Syracuse University
Syracuse University Research Facility and Collaborative EnvironmentNot a member yet
20376 research outputs found
Sort by
Investigating the Topic, Impact, and Resource Gap Between Academia and Industry in the Field of Artificial Intelligence
No full textArtificial intelligence (AI) has seen fast development in industry and academia. Striking advancements from industry companies have stunned the research field, inviting a fresh perspective on the relationship between industry and academia. Industry research teams have played a critical role by developing large-scale computational infrastructure, curating extensive datasets, and publishing impactful AI research. As a result, industry-led AI research has impacted both fundamental and applied AI, blurring the distinction between academic and industrial research. Concerns have arisen regarding the growing resource gap between industry research teams and academia research teams. While some AI research topics remain accessible, research in areas such as large language models (LLMs) necessitate more resources such as computational power and data access: resources largely concentrated among industry companies and a few top universities. This disparity raises critical questions about transparency, replicability, and inclusiveness in AI research, particularly regarding the role of academia in shaping the landscape of AI research. This dissertation addresses three key research questions: (1) What are the differences between industry and academia AI research in terms of impact and novelty? (2) How do research topics differ between resource-limited institutions and resource-rich institutions? (3) Has research from resource-rich institutions lowered or heightened the barriers for resource-limited researchers in AI? To explore these questions, three studies were conducted. Study 1 quantitatively examines the differences in impact, novelty, disruptiveness, and state-of-the-art status between industry and academic AI research over the last 25 years. The study reveals that articles published by teams consisting exclusively of industry researchers tend to get more attention, with a higher chance of being highly cited, and more likely to produce state-of-the-art models. In contrast, academia teams publish the bulk of AI research and tend to produce more atypical and citation-disruptive work. The respective impact-novelty advantages of industry and academia are robust to controls for subfield, team size, seniority, and prestige. We find that academia-industry collaborations produce the most impactful work overall but do not have the novelty and citation-disruptive level of academia teams. Study 2 surveys existing methodologies in citation context analysis to support Study 3. Study 3 investigates how research topics in AI have evolved and whether the resource-intensive nature of AI research has constrained topic and methodology selection for resource-limited researchers. Our findings reveal that resource-rich institutions are more likely to engage in research areas that are gaining popularity, while resource-limited institutions tend to work on topics that are declining in prominence. This suggests that resource constraints influence the feasibility of pursuing certain research directions. Moreover, by analyzing citation intent, our result indicates that despite significant industry-led advancements, research from well-funded teams has not necessarily lowered the barrier to entry but has, in some cases, made it more challenging for resource-limited researchers to contribute. The findings contribute to a broader understanding of the relationship between academia and industry, the resource gap, and the shifting landscape in AI research. We identify the unique and nearly irreplaceable characteristics that academia and industry have for the healthy progress of the academic field of AI. We provide empirical evidence on how resource-demanding research is shaping the field. These insights will support policymakers, academic institutions, and industry stakeholders in fostering a more equitable and transparent AI research ecosystem
Online and Offline Learning for Embodied AI in Autonomous Systems
No full textEmbodied Artificial Intelligence (AI), which integrates physical embodiment with intelligent decision-making, is increasingly critical in advancing autonomous systems across diverse domains such as autonomous driving and robotic manipulation. This dissertation presents a comprehensive exploration of online and offline learning approaches for Embodied AI in autonomous systems, addressing both algorithmic innovations and dataset construction to overcome fundamental challenges in perception, decision-making, and control. Through five interconnected studies, we systematically advance the state of the art in deep reinforcement learning (DRL) and imitation learning for embodied control. First, we introduce CADRE, a cascade online DRL framework for vision-based autonomous urban driving that strategically decomposes complex driving tasks into perception and control subtasks. By pre-training a perception module that leverages the attention mechanisms to build inter-relationships between visual and control information, and employing distributed Proximal Policy Optimization with careful reward shaping and LSTM sequential modeling, CADRE achieves high success rates in challenging urban environments with dense traffic on the NoCrash benchmark. Second, we propose Adaptive Conservative Level in Q-Learning (ACL-QL), a flexible framework for offline DRL that enables fine-grained control of Q-function conservatism. The ACL-QL framework introduces two adaptive weight functions corresponding to out-of-distribution actions and dataset actions to dynamically shape the Q-function. Supported by detailed theoretical analysis, we implement neural networks as weight functions and construct surrogate and monotonicity losses to maintain better performance. Comprehensive experiments on standard offline DRL benchmarks demonstrate ACL-QL\u27s state-of-the-art performance and versatility across diverse scenarios. Third, we develop a Discrete Policy approach for multi-task imitation learning with language instruction, which learns action patterns in the latent space to better disentangle feature representations across different skills. This innovative strategy enables more effective transfer learning and skill composition in robotic systems. Through extensive simulations and real-world experiments, our approach demonstrates superior performance in multi-task settings compared to various state-of-the-art methods, offering a compelling new perspective on learning multi-task policies for embodied control. Fourth, we address the challenge of learning from imperfect demonstrations through a Self-Supervised Data Filtering (SSDF) framework. By calculating accurate quality scores using pre-trained transformers and performing weighted behavior cloning on high-quality imperfect demonstrations, SSDF significantly improves policy learning without requiring reward information or online exploration. The extensive experimental results in both simulation and real-world applications confirm SSDF\u27s ability to accurately select high-quality demonstrations from imperfect datasets, substantially boosting final performance. Finally, we introduce RoboMIND, a large-scale, multi-embodiment dataset for robot manipulation. This comprehensive resource includes four distinct embodiments with high-quality demonstrations across multiple tasks, objects, and skills, collected through an intelligent data platform with rigorous quality assurance. Our quantitative analyses highlight RoboMIND\u27s heterogeneous embodiments, diverse episode lengths, broad task coverage, and wide range of objects from domestic, industrial, kitchen, office, and retail scenarios. Experiments with popular imitation learning robot models reveal opportunities for improving accurate positioning and precise control. Together, these contributions form a cohesive exploration of learning approaches for embodied intelligence, spanning online and offline paradigms, with applications in autonomous driving and robotic manipulation. Our work establishes new methodologies and resources that address fundamental challenges in developing autonomous systems capable of robust perception, decision-making, and control in complex real-world environments
Social Media Fragility Indicators: A framework for evaluating social media as fragile states
No full textSocial Media Fragility Indicators: A framework for evaluating social media as fragile states
Presented at the Media Sociology Symposium, August 7th 2025
In the uncharted domain of cyberspace, social media platforms have stepped in and stepped up as state-like entities – social states – enacting governance structures supporting a variety of state services: security to safeguard user data, public services such as content moderation, regulations for civil behavior laid out in platform standards, and policing and sanctioning behavior. As social media has increasingly become the site for public discourse, platform policies are establishing these social media as ‘de facto regulators’ of opinion, information, and news, and how these are delivered by whom, to whom, and under what circumstances. Yet, these nascent social states are fragile. Platforms are challenged by new leadership directives, the problem of keeping up with the competing demands for content moderation, internal pressures from users, external pressures from domestic and non-domestic governments, and changing landscape of participants, external concerns, and social media features.
This research adopted the a ‘social media as state’ analogy to make sense of cases, controversies, and the continuously emergent practices on social media. While the ‘social media as state’ analogy has its limits, as social media have become more pervasive and seemingly independent of nation states, the analogy encourages attending to state-like functions as they are reproduced in the social organization of each social media state. Moreover, the state analogy provided the rationale for following on the idea of fragility from the Fund For Peace Fragile States Index, using their model to derive a set of fragility indicators for social media (https://fragilestatesindex.org/indicators/). Analysis of a wide range of literature on social media led to the derivation of a set of six Social Media Fragility Indicators which will be the focus of this talk. In brief, the indicators address: technical vulnerabilities and protections, internal social pressures, fragmentation, human rights, economic conditions and external pressures
Project Overview and Reflections
No full textThe following reflections come from faculty and Working Group members who participated in the project. Their insights reveal how engaging with language instruction and interdisciplinary content sparked new connections—not only for students, but for educators themselves. These perspectives highlight the value of stepping outside disciplinary boundaries and embracing the exploratory potential of language-centered teaching
Black Mothers\u27 ADHD Experiences, Attitudes, and Support-Seeking Decisions
No full textBackground: ADHD, particularly when left untreated, is associated with significant impairment in several domains. Black youth may still be under-diagnosed with ADHD, resulting in symptoms going untreated. Black mothers may determine non-professional delivered supports to be more appropriate and acceptable than medication for their children with ADHD. As they tend to be the primary healthcare decision makers, it is important to develop a greater understanding of Black mothers’ ADHD support-seeking decision making for their youth. Method: The present study used semi-structured interviews guided by Kleinman’s Explanatory Models of Illness to discuss Black mothers’ decision making for their youth with ADHD. Using Interpretative Phenomenological Analysis, the present study explored Black mothers’ (n = 12) (1) experience with and conceptualization of ADHD, and attitudes toward ADHD as a disorder and (2) treatment and support-seeking decisions for their children with ADHD. In addition to participating in semi-structured interviews, participants completed two measures: the Attention-Deficit/Hyperactivity Disorder – Rating Scale – 5th edition (ADHD-RS-5) and the Impairment Rating Scale (IRS) to characterize the symptom severity of their child’s ADHD. Results: ADHD symptom count across participants indicated significant symptoms of both inattention and hyperactivity/impulsivity. Further, the overall functioning rating of the study sample indicated significant overall impairment with the areas of most significant impairment being academic progress, relationship with playmates, and self-esteem. Using Interpretative Phenomenological Analysis, four superordinate themes emerged from the interviews: (a) ADHD as a problem to be solved, (b) an inclination to assign accountability and blame for ADHD, (c) chosen treatment/support must do more good than harm, and (d) understanding and acceptance of ADHD by others would be the ideal support. Discussion: The present findings indicate the importance of understanding and supporting an increase in providers’ willingness to learn from Black mothers as the treatment and support decision makers for their children. The dearth of accurate representation of ADHD and of medical and mental health care providers’ contributions in more accessible spaces may contribute to Black mothers’ lacking feelings of support. The present study provides a starting point for stakeholders to affect change in the area of mental health service provision, specifically support for ADHD, for Black youth
Mobility at the Margins: Insights into gender, class and climate change in Islamabad\u27s transit spaces
No full textThis dissertation sets out to explore how the urban built environment- particularly public transport infrastructure- shapes class, gender and climate inequality in Islamabad, Pakistan. I tried to understand this issue by using mixed methods, primarily consisting of semi-structured in depth interviews, supplemented by ethnographic fieldwork and a brief consultation of english-language newspaper archives. I conducted 95 interviews in total, and clocked in around 80 hours moving around the city and engaging in a mobile ethnography. The first empirical question that I asked relates to the state and urban transportation planning. I wanted to understand what assumptions regarding class-based, gender-based and climate-based inequality, are embedded in the design and planning process of public transport in Islambad? The second empirical question focuses on the gendered experiences of commuting using public transport. I explored the embodied experienes of low-income women and khwajasira while using public transport every day? How do they cope with sexual harassment and what infrastructural fixes can be used to challenge the existing design of transport and the city at large? Lastly, I wanted to understand how everyday mobility intersects with climate change. So I looked into the ways that two particular effects of climate change in Islamabad- rising heat levels and erratic rainfall- shape the use of public transport? How is climate change shaping the mobility of low-income commuters and what additional costs are they bearing as a result? My findings suggests that big transport infrastructure, such as Bus Rapid Transit (BRT), represents more than mobility in places such as Pakistan. The ‘development as infrastructure’ ideology propagated by the state has successfully manufactured consent among a big part of the population. Living in crowded urban centers, with no access to air-conditioning in extreme levels of heat, makes transport infrastructure such as the Metrobus look very appealing even if it comes with issues such as time poverty and a lack of first-and-last mile connectivity. It can fit into an orderly vision of the city and maintain its classed character. The city authorities are not interested in developing a holistic vision of the city by connecting the Metrobus to loosely regulated minibus taxis. Therefore, people resort to using the two-wheeler as it provides speed and door-to-door access with better mileage and is considerably cheaper than a car. Meanwhile, projects such as the Metrobus have been opposed by many civil society and environmental activists who argue that such infrastructure comes at the cost of displacing communities, increasing concrete cover, challenging the spatial configuration of cities and a burden on the exchequer. These ideological contestations reflect social structures such as gender and class, as lived and embodied in Pakistan. Most female and khwajasira respondents preferred the Metrobus to minibus taxis due to two main reasons. First, it had gender segregated sections and second, it had cameras. However, they still reported experiencing some form of harassment going to and from Metrobus stops as well. Therefore, my findings confirm research from other parts of South Asia which says that women prefer gender segregated modes of transport. Not only that, but feminist design interventions, such as physical barriers between the men’s and women’s section have the potential to provide a more comfortable experience. Access to public transport and affordable housing are also inextricably linked. Those who live away from city centers, are forced to pay higher fares and experience greater time poverty. Lastly, the two major effects of climate change that I studied are rising heat levels and unpredictable rainfall. Both these cause major challenges as the first and last mile journey becomes even more difficult in sweltering heat or pouring rain. Overall, zooming in on how urban design interacts with mobility helps to underscore the complex nature of everyday mobility, which is tied to social class, gender, the effects of climate change, the built environment, labor laws and growing urbanization in various ways. I frame my dissertation within the ‘right to the city’ framework, by Henri Lefebvre to show how inequality and mobility are linked and experienced in Islamabad. I elaborate on contributions from feminist geography to add a gendered analysis to the ‘right to the city’. Lastly, I lay out concerns raised by urban theorists from the Global South that emphasize context-specific research, especially in the urban realm. By combining these, I argue that urban planning and design, as an exercise, must borrow lessons from urban sociology to design better, people-centric cities
MENTAL ILLNESS BETWEEN SYSTEMS AND RELATIONS: DEPRESSION, BIPOLAR DISORDER, AND ANXIETY IN MODERN CHINA
No full textHow do individuals in contemporary China experience and narrate mental illness, and what are the social origins of this mental distress? To answer these questions, this dissertation draws on long-term ethnographic fieldwork in three psychiatric hospitals, over 90 in-depth interviews with patients, families and psychiatrists, and more than 300 cases in clinics. It adopts a sociological approach to analyze these data to uncover how mental suffering emerges from the tensions between systems and relations in a rapidly transforming society. This dissertation develops the concept of the “colonization of the mental world” to theorize how mental life is increasingly colonized by the performance-driven structuring of emotion, the moral disciplining of intimate relationships, and the individualization of emotional responsibility. In primary and secondary schools, exam-centered discipline, quantification of student value, competitive and deprived peer relationships, and pathological family relationships are experienced by students as insecurity, self-blame and low self-esteem, anxiety and fear, and learning aversion. In universities, amid both the continuity and disjunction of the transition from secondary education, the continuous performance pressure exacerbated by the unequal distribution of cultural capital cultivates uncertainty about the future and heightened employment anxiety. The instrumentalization and rationalization of social relationships, coupled with pervasive social comparison and the dictates of the social clock, further exacerbates university students’ anxiety and alienation. In the workplace, neoliberal performance culture, Confucian hierarchical authority, and precarious labor regimes push workers into cycles of shame, continuous self-optimization, future anxiety, and burnout. Within families, the imbalance of emotional labor, emotional repression and gender discipline, and the moral anxiety in intergenerational responsibility convert intimate relationships into emotionally exhausting structures. These findings challenge the popular medicalized discourse of mental illness in social science by showing that mental suffering is rooted in the cumulative weight of systemic encroachment on mental life, like through academic pressures in education, moralized responsibility in families, and self-optimization demands in the workplace. By tracing how these systems, which include not only administrative and market-based logics in Habermas’ sense but also cultural and ethical modes of discipline, permeate the emotional fabric of everyday life, this dissertation reconceptualizes mental illness as a profoundly social and moral phenomenon, one that signals the erosion of subjectivity in modern society. Confronting this crisis requires not just clinical intervention or social welfare, but a sustained ethical and structural reconstruction of the mental world as a space of resonance, meaning, and relational vitality for contemporary Chinese people
The Many Futures of Work and Skill: Gender and Occupational Influences on Online Freelancers\u27 Skilling Experiences in Platform Work
No full textThis dissertation examines how people understand, develop, and apply skills in one emerging future of work: online freelancing on digital labor platforms. As more workers turn to new forms of work enabled by emerging technologies, they are reshaping not only how work is performed but also what it means to be skilled and how one becomes skilled. My dissertation focuses on online freelancing, recognizing that workers must build skills to navigate new technologies, interfaces, and expectations, often in environments that overlook their lived experiences and diverse challenges. Drawing on a longitudinal, mixed-methods study of 108 online freelancers on Upwork, I address two central questions: (1) how the nature of work and skill development is evolving, and (2) how these evolutions affect individuals differently depending on their social positions, such as gender, occupation, and other circumstances. A central contribution of this dissertation is a rethinking of skill in the future of work, not as a fixed human trait independent of context, but as an emergent experience shaped by the interactions among people, technologies, and social conditions. This situational and relational perspective helps explain why workers performing similar tasks may encounter vastly different skill learning curves, barriers, and outcomes. By surfacing these diverse realities, my dissertation also offers actionable insights for designing more sustainable and inclusive user experiences in platform-mediated work
