To Örebro University

Understanding active vision in naturalistic settings requires examining how the brain generates predictions and anticipates events under dynamic and uncertain conditions. The study of active vision necessitates investigation of coordinated interactions across distributed brain networks supporting diverse introspective and anticipatory functions while at the same time adopting a naturalistic stimulus design approach. Towards this, we propose a cognitive visuospatial complexity model that enables systematic parametrisation of dynamic visual stimuli for functional neuroimaging, behavioural experimentation, and psychophysics. The model supports controlled construction of graded complexity levels through parametric manipulation and is designed to investigate active vision and event‑based anticipation in ecologically valid scenarios. Our methodology defines an abstraction‑to‑realism axis capturing key prediction‑relevant dimensions of visuospatial complexity, including occlusions, contextual continuity, temporal regularity, event‑based anticipation, and constraints from commonsense and naïve physics. We additionally provide an accompanying stimulus dataset that demonstrates how the model can be implemented in practice for neuroimaging applications. The framework provides a standardised and reproducible stimulus‑design space, enhances cross‑study comparability, enables multimodal data integration, and supports the testing computational models of active vision and predictive processing. Our aim is to advance systematic methodological foundations for the neurocognitive and behavioural study of active vision under ecologically valid naturalistic conditions.

Studies in visual cognition highlight the importance of visual, spatial, and auditory cues in influencing human attention. Such cues often tend to be indicative of actions or events, thereby serving as predictive indicators in both passive observation as well as in interactive engagement. Our research focuses on visual attention in passive observation, particularly examining the manner in which visual, spatial, and auditory cues —henceforth visuoauditory (shorthand) cues— influence attention on everyday multimodal interaction. We systematically develop a visuoauditory event model for investigating visual attention in naturalistic embodied settings. Rooted in this event model, we explore the influence of five select visuoauditory cues —namely, speaking, gaze, relative motion, hand action, and visibility— on visual attention. Our analysis utilizes eye-tracking data from (90) participants observing (27) carefully designed naturalistic event scenarios and correlating their attentional metrics with the select visuoauditory cues in the backdrop of the developed event model. Findings reveal strong associations between attention and both intra-modal (irrespective of other cues) and cross-modal (combined with other cues) cueing effects, thereby highlighting the nuanced interplay amongst the cues influencing attentional patterns. We develop a systematic and generalized method for analyzing interactions and behavioral parameters, thereby characterizing the impact of visuoauditory cues on attentional dynamics. Our methodology, combined with the obtained insights into the attentional cueing effects, provides an analytical framework explicating the manner in which everyday (interactive) events directly drive attention under naturalistic conditions. This facilitates not only the precise modeling of behavior and attention allocation but also offers a high-level ‘experimental lens’ for examining interactions in relation to behavioral parameters. Taken together, our methodological and behavioral findings are well-positioned to benefit multiple fields, particularly by advancing human-centered design across diverse application domains. Lastly, towards promoting open-science and for wider dissemination, the complete experimental basis of this research —e.g., event-scenarios, high-quality annotated data, data-set supplementary— have been independently documented together with instructions to use and access experimental data.

Commonsense visual sensemaking requires robust mechanisms to extract scene elements and other visual features from perceived imagery, as well as rich conceptual commonsense knowledge and reasoning to interpret the perceived interactive dynamic. In this context, we position ongoing work on using integrated Vision Language Models (VLM) and Answer Set Programming (ASP) based reasoning about (dynamic) spatial configurations using VLMs as a neural foundation for symbolic modeling of dynamic scene structures. We sketch preliminary results highlighting practical usability by application to the task of Visual Question Answering (VQA) with the STRIDE-QA driving dataset.

The long-standing goal of creating a comprehensive, multi-purpose knowledge resource, reminiscent of the 1984 Cyc project, still persists in AI. Despite the success of knowledge resources like WordNet, ConceptNet, Wolfram|Alpha and other commercial knowledge graphs, verifiable, general-purpose, widely available sources of knowledge remain a critical deficiency in AI infrastructure. Large language models struggle due to knowledge gaps; robotic planning lacks necessary world knowledge; and the detection of factually false information relies heavily on human expertise. What kind of knowledge resource is most needed in AI today? How can modern technology shape its development and evaluation? A recent AAAI workshop gathered over 50 researchers to explore these questions. This paper synthesizes our findings and outlines a community-driven vision for a new knowledge infrastructure. In addition to leveraging contemporary advances in knowledge representation and reasoning, one promising idea is to build an open engineering framework to exploit knowledge modules effectively within the context of practical applications. Such a framework should include sets of conventions and social structures that are adopted by contributors.

We present a general framework for declaratively grounded visual commonsense (reasoning) about embodied interaction in naturalistic, in-the-wild settings relevant to a range of AI application domains. The core computational capabilities of the framework pertaining visual commonsense are driven by a robust neurosymbolic architecture primarily consisting of: (1) answer set programming based modelling of foundational aspects pertaining spatio-temporal dynamics, encompassing space, time, events, action, motion; (2) modularly integrated visual computing techniques constituting the neural substrate linking quantitative perceptual features serving as low-level counterparts to high-level semantic characterisations of (inter)active visual commonsense.

Practically, we also present a first open-release of the developed framework with the aim to promote independent extensions and real-world applied KRR. The release comprises: (a) demonstrated case-studies in domains such as autonomous driving, psychology and media studies; (b) systematic evaluation mechanisms for community benchmarking; and (c) supporting material such as tutorials and datasets.

Everyday driving relies on cognitive functions such as visual search, change detection and predictive attention for maintaining a high-level mental model of the world. Such functions are especially constrained by limited cognitive resources under complex conditions. Understanding the manner in which environmental complexity influences cognitive driving performance is key to improving road safety and developing human-factors guided driver assistance systems. Towards this, we explore the relationship between visuospatial and temporal complexity, and drivers’ situational awareness and decision-making. Particularly, we address two research questions: (I) How environmental complexity (visuospatial and temporal) impacts (in)attentional blindness; and (II) How does visual cueing impact anticipatory attention. In a naturalistic VR study, 150 participants drove in an immersive urban environment with varying complexity levels while responding to conditions such as overtaking, pedestrian crossings, occlusions, safety-criticality. The experiment, rooted to real-world driving data, incorporates a systematic cognitive model of environmental complexity consisting of dynamic visual, spatial, multimodal interaction features. Furthermore, we also analyse multimodal behavioural data, including gaze patterns, head movements, steering and braking. Results show that: (1) High visuospatial complexity increases inattentional blindness, with drivers compensating by prioritizing critical events; (2) Temporal complexity impairs detection performance, especially when events occur in rapid succession (1 sec) or when they last 5-10 seconds; and (3) Visual cueing improves anticipation, detection performance, and reaction time. These findings highlight the importance of adaptive strategies in overcoming cognitive limitations, providing insights for designing safety systems and training protocols for improving or assessing driver performance in everyday driving environments.

A key challenge in embodied, inter(active) vision is reasoning over alternative hypotheses about the dynamics of perceived objects and events, be it for real-time or even offline interpretation. Towards this, we address the problem of generating and ranking grounded visuospatial hypotheses based on a semantically encoded notion of hypothesis preference. Driven by probabilistic Answer Set Programming (ASP), we propose a general framework for modeling and reasoning about diverse preference types tailored to visuospatial interpretation tasks. The effectiveness of our probabilistic visuospatial hypotheses ranking method is demonstrated and evaluated with a community benchmark of Multi-Object Tracking (MOT17), where modeling uncertainty and preference is critical for robust scene interpretation. Furthermore, practical examples also showcase how semantically driven reasoning with preferences can be effectively used in real-world visual sensemaking tasks.

Situation awareness in driving involves detection of events and environmental changes. Failure in detection can be attributed to the density of these events in time, amongst other factors. In this research, we explore the effect of temporal proximity, and event duration in a change detection task during driving in VR. We replicate real-world interaction events in the streetscape and systematically manipulate temporal proximity among them. The results demonstrate that events occurring simultaneously deteriorate detection performance, while performance improves as the temporal gap increases. Moreover, attentional engagement to an event of 5-10 sec leads to compromised perception for the following event. We discuss the importance of naturalistic embodied perception studies for evaluating driving assistance and driver’s education.