1,721,009 research outputs found
Experiencing free will: electrophysiological correlates of preparation and monitoring of intentional actions
Full text linkIn the last decades there has been growing interest in cognitive neuroscience for the understanding of neural underpinnings of voluntary motor actions. The interest in willed behaviour is somehow fuelled by the philosophical problem of free will. A fundamental aspect of the experience of free will is the experience of intention, that is, the experience of planning or being about to do something.
The aim of the present thesis was to examine neurophysiological processes associated with preparation and monitoring of intentional actions, by employing Event-Related Potentials (ERP).
In Experiment 1, I examined whether action-monitoring (i.e. processes reflecting the monitoring of the consequences of actions) is involved in the subjective experience of intention. In particular, an intentional action task and an action-monitoring approach were combined in order to investigate whether post-action ERP components reflecting action-monitoring are involved in people’s experience of ‘when’ they become aware of their intention to act. Although the idea that post-action events can influence the reported time of intentions might seem counterintuitive, empirical evidence suggests that intentions can be partially inferred from events occurring after action execution. Here it was demonstrated that the time at which people become aware of their intention to act is partially inferred from the apparent time of the motor response, rather than the actual response. In addition, a specific ERP component, namely action-effect negativity (Nae), was found to reflect the comparison between the representation of the expected action effect and the actual effect. These findings suggest that conscious intentions are not entirely based on action preparation processes, but they are partially inferred from post-action events.
Experiment 2 extended this finding by showing that ERP components reflecting motor preparation and action-monitoring are specifically enhanced in intentional actions. Participants performed self-paced voluntary actions and attended either to their intention to act or to the actual movement. When they attended to their intention, brain activity reflecting motor preparation – i.e. the readiness potential (RP) – was increased. This result confirms previous evidence that preparation of intentional actions involves the anticipation of the effects of the action itself and that the representation of the intended effect is reflected by the RP. Also the Nae was larger when participants attended to their intention, as compared to when they attended to the movement. This finding is taken as evidence that action-monitoring plays a crucial role in binding together the representation of the intended outcome and the actual action-effect.
Experiment 3 aimed at investigating whether brain correlates of intentional motor preparation can be influenced by abstract beliefs such as beliefs about free will. Neurophysiological activity was recorded while participants executed self-paced key presses and we found that the RP, that reflects intentional action preparation, was reduced in individuals previously induced to disbelieve in human will. This effect was evident more than 1 second before participants consciously decided to move, suggesting that our manipulation affects intentional actions at a preconscious stage. These findings indicate that abstract belief systems might have a much more fundamental impact than we ever thought.Negli ultimi decenni si è sviluppato, nell’ambito delle neuroscienze cognitive, un crescente interesse per la comprensione delle basi neurofisiologiche delle azioni intenzionali. Il comportamento intenzionale, o volontario, è strettamente connesso al problema filosofico del libero arbitrio. Un aspetto importante dell’esperienza volitiva è l’esperienza di intenzione, che può essere definita come la consapevolezza di pianificare o di essere sul punto di fare qualcosa volontariamente.
L’obiettivo della ricerca era di studiare i processi neurofisiologici associati alla preparazione e al monitoraggio delle azioni volontarie tramite l’utilizzo dei Potenziali Evento-Relati (ERP).
Nell’Esperimento 1, si è voluto indagare se i processi che riflettono il monitoraggio degli effetti di un’azione motoria siano implicati nell’esperienza soggettiva di intenzione. In particolare, un compito ideato per lo studio dell’azione intenzionale è stato combinato con un approccio derivato dalla letteratura sul monitoraggio dell’azione, al fine di esaminare se componenti ERP seguenti all’azione motoria fossero implicate nell’esperienza di quando la persona ritiene di aver avuto l’intenzione di agire. L’idea che eventi successivi all’azione possano influenzare l’esperienza dell’intenzione, può sembrare controintuitiva; tuttavia evidenze empiriche hanno dimostrato che le intenzioni possono basarsi, almeno parzialmente, su un processo inferenziale che deriva dalla valutazione di eventi successivi all’esecuzione dell’azione. I risultati dell’esperimento hanno dimostrato che nel riportare quando hanno avuto l’intenzione, i soggetti erano influenzati dalla risposta apparente, derivante da una manipolazione del feedback uditivo, piuttosto che dalla effettiva risposta motoria. Inoltre, una specifica componente ERP, denominata action-effect negativity (Nae), era legata al confronto tra la rappresentazione degli effetti attesi dell’azione e quegli effettivi. Questi risultati dimostrano che le intenzioni coscienti non sono basate solamente su processi legati alla preparazione dell’azione, ma sono influenzate anche da processi di tipo inferenziale.
Nell’Esperimento 2 sono stati approfonditi gli aspetti legati al monitoraggio dell’azione indagati nel primo esperimento. È stato dimostrato che componenti ERP associate sia alla preparazione che al monitoraggio dell’azione motoria sono più pronunciate nelle azioni intenzionali. I soggetti eseguivano dei semplici movimenti (pressione pulsante), in modo del tutto volontario e senza costrizioni temporali. Durante il compito, i soggetti dovevano prestare attenzione alla loro intenzione ricompiere il movimento oppure al movimento stesso. Quando prestavano attenzione all’intenzione, l’attività neurofisiologica associata alla preparazione motoria, rappresentata dal readiness potential (RP), era maggiore. In linea con precedenti evidenze sperimentali, questo risultato indica che nelle azioni intenzionali gli effetti dell’azione stessa vengono anticipati e che la rappresentazione degli effetti è associata al RP. Inoltre, anche l’ampiezza della Nae era maggiore quando i soggetti prestavano attenzione all’intenzione, rispetto a quando essi prestavano attenzione al movimento stesso. Da una parte, questo risultato suggerisce che il monitoraggio dell’azione ha un ruolo nel confrontare la rappresentazione degli effetti attesi e la rappresentazione degli effetti effettivi; dall’altra, enfatizza il ruolo dei processi di monitoraggio nell’esperienza soggettiva dell’azione intenzionale.
L’Esperimento 3 aveva l’obiettivo di indagare se i correlati neurofisiologici di preparazione motoria possono essere modulati da credenze astratte sul libero arbitrio. È stata registrata l’attività neurofisiologica mentre ai soggetti veniva chiesto di premere a piacimento un pulsante, senza alcuna costrizione temporale. È stato evidenziato che i soggetti indotti a credere che il libero arbitrio è un’illusione mostravano un ridotto RP. Questo effetto, che dimostra una riduzione dell’attività neurofisiologica associata alla preparazione del movimento, era evidente più di un secondo prima che i soggetti decidessero di effettuare il movimento. Ciò suggerisce che indurre una prospettiva deterministica, in cui il libero arbitrio viene considerato un’illusione, ha un effetto nelle stadi pre-consci della preparazione delle azioni intenzionali. Questi risultati dimostrano che sistemi astratti di credenze, come la credenza nel libero arbitrio, hanno un impatto ad un livello molto basilare del comportamento umano
Understanding Multimedia Content with Prior Knowledge
No full textVisual-textual grounding is a challenging task that involves associating language with visual objects or scenes, and it has become a popular research area due to its importance in various applications. Traditionally, visual-textual grounding has been solved by relying on information from images and textual phrases. However, incorporating additional prior knowledge, such as a graph, could potentially enhance the performance and accuracy of visual-textual grounding models. The graph is a discrete structure that can represent any kind of information that can be used to solve the grounding task.
In this Ph.D. thesis, a formal probabilistic framework is proposed to consider all three modalities: image, text, and graph. The framework allows for the analysis of existing works and the development of a novel approach to visual-textual grounding based on an innovative factorization of probabilities. The adoption of the probabilistic approach is crucial for accounting for the inherent uncertainties in solving the task.
In addition, this thesis presents two contributions to improve the traditional visual-textual grounding task. The first contribution regards a new loss function for training visual-textual grounding models in a supervised setting. Indeed, the models in the literature are typically constituted by two main components that focus on how to learn useful multi-modal features for grounding and how to improve the predicted bounding box of the visual mention, respectively. Finding the right learning balance between these two sub-tasks is not easy, and the current models are not necessarily optimal with respect to this issue.
The second contribution consists of a model tackling the weakly-supervised visual-textual grounding. The proposed model is based on the principle of first predicting a rough alignment among phrases and boxes, adopting a module that does not require training, and then refining those alignments using a learnable neural network. The model is trained to maximize the multimodal similarity between an image and a sentence describing that image while minimizing the multimodal similarity of the same sentence and a new unrelated image, carefully selected so as to help as much as possible during training.
The object detector plays a fundamental role in solving the visual-textual grounding task. It should be able to identify many different objects and classify them correctly. Nevertheless, increasing the number of objects to be recognized usually leads to a more challenging classification problem. The importance of the correct classification of an object is even greater when considering the graph in the resolution of the visual-textual grounding task. In fact, the semantic information conveyed through the classes is crucial to identify the graph nodes that best characterize the objects depicted in the image. In literature, the most common approach is to use an object detector trained to detect 1600 different classes of objects. However, those classes are noisy and impair the performance of the object detector. To solve this problem, this document proposes also a new set of clean labels to use for training object detectors on the Visual Genome dataset.
To conclude, this thesis introduces a new object detector that can be conditioned by nodes of the WordNet graph to search for objects in images. In particular, the conditioned object detector can be deployed to estimate a component of the probability distribution factorization designed thanks to the probabilistic framework.
Overall, this Ph.D. thesis contributes to the study of visual-textual grounding and provides tools and insights that have the potential for developing advanced approaches and applications within this domain.Visual-textual grounding is a challenging task that involves associating language with visual objects or scenes, and it has become a popular research area due to its importance in various applications. Traditionally, visual-textual grounding has been solved by relying on information from images and textual phrases. However, incorporating additional prior knowledge, such as a graph, could potentially enhance the performance and accuracy of visual-textual grounding models. The graph is a discrete structure that can represent any kind of information that can be used to solve the grounding task.
In this Ph.D. thesis, a formal probabilistic framework is proposed to consider all three modalities: image, text, and graph. The framework allows for the analysis of existing works and the development of a novel approach to visual-textual grounding based on an innovative factorization of probabilities. The adoption of the probabilistic approach is crucial for accounting for the inherent uncertainties in solving the task.
In addition, this thesis presents two contributions to improve the traditional visual-textual grounding task. The first contribution regards a new loss function for training visual-textual grounding models in a supervised setting. Indeed, the models in the literature are typically constituted by two main components that focus on how to learn useful multi-modal features for grounding and how to improve the predicted bounding box of the visual mention, respectively. Finding the right learning balance between these two sub-tasks is not easy, and the current models are not necessarily optimal with respect to this issue.
The second contribution consists of a model tackling the weakly-supervised visual-textual grounding. The proposed model is based on the principle of first predicting a rough alignment among phrases and boxes, adopting a module that does not require training, and then refining those alignments using a learnable neural network. The model is trained to maximize the multimodal similarity between an image and a sentence describing that image while minimizing the multimodal similarity of the same sentence and a new unrelated image, carefully selected so as to help as much as possible during training.
The object detector plays a fundamental role in solving the visual-textual grounding task. It should be able to identify many different objects and classify them correctly. Nevertheless, increasing the number of objects to be recognized usually leads to a more challenging classification problem. The importance of the correct classification of an object is even greater when considering the graph in the resolution of the visual-textual grounding task. In fact, the semantic information conveyed through the classes is crucial to identify the graph nodes that best characterize the objects depicted in the image. In literature, the most common approach is to use an object detector trained to detect 1600 different classes of objects. However, those classes are noisy and impair the performance of the object detector. To solve this problem, this document proposes also a new set of clean labels to use for training object detectors on the Visual Genome dataset.
To conclude, this thesis introduces a new object detector that can be conditioned by nodes of the WordNet graph to search for objects in images. In particular, the conditioned object detector can be deployed to estimate a component of the probability distribution factorization designed thanks to the probabilistic framework.
Overall, this Ph.D. thesis contributes to the study of visual-textual grounding and provides tools and insights that have the potential for developing advanced approaches and applications within this domain
RGCVAE: relational graph conditioned variational autoencoder for molecule design
No full textIdentifying molecules that exhibit some pre-specified properties is a difficult problem to solve. In the last few years, deep generative models have been used for molecule genera- tion. Deep Graph Variational Autoencoders are among the most powerful machine learning tools with which it is possible to address this problem. However, existing methods strug- gle to capture the true data distribution and tend to be computationally expensive. In this work, we propose RGCVAE, an efficient and effective Graph Variational Autoencoder based on: (i) an encoding network exploiting a new powerful Relational Graph Isomor- phism Network; (ii) a novel probabilistic decoding component. Compared to several State- of-the-Art VAE methods on two widely adopted datasets, RGCVAE shows State-of-the-Art molecule generation performance while being significantly faster to train. The Python code implementing RGCVAE is openly accessible for download at: https://github.com/drigoni/ RGCVAE
Post-action determinants of the reported time of conscious intentions
Full text linkContains fulltext :
90759.pdf (Publisher’s version ) (Open Access)The question of whether our behavior is guided by our conscious intentions is gaining momentum within the field of cognitive neuroscience. It has been demonstrated that the subjective experience that conscious intentions are the driving force of our actions, is built partially on a post hoc reconstruction. Our hypothesis was that this reconstructive process is mediated by an action-monitoring system that compares the predicted and the actual sensory consequences of an action. We applied event-related potentials (ERP) to a variant of the Libet's task in which participants were asked to press a button and to report the time of decision-will judgment (W) -to press. We provided delayed auditory feedbacks after participants' action to signify an action time later than the actual action. We found that auditory feedbacks evoked a negative component in the 250-300 time range, namely action-effect negativity (N-AE), that is thought to reflect the activity of a system that detects violation from expectancies. We showed that the amplitude of the N-AE was sensitive to the delay of the auditory feedback, with a larger amplitude for more delayed feedbacks. Furthermore, changes in the N-AE were also associated with changes in the reported W. These results not only confirm that we infer the time we decided to act from events occurring after the response, but these results also indicate that the subjective experience of when an action is decided is influenced by the activity of an action-monitoring system that detects mismatches between predicted and actual sensory consequences of the actions.9 p
Happiness in action: The impact of positive affect on the time of the conscious intention to act
Full text linkThe temporal relationship between our conscious intentions to act and the action itself has been widely investigated. Previous research consistently shows that the motor intention enters awareness a few hundred milliseconds before movement onset. As research in other domains has shown that most behavior is affected by the emotional state people are in, it is remarkable that the role of emotional states on intention awareness has never been investigated. Here we tested the hypothesis that positive and negative affects have opposite effects on the temporal relationship between the conscious intention to act and the action itself. A mood induction procedure that combined guided imagery and music listening was employed to induce positive, negative, or neutral affective states. After each mood induction session, participants were asked to execute voluntary self-paced movements and to report when they formed the intention to act. Exposure to pleasant material, as compared to exposure to unpleasant material, enhanced positive affect and dampened negative affect. Importantly, in the positive affect condition participants reported their intention to act earlier in time with respect to action onset, as compared to when they were in the negative or in the neutral affect conditions. Conversely the reported time of the intention to act when participants experienced negative affect did not differ significantly from the neutral condition. These findings suggest that the temporal relationship between the conscious intention to act and the action itself is malleable to changes in affective states and may indicate that positive affect enhances intentional awareness
When people matter more than money: An ERPs study
No full textIn the present study, we showed that, in a social gambling task, individuals are influenced more by the type of social interaction than by the pattern of gains and losses. More precisely, the neural responses, as well as the level of pleasantness/unpleasantness following gains and losses, are modulated by social interaction factors. Here we present an Event-Related Potentials (ERPs) study in which three groups of participants were compared. Subjects were engaged in gambling tasks differing with regard to social factors: in a first condition, there was no social context; in a second condition, participants compared their outcomes with those of another individual; in a third condition, participants competed for a limited amount of money with another contender. In all conditions, all participants were revealed the outcome of an unselected alternative (non-obtained outcome) prior to the payoff associated with their selected option (obtained outcome). In addition, affective ratings were measured after the outcomes were presented. In the group without social context, ERPs results replicated previous findings. Interestingly, the P200 was modulated by varying social contexts, suggesting that attentive resources allocated to payoffs in comparison and competitive situations are decreased presumably in favor of social cues. Furthermore, Feedback Related Negativity (FRN) was predictive of the subjective feeling of pleasantness/unpleasantness following monetary outcomes. The present data provide information about neural and cognitive processing underlying economic decision-making when other individuals are involved
- …
