In the recent years, a huge success has been accomplished in prediction of human eye fixations. Several studies employed deep learning to achieve high accuracy of prediction of human eye fixations. These studies rely on pre-trained deep learning for object classification. They exploit deep learning either as a transfer-learning problem, or the weights of the pre-trained network as the initialization to learn a saliency model. The utilization of such pre-trained neural networks is due to the relatively small datasets of human fixations available to train a deep learning model. Another relatively less prioritized problem is amount of computation of such deep learning models requires expensive hardware. In this dissertation, two approaches are proposed to tackle abovementioned problems. The first approach, codenamed DeepFeat, incorporates the deep features of convolutional neural networks pre-trained for object and scene classifications. This approach is the first approach that uses deep features without further learning. Performance of the DeepFeat model is extensively evaluated over a variety of datasets using a variety of implementations. The second approach is a deep learning saliency model, codenamed ClassNet. Two main differences separate the ClassNet from other deep learning saliency models. The ClassNet model is the only deep learning saliency model that learns its weights from scratch. In addition, the ClassNet saliency model treats prediction of human fixation as a classification problem, while other deep learning saliency models treat the human fixation prediction as a regression problem or as a classification of a regression problem.

No description available.

Problemformulering och syfte: Tidigare forskning visar att media och dess innehåll har stor påverkan på hur samhället ser på kvinnor och skapar en förväntan på hur kvinnor bör vara utifrån stereotyper och normer. Om inte kvinnor representeras på magasinomslag skapas klyftor och ojämlikhet. Syftet är att undersöka hur kvinnor representeras visuellt på ett urval av Rolling Stones magasinsomslag. Metod: Kvantitativ innehållsanalys användes för att undersöka andelen Enbart kvinna/kvinnor gentemot andelen Enbart man/män, Både män och kvinnor och Övrigt på omslagen av Rolling Stone mellan 2010–2020. Semiotisk bildanalys användes för att undersöka tio omslag mellan åren 2010–2020. Dessa valdes genom att ta det första numret varje år med Enbart kvinna/kvinnor på omslaget. Resultat: Den kvantitativa innehållsanalysen visar att andelen omslag med Enbart kvinna/kvinnor ökade procentuellt från 2017. År 2016 fanns inget omslag med Enbart kvinna/kvinnor. Den semiotiska bildanalysen kartlägger de olika visuella elementen och hur dessa bidrar till hur kvinnor representeras visuellt samt att detta i majoriteten av omslagen bidrar till stereotyp representation.

Our ability to be active agents in the world depends on our cognitive system to collect complex multisensory information, i.e. information coming from different senses, and integrate it appropriately. One fundamental topic of interest in the study of cognition is to understand the consequences of deafness on the organization of brain functions, specifically when one sensory modality is either lost or the information coming from that sensory modality is limited. In my work I used the spatial cueing paradigm to study how visual attention and selection is affected by diverse grades of congenital or acquired deafness in different life stages. The goal of the first study was to validate an integrated approach of covert and overt orienting to study social and non-social cueing of attention in hearing adults. Specifically, I examined manual and oculomotor performance of hearing observers performing a peripheral discrimination task with uninformative social (gaze cue) and non-social cues (arrow cue). In Experiment 1 the discrimination task was easy and eye movements were not necessary, whereas in Experiment 2 they were instrumental in identifying the target. Validity effects on manual response time (RT) were similar for the two cues in Experiment 1 and in Experiment 2, though in the presence of eye movements, observers were overall slower to respond to the arrow cue compared to the gaze cue. Cue-direction had an effect on saccadic performance before the discrimination was presented and throughout the duration of the trial. Furthermore, I found evidence of a distinct impact of the type of cue on diverse oculomotor components. While saccade latencies were affected by whether the cue was social or not, saccade landing positions were not affected by cue-type. Critically, the manual validity effect was predicted by the landing position of the initial eye movement. This work suggests that the relationship between eye movements and attention is not straightforward. In hearing adults, in the presence of eye movements, saccade latency was related to the overall speed of manual response, while eye movements landing position was closely related to manual performance in response to the validity of the cues. In the second study, I used the above-mentioned approach to investigate the impact of early profound deafness on the oculomotor control and orienting of attention to social and non-social cues. Previous research on covert orienting to the periphery suggests that early deaf adults are less sensitive to uninformative gaze cues, though were equally or more affected by non-social arrow cues. The aim of this second study was to investigate whether spontaneous eye movement behavior helps explain the reduced contribution of this social cue in deaf adults. Twenty-five deaf and twenty-five age-matched hearing observers took part in the experiment. In both groups, the cueing effect on RT was comparable for the gaze- and arrow-cue, although deaf observers responded significantly slower than hearing controls. While deaf and hearing observers responded equally to the cue presented in isolation, deaf participants relied significantly more on eye movements than hearing controls once the discrimination target was presented. Saccade landing position in the deaf group was affected by validity but not by cue type while latency was not modulated by these factors. Saccade landing position was also strongly related to the magnitude of the validity effect on RT, such that the greater the difference in saccade landing position between invalid and valid trials, the greater the difference in manual RT between invalid and valid trials. This work suggests that the contribution of overt selection in central cueing of attention is more prominent in deaf adults and determines the manual performance. The increase in eye movements and overall slower responses in deaf observers may act as an adaptive strategy to balance the need for accuracy in a context where vision and visual attention are used to monitor the surrounding environment in the absence of auditory input. This tendency to emphasize accuracy of response at the cost of responding more slowly seems to allow them to maintain the same level of cue-driven performance as their hearing peers. In the third study I focused on partial hearing loss. Little is known on the consequences of pure presbycusis, which is usually associated with aging (Age-related Hearing Loss, ARHL). In this case, auditory information is still present, although linked to an amount of uncertainty regarding its usefulness. In this study I started to investigate the role of ARHL on cognition considering covert orienting of attention, selective attention and executive control. I compared older adults with and without mild to moderate presbycusis (26-60 dB) performing 1) a spatial cueing task with uninformative central cues (social vs. non-social cues), 2) a flanker task and 3) a neuropsychological assessment of attention. Notably, while hearing impaired individuals responded as equally fast as their normally hearing peers, they were characterized by reduced validity effects on spatial cueing of attention, though no additional group differences were found between the impact of social and non-social cues. Hearing impaired individuals also demonstrated diminished performance on the Montreal Cognitive Assessment (MoCA) and on tasks requiring divided attention and flexibility. Conversely, overall response times and flanker interference effects were comparable across groups. This work indicates that while response speed and response inhibition appear to be preserved following mild to moderate presbycusis, orienting of attention, divided attention and the ability to flexibly allocate attention, are more deteriorated in older adults with ARHL. These findings suggest that presbycusis might exacerbate the detrimental influences of aging on visual attention. Taken together, the findings of my research project highlight the different role hearing loss may play at different life stages. On the one hand, congenital and early deafness seems to induce cognitive and behavioral compensations, which may encompass oculomotor behavior as well; these changes occur progressively during development and may reflect experience-dependent plasticity. On the other hand, late-life compensations in vision and visual attention in older adults with presbycusis may not take place or do not effectively reduce the negative impact of the auditory impairment. Rather, my data suggest that in this population a deficit in audition may consequently lead to a deficit in visual attention. Future lines of research can aim to better characterize other aspects of attention in the aging population with presbycusis, e.g. peripheral visual attention and the relationship between covert and overt attention. Finally, future research may also consider intervention through early diagnosis and treatment by means of hearing aids, which can be beneficial to cognitive functions and might delay or even prevent cognitive decline in this population, in which sensory compensation may not be sufficient.

This study investigates the representations of librarians in movies spanning from the years 2012 to 2022 from a gender perspective using thematic analysis. Being a female-dominated profession with an increasing use of technology, the goal is to see how librarians are being portrayed today and how that might affect the perceptions of librarianship. The revealed themes address: (a) librarians being portrayed as feminine, (b) librarians being portrayed as sexual objects to please the male audience, (c) libraries as a place for knowledge, (d) librarian profession being chosen out of necessity, and (e) bias toward traditional librarianship. This study indicates a change in portrayals of librarians in movies, but the librarian characters are nevertheless being depicted in a negative light. In order to improve the perceptions of librarians, this study suggests that a more diverse representation of librarians and their tasks is needed.

No description available.

No description available.

No description available.

The objective of this paper is to develop a theory for analysing the practice of representing the world with visualizations in the form of different images and the scientific gaze that renders these images meaningful. These images are often combined with text, since the natural sciences are lexivisual practices. For natural scientists these visual and lexivisual representations appear totally unproblematic. Scientific representations can be many kinds of pictures, graphs, tables or diagrams and are seen as vehicles of knowledge; in their concrete and tangible form we call them images but they can also be displayed, screened and projected. I will not directly deal with mental images or conceptions of the world, only as one way to understand what renders the tangible images meaningful. But of course, the conflict between a culturalistic and a natural scientific conception of the world runs through this paper, as the empirical material is natural scientific and the analysis is culturalistic.

A lack of alarm management for industrial control rooms has led to frequent alarm floods that have the potential to overwhelm operators within minutes. One approach to managing alarm floods would be altering the salience of alarms that operators might already notice, thereby reducing the disruption on workflow and attention for managing uninformative alarms. This research investigated the central hypothesis that eye fixations could supply passive input to acknowledge alarms anticipated by the operators and thereby improve their overall task performance. A dual-task experiment recruiting 24 participants was conducted to compare three gaze-based alarm acknowledgement methods –Proximity, Prediction, and Entropy- against no acknowledgement across three types of scenarios – Near-threshold, Trending, and Fluctuation. The gaze-based acknowledgement methods reduced visual and auditory salience of alarms as a function of the number of fixations on parameters as well as characteristics of the parameter known to influence operator monitoring behaviors. The participants performed an alarm monitoring task while controlling a continuous parameter within an acceptable range. While participants showed a preference for all of three gaze-based acknowledgment methods, performance of the parameter control task did not improve with gaze-based acknowledgement. Scenario types, as defined by the behavior of the parameters, exhibited a significant effect on the performance of the parameter control task, suggesting a greater influence on participant attention than the reduced salience associated with the gaze-based acknowledgments. Additional analysis revealed that gaze-acknowledgements are higher in scenarios with the most suitable for the gaze-based acknowledgement methods, although the participants did not show any gaze-based acknowledgements and did not make a prediction of an alarm for a significant portion of the trials, suggesting a lack of resource allocation to the alarm monitoring task. This result suggests that the effectiveness of gaze-based acknowledgement may depend on the combination of on-going tasks. Taken together, the experimental results showed some utility of user gaze in managing alarms given how acknowledgement occurred more often when the acknowledgement methods and parameters matched; however, further design research is necessary to translate the utility into clear performance or productivity benefits. / Master of Science / Industrial control rooms are notorious for having too many alarms triggered within minutes and operators are hindered by responding to these alarms as opposed to the actual process faults. Existing alarm management research and applications have already reduced nuisance alarms by filtering out those correlated to one another according to historical data or plant models. However, existing approaches have not eliminated the process parameters that operators already expect to reach alarm thresholds. In other words, current alarm management has not adapted for operator awareness of impending alarms. This study explored how eye-tracking might be used to acknowledge alarms anticipated by operators, thereby reducing uninformative alarms and interruption to operator work. The participants performed an alarm monitoring task while trying to maintain a fluctuating parameter within an acceptable range. While participants liked the gaze-based acknowledgement methods, their performance on the parameter control task did not improve over conditions without any alarm acknowledgement. The alarm monitoring task may not have received sufficient attention to induce an observable benefit. The characteristics of the parameter seemed to have a larger effect on participants' attention than the muted alarm presentation associated with the gaze-based acknowledgment. Further research is necessary to refine the current design to induce the postulated attention and performance benefits with gaze-based acknowledgement.