39339 research outputs found
Sort by
AI in the Ballot Box: Four Actions to Safeguard Election Integrity and Uphold Democracy
No full textTechnologies have long influenced elections, both positively and negatively, shaping their outcomes and the quality of public debate surrounding them. For example, the Internet enables citizens to organize more effectively than ever, empowering them to advocate for specific ideas and causes, but it is also a formidable channel for disinformation.
The rise of artificial intelligence (AI) presents significant new threats, including the multiplication of deepfakes, heightened cybersecurity risks, the emergence of manipulative persuasive agents, and the proliferation of synthetic data and fake accounts. At the same time, AI offers political actors a powerful tool to connect with voters, influence public opinion, and shape the flow of information. By tapping into existing trends in elections, AI has the potential to profoundly reshape the democratic process and influence election outcomes. Without proactive measures, however, AI could exacerbate worrisome trends such as political polarization and declining trust in democracy.
Governments must take decisive action regarding AI, particularly at a time when democracies around the world are facing increasing challenges and attacks on their elections. By acting on various fronts, they will shore up democratic systems, improve trust in society, and ensure that AI is leveraged responsibly to enhance the integrity of elections.This project was undertaken thanks to the contribution of the Fonds de recherche du Québec, CEIMIA, the Canada CIFAR Chair in AI and Human Rights at Mila and the University of Ottawa Research Chair in Technology and Society, and with the help of the Délégation du Québec à Rome and SIOI for the organization of the retreat
Design and Evaluation of Walking, Sit-To-Stand, and Stand-To-Sit Control Strategies for a Hip-Knee-Ankle-Foot Prosthesis with Motorized Hip Joint
No full textHip disarticulation (HD) amputation involves the removal of the entire lower limb and the hip joint, adversely affecting mobility and quality of life. Depending on their physical condition and life goals, some people with amputation are prescribed a hip-knee-ankle-foot (HKAF) prostheses to regain mobility. However, HKAF prostheses are known to have a high rejection rate compared to transfemoral and transtibial prostheses, primarily due to their excessive energy demands and the physical fitness required for effective use. Despite advancements in motorized prosthetic joints for the knee and ankle, innovation for HKAF prostheses has stagnated. This thesis addresses this gap by developing and evaluating adaptive control strategies for a motorized HKAF prosthesis to enhance mobility for HD amputees.
Based on preliminary mechanical development of the first viable motorized hip joint (Power Hip), this thesis developed and refined the electronics, sensors, and control system to enable people with hip level amputations to walk, sit, and stand. A prototype Powered Hip prosthesis was tested against a conventional passive prosthesis (Otto Bock Helix hip, C-Leg knee, Terion K2 foot) in a single HD participant. The Theia Markerless motion analysis system and Visual3D were used for kinematic and kinetic analyses. During walking, the Power Hip reduced pelvic tilt range from 22.77° ± 5.76° to 6.72° ± 1.49°, minimizing compensatory pelvic movements. Hip extension range improved from -0.22° ± 0.77° to -7.04° ± 2.85°, enabling a more natural stride by stabilizing the hip throughout the gait cycle. During sit-to-stand, ground reaction forces (GRF) on the prosthetic side increased from 0.30 ± 0.67 N/kg to 2.69 ± 0.34 N/kg, while stand-to-sit GRF rose from 4.28 ± 1.00 N/kg to 5.37 ± 0.52 N/kg. These enhancements improved load distribution, reducing intact-limb forces and aligning kinetic profiles more closely with transfemoral amputee patterns.
By achieving movement biomechanics comparable to transfemoral prosthesis users, this research reimagined what HKAF prostheses can achieve. This research lays the foundation for a new generation of user-friendly prosthesis that prioritize mobility, independence, and quality of life for people with hip-level amputations
Scene and Graph Domain Adaptation to Tackle Domain Shifts in Real-world Image Semantic Segmentation and Human Activity Recognition
No full textDeep learning has transformed computer vision by automating feature extraction and representation learning across multiple data modalities, surpassing traditional paradigms that relied on manual feature engineering. However, deploying deep learning models in real-world applications often encounters significant challenges in generalization due to domain shifts between the training dataset (source domain) and the deployment environment (target domain). When the source domain distribution is misaligned or represents only a subset of the target domain, the deep learning model’s performance tends to degrade in the target domain, limiting its scalability in many real-world applications.
The research in this thesis addresses the domain shifts challenge with the development of novel Domain Adaptation (DA) frameworks. Particularly, DA formulates adaptive models that can be trained on the source domain and adapted to the target domain where data is scarce. Moreover, unlike previous DA research that focuses on a single type of computer vision application, this thesis expands the research on individual DA frameworks for two different applications, that is, RGB image semantic segmentation and skeleton video human action analysis respectively, which can be detailed as follows.
1) The first proposed DA framework, named Enhanced Scene Domain Adaptation (ESDA), focuses on mitigating scene-based domain shifts for the task of semantic segmentation on street images. Street scene understanding is critical for autonomous driving and robotic navigation, but outdoor scenes often exhibit scene-based visual discrepancies caused by variations in lighting, weather, and regional characteristics. Therefore, ESDA introduces three innovative DA mechanisms: pixel-wise adversarial adaptation, prototypical knowledge adaptation, and a target-specific adaptation classifier, each learning a model trained on image samples from one region (e.g., a city) to achieve robust performance on samples from other regions (e.g., alternative cities) for semantic segmentation.
2) The second proposed framework, named Enhanced Graph Domain Adaptation (EGDA), addresses graph domain shifts in skeleton-video based human activity analysis. The latter has been widely utilized to address a variety of real-world video surveillance tasks, where human actions are represented by the trajectories of skeletal joints captured by acquisition systems. By proposing four novel methods, cross-view adaptation, cross-sensor adaptation, cross-sequence adaptation, and cross-permutation adaptation, EGDA effectively encourages the deep learning model to aggregate domain-invariant action dynamics from skeletal joint trajectories while dealing with the domain shift arising from varying camera perspectives, sensor configurations, and sequence lengths or activity ordering.
While evaluating the ESDA and EGDA frameworks, the research utilizes existing large-scale benchmarks to mimic various domain shift situations while creating pairs of a source domain and a target domain. For instance, in semantic segmentation, ESDA utilizes the dataset composed of synthetic street images as the source domain and evaluates the model on a real-world street image dataset. In human action analysis, EGDA leverages several large-scale skeleton datasets that are collected in different environments to mimic the domain shifts in skeleton data. Experimental results demonstrate that the proposed frameworks are effective for alleviating both scene and graph domain shifts from different data modalities and successfully improve the adaptability of the baseline deep learning models for image semantic segmentation and skeleton action analysis tasks.
The original contributions of this thesis include a comprehensive investigation of two types of domain shifts commonly encountered in computer vision: scene domain shifts in RGB image semantic segmentation and graph domain shifts in skeleton-based video human action recognition. Novel methodologies are introduced to enhance domain adaptation performance in these computer vision applications. The proposed frameworks enable adaptive deep learning models to be trained on large-scale benchmarks while easily adapting to real-world target domains where labeled data is scarce or unavailable. The original domain adaptation approaches are designed to enhance the base networks without additional computational complexity, ensuring efficient operation on the real-world target domain
Radiator Designs and Experimental Platforms for Near-Field Radiative Heat Transfer
No full textNear-field radiative heat transfer (NFRHT) consists of evanescent electromagnetic coupling occurring between two bodies at sub-wavelength distances, allowing to increase the radiative exchange beyond conventional laws of thermal radiation. NFRHT has demonstrated great potential for applications such as energy conversion and heat transfer control. In particular, multiple works predict that performances of near-field thermophotovoltaic (NFTPV) modules, in which a hot radiator is positioned at sub-wavelength distances from a cold photovoltaic (PV) cell, could significantly surpass those of current solid-state heat-to-electricity conversion technologies.
Despite its tremendous potential, experimental progress on NFRHT and NFTPV has been slow due to technical challenges and gaps in knowledge, three of which are at the core of this thesis. Firstly, only a limited number (i.e., fewer than 30) of NFRHT experimental platforms have been reported worldwide. The field therefore faces limited capabilities in characterizing novel materials for NFRHT. Secondly, NFTPV technology commands highly specialized PV cells. Unfortunately, most reported NFTPV platforms relied on basic PV cells fabricated in-house, or on commercially available photodetectors, resulting in modest performances. This leaves many promising PV materials, e.g., InAs, largely unexplored. Thirdly, theoretical research on new radiator materials for NFTPV applications has predominantly focused on one class of materials (i.e., plasmonic materials) that are difficult to experimentally investigate as their optical properties are contingent upon factors such as film annealing and deposition conditions. Therefore, the most suitable radiator material for NFTPV remains ambiguous.
To address these three challenges, our first objective is to develop experimental methods for NFRHT, our second objective is to achieve NFTPV measurements using an optimal bandgap (i.e., InAs-based) PV cell, and our third objective is to theoretically design an optimized radiator for an InAs-based NFTPV system.
Within the scope of our first objective, we experimentally demonstrate the potential of using nanomechanical resonators as a temperature sensor for NFRHT measurements. Our approach offers a high-precision flexible NFRHT platform that could facilitate the testing of many new interesting materials. We report measurements in the deep sub-wavelength regime (i.e., regime dominated by surface resonance coupling) without the need for custom-fabricated micro-devices. Moreover, we demonstrate that the use of nanomechanical resonators could allow fundamental advances on simultaneous measurement of NFRHT and Casimir forces. If we can distinguish the contributions from Casimir and NFRHT effects, nanomechanical resonators could provide a path to experimentally demonstrate the correction to the Casimir forces out of thermal equilibrium.
For our second objective, we aim to conduct experimental research on the NFTPV effect using InAs-based PV cells. These cells, optimized for near-field thermal radiation, were custom-fabricated by PV cell experts (Prof. Karin Hinzer group, uOttawa, and Prof. Zbig Wasilewski group, University of Waterloo) in close collaboration with our group during this work. Unfortunately, the geometry of our radiator significantly limited the capabilities of our experimental platform, preventing NFTPV measurements. In fact, the large thermal mass radiator employed in our initial approach precluded high-temperature scans and greatly complicated the alignment process. In the future, we, therefore, recommend a revised approach relying on a localized heating element, enabling vibrational contact detection.
Finally, regarding our third objective, using a one-dimensional fluctuational electrodynamics model, we theoretically explore new radiator materials for NFTPV, focusing specifically on reproducible crystalline materials. More precisely, we investigate spectral electromagnetic coupling of a near-field thermal radiator with a PV cell in order to enhance spectral efficiency and total output power. We find that when the radiator and PV cell are both made of InAs, nearly a threefold improvement of spectral efficiency is possible compared to a silicon radiator with the same InAs cell. This enhancement reduces subgap thermal transfer while maintaining power output. Through this work, we also uncover an overestimation of free carrier absorption in InAs dielectric function models. We propose a corrective model and demonstrate that it accurately represents the absorption at moderate doping levels but could be further refined for improved accuracy at higher doping levels
Medication management for older adults in interprofessional primary care teams: a qualitative interview study of family health teams in Ontario, Canada
No full textAbstract Background Team-based, interprofessional primary care models are arguably well positioned to care for patients with polypharmacy as they often have a pharmacist or allied health professionals to support patients with medication management. However, little is known about how teams work together to manage medications. This study aimed to explore how a team-based primary care organization including a mix of physicians and interdisciplinary health providers (IHPs), called Family Health Teams (FHTs), manage medications for older adults. Methods We conducted semi-structured interviews (n = 38) with administrators, family physicians, and IHPs from six FHTs in Ontario, Canada. We followed the thematic analysis steps outlined by Braun and Clarke and adapted the approach to use a codebook. Results Four themes were identified: (1) strategic goals and internal policies; (2) tailored programs and supports; (3) diverse team configurations and roles; and (4) teamwork and collaboration. Findings revealed variation in the ways physicians and IHPs worked together to manage medications for older adults and that different approaches to care and physician communication preferences were identified as challenges to medication management. Trust was an important factor in medication management among teams; the more physicians interacted with IHPs, the more comfortable and trusting they were in giving them an active role in patient care. Regardless of the approach to medication management, participants agreed that physicians ultimately had the final say in patient care. Conclusions Despite an emphasis on teamwork in FHTs, there were few examples of true collaboration and shared care for medication management. To support older adults and others with complex health needs, opportunities to improve teamwork, strengthen collaboration, and optimize team composition should be identified and pursued
The Impact of a Mediterranean-Based Diet on Cognitive, Inflammatory, and Neurotrophic Impairments Induced by a Chronic Social Defeat Stressor in Male C57BL/6N Mice
No full textBackground: Stress-related neuropsychiatric disorders have overlapping impairments in social and non-social cognition. Limited efficacy of treatments targeting these symptoms and the continued increase in prevalence in these disorders has necessitated alternative strategies. The Mediterranean (Med) diet has been shown to improve depression, anxiety, and cognitive deficits in clinical studies. Using a mouse model of chronic social stress, this study investigated the potential of a mouse-adjusted Med-based dietary intervention to mitigate social and non-social cognitive impairments and limit changes in brain neurotrophic and inflammatory factors.
Methods: Male C57BL/6N mice were randomly assigned to a Control or a Med-based diet. After a 14-day acclimatization period to the diets, both groups were either subjected to 10 consecutive days of chronic social defeat stress (CSDS) or to a no stressor control condition. Cognitive tests were conducted 24 hours after the last stressor or control session. The ventral hippocampus was collected 24 hours following the last cognitive test and analyzed for the mRNA expression of pro-inflammatory cytokines, microglial markers, and neurotrophic factors.
Results: The CSDS regimen increased social avoidance behaviours and altered the hippocampal expression of neurotrophin-3 and Tropomyosin receptor kinase B. In CSDS mice, the Med-based diet improved long-term memory and reduced hippocampal tumor necrosis factor alpha but promoted social avoidance behaviours, impaired spatial reference memory, and decreased hippocampal brain-derived neurotrophic factor.
Conclusion: Dietary interventions in male mice may have differential effects on cognitive and hippocampal health in the context of chronic social stress, requiring further investigation into its use as an adjunctive therapy for cognitive deficits in stress-related neuropsychiatric disorders
Réseaux socionumériques et pratiques informationnelles des journalistes haïtien.ne.s de la zone métropolitaine: affaire petrocaribe (2018-2020)
No full textNotre travail porte sur la réception des contenus des réseaux socionumériques et leur influence sur les médias conventionnels dans la lutte contre la corruption, une pratique mondiale émergente dont Haïti a fait l'expérience dans le cadre de l'affaire petrocaribe de 2018 à 2020.
Comme objectif principal, nous cherchons à comprendre l'interprétation que faisaient les journalistes haïtien.ne.s des publications sur la corruption, postées dans les réseaux socionumériques dans le contexte du mouvement social #petrocaribechallenge et comment ces journalistes les intégraient dans leur travail au quotidien.
Nous avons retenu la théorie du sense-making de l'américaine Brenda Dervin pour répondre à notre question générale de recherche qui est la suivante : Dans quelle mesure les contenus en lien avec la corruption véhiculés dans les réseaux socionumériques ont-ils influencé les pratiques informationnelles des journalistes de la zone métropolitaine entre 2018 et 2020 ?
Nous avons utilisé l'entrevue semi-dirigée pour collecter des données qualitatives auprès de 11 journalistes professionnel.le.s de la zone métropolitaine que nous avons subdivisé en deux catégories : journaliste « confirmé.e » et journaliste « junior ». Tous ont travaillé sur l'affaire petrocaribe. Nous avons analysé les verbatim des entretiens.
Nos résultats montrent que les journalistes ont interprété et utilisé différemment les contenus sur la corruption publiés sur les pages Facebook et comptes Twitter (actuellement X) des « petrochallengers » et de la Cour supérieure des comptes et du contentieux administratif (CSC-CA). L'appropriation et l'intégration de ces publications dans leur rendement professionnel étaient en partie déterminées par le nombre d'années d'expérience qu'ils avaient dans le métier
Distributed Representations of Topics
No full textIn an era where around 330 million terabytes of data are generated each day, it is crucial to have effective methods for extracting knowledge and structure from this vast amount of information. Topic modeling is a technique for extracting themes, topics, and structure within large data sets which allow for organizing, searching and making sense of the data efficiently. It is a fundamental technique that has a lot of downstream uses in information retrieval, recommender systems, content summarization, content tagging, trend detection, and many others. Some major challenges of topic modeling are finding the right resolution of topics, labeling the topics, segmenting text by topics, evaluating topic model performance, and dealing with topic change over time. The most widely used methods for topic modeling are Latent Dirichlet Allocation and Probabilistic Latent Semantic Analysis. They are probabilistic generative models and, despite their popularity, they have several weaknesses. In order to achieve optimal results, they often require the number of topics to be known. They need custom stop-word lists, stemming, and lemmatization. Lastly, they model topics as distributions over a vocabulary which necessarily make uninformative words the most probable in a topic. Modern neural topic modeling approaches have tackled some of these problems, but none have been able to solve all of them. We introduce distributed representations of topics where topics are vectors in a semantic vector space. We redefine topics to be the most informationally representative of documents rather than representing an underlying distribution over a vocabulary. Our novel topic modeling approach uses document contextual token embeddings. It creates hierarchical topics, finds topic spans within documents, and labels topics with phrases rather than just words. We propose a density-based agglomerative clustering for semantic vector spaces, which is essential for topic hierarchies. Most previous topic modeling evaluation methods focus on topic coherence without evaluating how well topics represent the documents specifically assigned to a topic, leaving a gap in topic model evaluation. To close this gap, we propose the use of BERTScore and topic information gain to evaluate topic coherence and to evaluate how informative topics are of the underlying documents in addition to the existing topic coherence measures
Phenotypic and Functional Characterization of Surgery Induced Myeloid Derived Suppressor Cells
No full textRemoving the bulk of tumor burden with surgery is critical for the recovery of patients with solid malignancies. Surgery, however, induces many physiological changes which impair both NK cell cytotoxicity and cytokine secretion giving circulating tumor cells an opportunity to escape and form distant metastases. The rise of MDSCs in the postoperative landscape has been identified as a major contributor of postoperative NK cell dysfunction. Lack of characterization of both SxMDSC phenotype and suppressive mechanisms are a major challenge in targeting these cells to improve long-term outcomes for patients. This work gives a previously unseen detailed phenotypic description of SxMDSCs, showing a phenotypic switch toward an “M2 like” phenotype after tumor resection. Furthermore, preliminary data from this study identifies a contact-dependent mechanism for SxMDSC suppression of NK cell cytotoxicity and indicates they are likely not a major contributor of suppression NK cell cytokine or cytotoxic granule secretion
The Driving Factors of Canada's Foreign Policy and History Concerning Israel and Palestine
No full textOne of the most prominent conflicts in the modern world is that regarding Israel and Palestine and Canada's role within this conflict has been apparent before the creation of the Israeli state. Literature has researched and demonstrated the United States' role in this conflict however, minimal literature has been produced for Canada's foreign policy role and attitudes towards Israel and Palestine. With the use of secondary literature, this thesis uses qualitative data to display four driving factors of Canadian foreign policy in Israel and Palestine. It also presents several Canadian Prime Ministers to demonstrate the various policies, actions, and attitudes towards notable Israeli-Palestinian events within their governments, and the driving factors that were used to implement these policies. This thesis will demonstrate that Canada's stated interests and values differ from their implied interests and values, as the first favours human security and peacekeeping initiatives, while the latter is rooted in racism and Western supremacy