49755 research outputs found
Sort by
Advancing Dialogue Systems for Social Good: Human-Robot Interaction, Knowledge Integration, and Real-World Applications
No full textDialogue systems are increasingly being developed to address societal needs, offering opportunities for meaningful interactions that contribute to social good. This thesis explores the development of dialogue systems by addressing key challenges, including mitigating the risks of self-anthropomorphism to ensure ethical and trustworthy interactions, integrating knowledge to enhance accuracy and reliability, and deploying these systems in practical applications such as anti-scam systems and curriculum-aligned educational chatbots.
We begin by examining self-anthropomorphism in dialogue systems, where AI mimics human identities and behaviors, potentially leading to misplaced trust and unrealistic expectations. Our initial work, Robots Don’t Cry, provides an analytical foundation by assessing how feasible human-like utterances appear to users across diverse dialogue datasets. These findings reveal that many anthropomorphic responses are perceived as inappropriate or misleading, regardless of the AI's embodiment. Building on these insights, our follow-up work, Pix2Persona, introduces a dataset that systematically models and transforms dialogue responses along the anthropomorphism spectrum. This enables dialogue systems to dynamically adjust their persona and language style, aligning with ethical considerations and user expectations across different applications.
Next, we tackle the challenge of ensuring that dialogue systems provide reliable, knowledge-grounded responses, an essential component of socially beneficial interactions. Many systems struggle to generalize across unseen topics or to integrate diverse external information sources effectively. To address this, we develop a unified knowledge-based framework that consolidates resources such as wikis, knowledge graphs, and dictionaries into a cohesive input format for language models. This approach enhances the contextual accuracy and robustness of responses across multiple domains, enabling dialogue systems to support real-world applications more safely and effectively.
We demonstrate the practical impact of these approaches through two dialogue system applications: an anti-scam system designed to help users resist fraudulent schemes, and a curriculum-driven EduBot that provides structured conversational practice aligned with textbook content. In addition, we explore the use of vision-language models to support dialogue systems for people with blindness and low vision. We evaluate these models on fundamental visual reasoning tasks related to navigation, offering insights into their potential and limitations for future multimodal assistive systems.
Through these contributions, this thesis illustrates how dialogue systems can be responsibly designed, knowledge-grounded, and practically deployed to advance social good and improve lives
The Role of the Therapeutic Alliance in a Peer Mentorship Program for Individuals with Eating Disorders
No full textPurpose:
The therapeutic alliance is widely recognized as a critical component in successful psychotherapy outcomes. Peer mentorship programs leveraging mentors' lived experiences have shown promise in enhancing engagement and promoting recovery in individuals with eating disorders. However, the specific mechanisms through which therapeutic alliance operates within peer mentorship contexts remain unclear. This study examined the role of therapeutic alliance in facilitating engagement and improved clinical outcomes while exploring patient perception of helpful program elements in a peer mentorship program specifically designed for individuals with eating disorders.
Method:
Forty participants diagnosed with eating disorders were enrolled in a randomized controlled trial comparing a peer mentorship intervention (mentors with lived experience of eating disorders) against a social support intervention (mentors without eating disorder experience). Participants were explicitly informed about the difference between conditions regarding mentors' lived experience prior to their participation. Participants completed the Working Alliance Inventory–Short Revised (WAI-SR) to measure therapeutic alliance, alongside assessments of eating disorder symptoms (Eating Pathology Symptoms Inventory), anxiety (State-Trait Anxiety Inventory), depression (Patient Health Questionnaire-9), and eating disorder-related quality of life (Eating Disorder Quality of Life). Additionally, qualitative responses regarding the perceived helpfulness of the mentorship were analyzed using thematic analysis.
Results:
Participants in the peer mentorship group reported significantly stronger therapeutic alliances (p < .001, Cohen’s d = 1.48) compared to the social support group. A regression analysis controlling for group type found that higher WAI-SR alliance scores were associated with a greater reduction in depression scores (PHQ-9 symptom slope: B = –0.021, t(36) = –2.173, p = .036). In contrast, alliance scores were not significantly related to changes in anxiety symptoms (STAI slope: B = –0.029, t(36) = –1.732, p = .092) or in eating disorder symptoms (EPSI-BD slope: B = –0.008, t(36) = –0.760, p = .452). There was no significant relationship between alliance and change in eating disorder-specific quality of life (EDQOL slope: B = –0.001, t(36) = –0.490, p = .627). Therapeutic alliance was also significantly associated with increased participant engagement, including higher session attendance and longer participation duration. Qualitative analysis revealed key relational benefits unique to peer mentorship, notably emotional validation, authentic understanding, and recovery-oriented narratives, which participants identified as critical for fostering trust and engagement.
Limitations and Future Directions:
Several limitations should be noted, including the relatively small and homogeneous sample, reliance on self-report measures, and potential self-selection bias from participants opting into the study after waitlist assignment. Future research should employ larger, more diverse samples and incorporate objective clinical assessments alongside self-reports. Longitudinal studies investigating the durability of therapeutic alliance effects and hybrid models integrating peer mentorship with structured, evidence-based treatments are recommended to further understand and optimize mentorship interventions.
Conclusions:
Therapeutic alliance emerged as a central factor in improving depressive symptoms and enhancing participant engagement in peer mentorship programs for eating disorders, primarily driven by the relational benefits associated with mentors' lived experiences. These findings advocate for mentorship programs emphasizing relational competencies over precise diagnostic or demographic matching. Integrating peer mentorship's relational strengths with structured, evidence-based therapeutic interventions appears promising for effectively addressing both engagement and clinical symptom improvement, ultimately supporting sustained recovery trajectories for individuals with eating disorders
Disciplining Development: US Universities and Humanist Worldmaking in the Black Atlantic
No full textAt the intersection of comparative literature and international education studies, this dissertation examines how African and diasporic intellectuals endeavored to shape new, liberating networks of cultural exchange during the decolonizing period of the 1960s, and how American academia emerged as both an enabling and constraining mediator of this development.
Three case studies, each representing a hub of the Atlantic world, illustrate different facets of this history. From 1959-1961, Tom Mboya’s student airlifts mobilized a cooperative grassroots strategy for expanding East Africans’ access to higher education, before the project was absorbed and redirected by US philanthropic corporations with an elitist mentality. Inspired by his own educational experiences, Ghanaian author Ayi Kwei Armah’s novels Fragments (1970) and Why Are We So Blest? (1972) dramatize the role of such philanthropy in binding Black intellectuals and postcolonial literary culture to a neoimperial economic system.
Like Armah himself, these novels look beyond Western institutions to imagine alternative cultural infrastructure for independent Pan-African development. Jorge Amado’s Afro-Brazilian novel Tent of Miracles (1969) illustrates how even staunchly subaltern cultural initiatives may be absorbed by the coercive force of international capitalism, but it finds hope for humanist renewal despite the inevitability of compromise.
Together, these stories illuminate how Black intellectuals in the postcolonial South Atlantic reimagined education, the politics of knowledge, and the future of cultural circulation amid the contested rise of a US-dominated international culture market
A remote input of African dust to Last Glacial Europe
No full textDuring the Last Glacial Maximum global surface air temperatures were up to 6 °C lower than preindustrial levels and the mineral dust cycle was highly active, with global dust loading two to four times higher than during the Holocene. Loess deposits and Greenland ice cores show peak dust concentrations during this time. While Asian sources were traditionally seen as the main contributors to dust transported to Greenland, recent studies using geochemical methods suggest a mix of Asian, North African, and European sources. Europe experienced intense dust activity, with mineral particles largely emitted from regional sources. Here we present the trace elements, and strontium and lead isotopes from Last-Glacial Maximum samples collected at 15 sites across Europe. The results reveal that fine dust originated from remote sources, potentially northern Africa. Earth System model simulations support this finding, highlighting Northern Africa’s substantial role in dust deposition during glacial periods across the Northern Hemisphere
Automated Model Discovery & Explanation Generation for Physicochemical Systems using Artificial Intelligence
No full textThe advent of powerful computational resources coupled with substantial progress in algorithms and increased dataset size, have led to the development of machine learning (ML) models for physicochemical systems. Such models often come at the cost of interpretability and lack of explainability by a domain expert, thereby limiting its usage. Unlike applications such as game playing, recommendation systems, or even chatbots which have permeated everyday life in recent times, science and engineering systems are steeped in first-principles knowledge that must be leveraged to render meaningful explanations of the mathematical relations that attempt to model complex physicochemical systems. Accordingly, there is a need to develop models that can be used to subsequently explain salient aspects of the physicochemical phenomena. In this work, an end-to end data-driven model discovery engine and explanation generation artificial intelligence (AI) system is developed, which is tested on two real-world case studies of varying scales.
Chapter 1 introduces the need for combining first-principles knowledge into the modeling workflow in order to obtain more meaningful results. As shall be explained, the cost of a mistake in the sciences and engineering disciplines may prove to be fatal, and thus it is imperative that the models generated be explainable.
Chapter 2 outlines a data-driven symbolic model discovery engine that outputs an ensemble of best performing models when provided data. This system relies on the a priori inclusion of first-principles knowledge of the system being modeled, resulting in meaningful functional transformations.
Chapter 3 expands on the algorithm presented in the preceding chapter, to model systems of ordinary and partial differential equations, and presents the efficacy of the approach across a wide variety of case studies.
Chapter 4 applies the modeling engine developed in the preceding chapters to a bubble column aeration problem with real-world data. The resulting models obtained are compared to the analytical ground-truth, such that the improvement over the analytical model can be captured clearly—something that would not have been possible as effectively using a black-box modeling approach.
Chapter 5 applies the modeling engine to a structure-to-property prediction problem concerning the isothermal adsorption capacity of three adsorbates on zeolite structures. Varying in scale compared to the bubble column aeration system, valuable insights about the most descriptive structural properties were gained by virtue of the interpretable modeling engine.
Having successfully developed interpretable ML models for the physicochemical systems in the preceding chapters, Chapter 6 outlines the development of a large knowledge model (LKM) for automatically generating explanations from ML models. This first requires the extraction and organization of domain-knowledge from textual sources, followed by a hierarchical explanation generation strategy that yields increasingly natural language explanations. This blend of combining domain-knowledge with the natural language capabilities of modern tools such as large language models allows for explanation generation of ML models describing complex physicochemical systems. Finally, the dissertation concludes by summarizing the work undertaken, and outlines potential future directions of research
Becoming Multilingual/Multicultural in a Spanish-English Dual Language Bilingual Program: Partnering with Chinese Bilingual Families and Communities
No full textThis qualitative study aimed to explore the motivations and strategies employed by Chinese bilingual families for enrolling their children in a Spanish-English dual language bilingual program, with a focus on how they support their child’s development of multilingualism and academic progress in a third language.
Using Cultural Historical Activity Theory (CHAT) and transnational feminist frameworks, the study examined how families from non-Latinx backgrounds leverage their transnational experiences, multilingual resources, and community assets to support their child’s linguistic and cultural development. Specifically, it investigated how these families navigate the challenges of a dual language program where neither of the instructional languages (Spanish or English) was their home language. Data collection involved semi-structured individual interviews, photo elicitation, activity-based Zoom focus group interviews, voice journal entries, students' homework packets, and the researcher’s field notes.
The findings indicated that family members explored multiple pathways in the home for learning to learn, including negotiating technological tensions, fostering intergenerational partnerships, and preparing children for independence, all of which contribute to the development of learning agencies and global citizenship. Additionally, the findings emphasized culturally relevant and multidimensional learning facilitated by support from diverse communities, highlighting the importance of school communication and trust, the multi-layered sense of community for adult family members, and the impact of bicultural and bilingual neighborhoods on children’s development and cultural preservation.
Moreover, this study explored the mediational role of transnational and transcultural experiences in multilingual homes, examining unrecognized transnational resources that shaped children’s learning, including long-practiced multilingual settings and previous experiences. It also highlighted strategies for preserving language and culture in transnational families, such as access to resources from their country of origin and embracing transnational hybridity, while leveraging Chinese cultural values and family expectations to support children's educational development.
This study also examined the impact of the COVID-19 pandemic on minoritized learners, focusing on their experiences with online classes and the strategies employed to strengthen social and emotional connections during remote learning. The findings of this study offered valuable insights for educators, school leadership, and researchers to create culturally and ethnically responsive learning environments for families and students who are often marginalized in traditional educational settings
Discovery of programmable RNA-guided mechanisms for transposition and gene regulation
No full textCRISPR-Cas systems are widespread in bacteria and archaea, providing potent mechanisms foradaptive immunity. Their reprogrammable, RNA-guided nucleic acid targeting and cleavage capabilities have been harnessed to create high-precision, next-generation genome editing tools. Although CRISPR-Cas systems are thought to have evolved from transposon-encoded nucleases, the extent to which these homologous proteins can perform RNA-guided functions, and how these systems affect transposition, remain largely unexplored. Moreover, numerous transposon-derived nucleases and Cas homologs are found alongside novel genes unrelated to bacterial immunity.
These associations, often involving nuclease-inactivated Cas-like proteins, suggest co-option and neofunctionalization of RNA-guided DNA targeting capabilities. Through bioinformatic analyses, multiple such systems were identified, unveiling a diverse repertoire of repurposed CRISPR-Cas- like pathways. Next-generation sequencing, combined with biochemical and structural approaches, elucidated the mechanisms by which CRISPR-associated transposons (CASTs) achieve efficient DNA integration and revealed how transposon-encoded Cas homologs facilitate transposon spread.
Furthermore, novel RNA-guided regulators were discovered that either repress or activate gene expression in a fully programmable manner. Strikingly, these RNA-guided activation systems operate independently of canonical promoter motif constraints, expanding our understanding of bacterial transcription. This work underscores the functional versatility of RNA-guided systems, highlighting their roles in pathways beyond antiviral immunity. These discoveries provide a roadmap for developing enhanced gene-editing technologies and novel tools enabling programmable gene expression
Modulation of Stratosphere-to-Troposphere Transport of Ozone by North Pacific Weather Patterns, version 1
No full textThere is a new version of this dataset available at https://doi.org/10.7916/135z-jg21.
Stratosphere-to-troposphere transport (STT) of ozone is a major natural source of tropospheric ozone and is tightly linked to both stratospheric and tropospheric circulations. However, the complex interactions of drivers make it challenging to understand the subseasonal variability of STT. In this study, we apply a weather pattern perspective to address the midlatitude internal variability of STT over western North America, the strongest global hotspot. Six weather patterns are identified using self-organizing maps applied to year-round 500 hPa geopotential height anomalies from a nudged WACCM6 simulation. Extreme STT occurs preferentially in the pattern with a deep trough south of Alaska, which resembles the Pacific Trough regime. A larger ozone source and an aligned strong jet lead to frequent intrusions in this specific pattern. This framework is able to explain the ENSO impact on STT variability and its temporal evolving preference. It also suggests potential predictability of stratospheric intrusion through weather patterns
Field under Constraint: Soviet Literary Journals and Autonomy in Soviet Literary Life
No full textField under Constraint: Soviet Literary Journals and Autonomy in Soviet Literary Life examines how autonomy in cultural production can emerge and be sustained within institutions expressly designed to suppress it. Focusing on six major Soviet literary journals between 1956 and 1990, the dissertation argues that autonomy in Soviet literature was not confined to dissident or underground activity but was embedded in official institutions themselves. Drawing on archival records, editorial correspondence, interviews, and a large bibliographic dataset (more than 35,300 publications), it demonstrates that journals developed distinctive repertoires and discretionary strategies in response to shifting political campaigns, institutional ambiguities, and audience expectations.
Theoretically, the study integrates Bourdieu’s field theory with insights from theories of gradual institutional change. It shows that autonomy was relational, incremental, and situational, devised through the contradictions between political authority and literary legitimacy, and sustained through editorial practices of discretion. By combining structural analysis with qualitative case studies of Iunost’, Znamia, and Oktiabr’, the dissertation reveals how journals transformed their assigned role as instruments of ideological dissemination into precarious yet durable sites of innovation and change. More broadly, it demonstrates that authoritarian systems are never fully closed. The very institutions designed to enforce conformity generated contradictions that enabled autonomy
Recasting BEPS’ Pillar 2 for green FDI in developing countries
No full textThis Perspective juxtaposes Pillar 2 minimum-tax rules and outcomes with the urgent need for increased climate investment in developing states. It rationalizes the role of incentives in climate-aligned investment, international-law principles such as the common-but-differentiated-responsibilities, and proposes modifications to Pillar 2 to facilitate investment and development objectives