Plasticity and neuromodulation of the extended recurrent visual network

The extended visual network, which includes occipital, temporal and parietal posterior cortices, is a system characterized by an intrinsic connectivity consisting of bidirectional projections. This network is composed of feedforward and feedback projections, some hierarchically arranged and others bypassing intermediate areas, allowing direct communication across early and late stages of processing. Notably, the early visual cortex (EVC) receives considerably more feedback and lateral inputs than feedforward thalamic afferents, placing it at the receiving end of a complex cortical processing cascade, rather than just being the entrance stage of cortical processing of retinal input. The critical role of back-projections to visual cortices has been related to perceptual awareness, amplification of neural activity in lower order areas and improvement of stimulus processing. Recently, significant results have shown behavioural evidence suggesting the importance of reentrant projections in the human visual system, and demonstrated the feasibility of inducing their reversible modulation through a transcranial magnetic stimulation (TMS) paradigm named cortico-cortical paired associative stimulation (ccPAS). Here, a novel research line for the study of recurrent connectivity and its plasticity in the perceptual domain was put forward. In the present thesis, we used ccPAS with the aim of empowering the synaptic efficacy, and thus the connectivity, between the nodes of the visuocognitive system to evaluate the impact on behaviour. We focused on driving plasticity in specific networks entailing the elaboration of relevant social features of human faces (Chapters I & II), alongside the investigation of targeted pathways of sensory decisions (Chapter III). This allowed us to characterize perceptual outcomes which endorse the prominent role of the EVC in visual awareness, fulfilled by the activity of back-projections originating from distributed functional nodes.

Distraction by task-irrelevant stimuli: the effects of endogenous spatial attention and visual working memory load

Salient stimuli, like sudden changes in the environment or emotional stimuli, generate a priority signal that captures attention even if they are task-irrelevant. However, to achieve goal-driven behavior, we need to ignore them and to avoid being distracted. It is generally agreed that top-down factors can help us to filter out distractors. A fundamental question is how and at which stage of processing the rejection of distractors is achieved. Two circumstances under which the allocation of attention to distractors is supposed to be prevented are represented by the case in which distractors occur at an unattended location (as determined by the deployment of endogenous spatial attention) and when the amount of visual working memory resources is reduced by an ongoing task. The present thesis is focused on the impact of these factors on three sources of distraction, namely auditory and visual onsets (Experiments 1 and 2, respectively) and pleasant scenes (Experiment 3). In the first two studies we recorded neural correlates of distractor processing (i.e., Event-Related Potentials), whereas in the last study we used interference effects on behavior (i.e., a slowing down of response times on a simultaneous task) to index distraction. Endogenous spatial attention reduced distraction by auditory stimuli and eliminated distraction by visual onsets. Differently, visual working memory load only affected the processing of visual onsets. Emotional interference persisted even when scenes occurred always at unattended locations and when visual working memory was loaded. Altogether, these findings indicate that the ability to detect the location of salient task-irrelevant sounds and identify the affective significance of natural scenes is preserved even when the amount of visual working memory resources is reduced by an ongoing task and when endogenous attention is elsewhere directed. However, these results also indicate that the processing of auditory and visual distractors is not entirely automatic.

Mind the gaps: cognitive-inspired AI for high-level visual sensemaking. Towards abstract concept image classification

The abundance of visual data and the push for robust AI are driving the need for automated visual sensemaking. Computer Vision (CV) faces growing demand for models that can discern not only what images "represent," but also what they "evoke." This is a demand for tools mimicking human perception at a high semantic level, categorizing images based on concepts like freedom, danger, or safety. However, automating this process is challenging due to entropy, scarcity, subjectivity, and ethical considerations. These challenges not only impact performance but also underscore the critical need for interoperability. This dissertation focuses on abstract concept-based (AC) image classification, guided by three technical principles: situated grounding, performance enhancement, and interpretability. We introduce ART-stract, a novel dataset of cultural images annotated with ACs, serving as the foundation for a series of experiments across four key domains: assessing the effectiveness of the end-to-end DL paradigm, exploring cognitive-inspired semantic intermediaries, incorporating cultural and commonsense aspects, and neuro-symbolic integration of sensory-perceptual data with cognitive-based knowledge. Our results demonstrate that integrating CV approaches with semantic technologies yields methods that surpass the current state of the art in AC image classification, outperforming the end-to-end deep vision paradigm. The results emphasize the role semantic technologies can play in developing both effective and interpretable systems, through the capturing, situating, and reasoning over knowledge related to visual data. Furthermore, this dissertation explores the complex interplay between technical and socio-technical factors. By merging technical expertise with an understanding of human and societal aspects, we advocate for responsible labeling and training practices in visual media. These insights and techniques not only advance efforts in CV and explainable artificial intelligence but also propel us toward an era of AI development that harmonizes technical prowess with deep awareness of its human and societal implications.