. Für Entscheidungen unter Konflikt bedeutet dies, dass sich die Prozesse der Konfliktlösung und anpassung durch ihre direkte Interaktion im kognitiven System gegenseitig erzeugen. Zur Überprüfung dieser Hypothese werden innerhalb der Arbeit generelle empirische Strategien entwickelt, welche die Untersuchung von Entscheidungsprozessen auf verschiedenen Zeitskalen ermöglichen. Im empirischen Teil der Arbeit werden sodann zwei dieser Strategien zur Anwendung gebracht, um den Erkenntnisgewinn des dynamischen Ansatzes zu illustrieren. Zunächst wird in einer EEG-Studie eine Frequency-Tagging-Methode (z.B. Müller & Hübner, 2002; Müller, Andersen, & Keil, 2007) auf die Untersuchung der kognitiven Kontrollprozesse in einer Flanker-Aufgabe (Eriksen & Eriksen, 1974) adaptiert. Die neue Kombination einer kontinuierlichen neurophysiologischen Methode und eines klassischen Konflikt-Paradigmas ermöglicht die gleichzeitige Untersuchung kontinuierlicher Veränderungen der Aufmerksamkeit auf relevante und irrelevante Information. Die Ergebnisse der Studie stützen die Hypothese einer direkten Interaktion von Prozessen der Konfliktlösung und -anpassung und stellen bereits einen Widerspruch zur Conflict Monitoring Theory dar. Als weitere empirische Strategie wird in zwei Experimenten die Methode des Maus-Tracking (z.B. Buetti & Kerzel, 2009; Song & Nakayama, 2009; Spivey, Grosjean, & Knoblich, 2005) im Rahmen einer Simon-Aufgabe (Simon, 1969) eingesetzt. Die erneute Kombination einer kontinuierlichen Methode, diesmal auf Reaktionsebene, mit einem klassischen Konflikt-Paradigma erlaubt die Messung von Verhaltenstendenzen im Verlauf des gesamten Entscheidungsprozesses. Mit Hilfe einer neu entwickelten regressionsbasierten Analysemethode werden die Subprozesse einzelner Entscheidungen separiert und Einblicke in die Dynamik von Konfliktlösung und -anpassung gewonnen. Die Ergebnisse zeigen ein komplexes Muster zeitlicher Interaktion zwischen den beiden kognitiven Kontrollprozessen, wobei die Konfliktanpassung zeitlich unabhängig von der Verarbeitung irrelevanter Information ist. Dies steht erneut im Widerspruch zu Annahmen der Conflict Monitoring Theory. Zusammenfassend stützen die empirischen Ergebnisse die dynamische Hypothese der kontinuierlichen Interaktion rückgekoppelter Komponenten und werden im nächsten Schritt in einem dynamisch-konnektionistischen Netzwerkmodell integriert. Als Alternative zum Modell der Conflict Monitoring Theory verzichtet es entsprechend dem dynamischen Ansatz auf ein Conflict Monitoring Modul (Botvinick et al., 2001). Es verfügt stattdessen über Verarbeitungs-Prozesse auf verschiedenen Zeitskalen (Kiebel, Daunizeau, & Friston, 2008) und eine Rückkopplung zwischen der Netzwerkschicht, die der Informationsverarbeitung dient, und jener, die der Zielrepräsentation dient (Gilbert & Shallice, 2002; Cohen & Huston, 1994). Die Ergebnisse der Simulation zeigen, dass das Modell sowohl die klassischen Befunde zur Konfliktlösung und anpassung (z.B. Gratton, Coles, & Donchin, 1992), als auch das in den empirischen Studien gefundene kontinuierliche Datenmuster von Entscheidungsprozessen reproduziert. Die empirischen Befunde und die Ergebnisse der Modellierung bestätigen somit die postulierte dynamische Hypothese, dass sich Effekte auf verschiedenen Zeitskalen gegenseitig bedingen und einander hervorbringen. Dies verdeutlicht den komplementären Wert des dynamischen Ansatzes zum modulorientierten Ansatz, welcher vielfach in der Neuroökonomie verfolgt wird. Der hier entwickelte DST-basierte Ansatz bietet somit sowohl ein komplementäres Denkmodell, welches wie der modulorientierte Ansatz eine Verbindung zwischen den Phänomenen auf neuronaler und Verhaltensebene herstellt, als auch neue empirische Methoden zur dynamischen Erforschung von Entscheidungen. Daraus wird abschließend eine Fokuserweiterung für die zukünftige Forschung abgeleitet: zum einen auf die kontinuierlichen Prozesse, welche zu einer Entscheidung führen, und zum anderen auf die Interaktionsdynamik dieser Prozesse. Die Arbeit schließt mit dem Bild eines Entscheidungsprozesses als einer selbstorganisierten, metastabilen Balance (z.B. Kelso, 1995) bei der Lösung verschiedener Entscheidungsdilemmata (Goschke, 2003).:Statement I Brief Contents III Contents V Figures IX Chapter 1 Introduction 1 Chapter 2 Decision making under conflict 3 Chapter 3 Investigating decision making under conflict dynamically 14 Chapter 4 Making decisions with a continuous mind 17 Chapter 5 The dynamics of cognitive control: evidence for within trial conflict adaptation from frequency tagged EEG 56 Chapter 6 How decisions evolve: the temporal dynamics of action selection 77 Chapter 7 Dynamic goal states: adapting cognitive control at different time scales without conflict monitoring 97 Chapter 8 General discussion 115 Chapter 9 Conclusion and outlook 123 References 130 Deutsche Zusammenfassung 153 Appendix I Supplementary material for chapter 5 159 Appendix II Model formulas for chapter 7 163

Alcohol use disorder (AUD) is characterized by excessive habitual drinking and loss of control over alcohol intake despite negative consequences. Both of these aspects foster uncontrolled drinking and high relapse rates in AUD patients. Yet, common interventions mostly focus on the phenomenological level, and prioritize the reduction of craving and withdrawal symptoms. Our review provides a mechanistic understanding of AUD and suggests alternative therapeutic approaches targeting the mechanisms underlying dysfunctional alcohol-related behaviours. Specifically, we explain how repeated drinking fosters the development of rigid drinking habits and is associated with diminished cognitive control. These behavioural and cognitive effects are then functionally related to the neurobiochemical effects of alcohol abuse. We further explain how alterations in fronto-striatal network activity may constitute the neurobiological correlates of these alcohol-related dysfunctions. Finally, we discuss limitations in current pharmacological AUD therapies and suggest non-invasive brain stimulation (like TMS and tDCS interventions) as a potential addition/alternative for modulating the activation of both cortical and subcortical areas to help re-establish the functional balance between controlled and automatic behaviour.

La théorie du processus dual stipule que la cognition humaine fonctionne selon deux modes distincts : l’un pour le traitement rapide, habituel et associatif, appelé communément "système 1" et le second, ayant un traitement plus lent, délibéré et contrôlé, que l’on nomme "système 2". Cette distinction indique une caractéristique sous-jacente importante de la cognition humaine : la possibilité de passer de manière adaptative à différentes stratégies de calcul selon la situation. Cette capacité est étudiée depuis longtemps dans différents domaines et de nombreux bénéfices hypothétiques semblent y être liés. Cependant, les réseaux neuronaux profonds sont souvent construits sans cette capacité à gérer leurs ressources calculatoires de manière optimale. Cette limitation des modèles actuels est d’autant plus préoccupante que de plus en plus de travaux récents semblent montrer une relation linéaire entre la capacité de calcul utilisé et les performances du modèle lors de la phase d’évaluation. Pour résoudre ce problème, ce mémoire propose différentes approches et étudie leurs impacts sur les modèles, tout d’abord, nous étudions un agent d’apprentissage par renforcement profond qui est capable d’allouer plus de calcul aux situations plus difficiles. Notre approche permet à l’agent d’adapter ses ressources computationnelles en fonction des exigences de la situation dans laquelle il se trouve, ce qui permet en plus d’améliorer le temps de calcul, améliore le transfert entre des tâches connexes et la capacité de généralisation. L’idée centrale commune à toutes nos approches est basée sur les théories du coût de l’effort venant de la littérature sur le contrôle cognitif qui stipule qu’en rendant l’utilisation de ressource cognitive couteuse pour l’agent et en lui laissant la possibilité de les allouer lors de ses décisions il va lui-même apprendre à déployer sa capacité de calcul de façon optimale. Ensuite, nous étudions des variations de la méthode sur une tâche référence d’apprentissage profond afin d’analyser précisément le comportement du modèle et quels sont précisément les bénéfices d’adopter une telle approche. Nous créons aussi notre propre tâche "Stroop MNIST" inspiré par le test de Stroop utilisé en psychologie afin de valider certaines hypothèses sur le comportement des réseaux neuronaux employant notre méthode. Nous finissons par mettre en lumière les liens forts qui existent entre apprentissage dual et les méthodes de distillation des connaissances. Notre approche a la particularité d’économiser des ressources computationnelles lors de la phase d’inférence. Enfin, dans la partie finale, nous concluons en mettant en lumière les contributions du mémoire, nous détaillons aussi des travaux futurs, nous approchons le problème avec les modèles basés sur l’énergie, en apprenant un paysage d’énergie lors de l’entrainement, le modèle peut ensuite lors de l’inférence employer une capacité de calcul dépendant de la difficulté de l’exemple auquel il fait face plutôt qu’une simple propagation avant fixe ayant systématiquement le même coût calculatoire. Bien qu’ayant eu des résultats expérimentaux infructueux, nous analysons les promesses que peuvent tenir une telle approche et nous émettons des hypothèses sur les améliorations potentielles à effectuer. Nous espérons, avec nos contributions, ouvrir la voie vers des algorithmes faisant un meilleur usage de leurs ressources computationnelles et devenant par conséquent plus efficace en termes de coût et de performance, ainsi que permettre une compréhension plus intime des liens qui existent entre certaines méthodes en apprentissage machine et la théorie du processus dual. / The dual-process theory states that human cognition operates in two distinct modes: one for rapid, habitual and associative processing, commonly referred to as "system 1", and the second, with slower, deliberate and controlled processing, which we call "system 2". This distinction points to an important underlying feature of human cognition: the ability to switch adaptively to different computational strategies depending on the situation. This ability has long been studied in various fields, and many hypothetical benefits seem to be linked to it. However, deep neural networks are often built without this ability to optimally manage their computational resources. This limitation of current models is all the more worrying as more and more recent work seems to show a linear relationship between the computational capacity used and model performance during the evaluation phase. To solve this problem, this thesis proposes different approaches and studies their impact on models. First, we study a deep reinforcement learning agent that is able to allocate more computation to more difficult situations. Our approach allows the agent to adapt its computational resources according to the demands of the situation in which it finds itself, which in addition to improving computation time, enhances transfer between related tasks and generalization capacity. The central idea common to all our approaches is based on cost-of-effort theories from the cognitive control literature, which stipulate that by making the use of cognitive resources costly for the agent, and allowing it to allocate them when making decisions, it will itself learn to deploy its computational capacity optimally. We then study variations of the method on a reference deep learning task, to analyze precisely how the model behaves and what the benefits of adopting such an approach are. We also create our own task "Stroop MNIST" inspired by the Stroop test used in psychology to validate certain hypotheses about the behavior of neural networks employing our method. We end by highlighting the strong links between dual learning and knowledge distillation methods. Finally, we approach the problem with energy-based models, by learning an energy landscape during training, the model can then during inference employ a computational capacity dependent on the difficulty of the example it is dealing with rather than a simple fixed forward propagation having systematically the same computational cost. Despite unsuccessful experimental results, we analyze the promise of such an approach and speculate on potential improvements. With our contributions, we hope to pave the way for algorithms that make better use of their computational resources, and thus become more efficient in terms of cost and performance, as well as providing a more intimate understanding of the links that exist between certain machine learning methods and dual process theory.