SVSofia Valanci-Aroesty

Sofia Valanci-Aroesty

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Sofia Valanci-Aroesty has a total of 85 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Twelve tips for teaching research skills in the age of agentic AI: A guide for health professions educators

    by Neil Mehta, Jennifer Benjamin, Seysha Mehta, Sofia Valanci, Heather MacNeill, Ken Masters

    Artificial Intelligence (AI) in research now spans a broad spectrum, from tools that assist with discrete tasks to agentic systems that can plan, execute and iterate across entire workflows with mi...

  • Between Incremental Change and Transformation: The Emerging Role of Generative AI in Continuing Professional Development Within Health Care Systems

    by Sofia Valanci-Aroesty, Vjekosav Hlede, Heather MacNeill, Álvaro Margolis, Ken Masters

    Introduction: Generative Artificial Intelligence (GenAI) tools are reshaping professional education by enabling tailored content generation, automating tasks, and enhancing creative output. While widely studied in undergraduate and postgraduate education, little is known about their role in Continuing Professional Development (CPD), where professionals function as system-level educational leaders rather than learners. Methods: A cross-sectional international survey (March–June 2025) explored GenAI adoption among CPD professionals. Using the validated Generative Artificial Intelligence Acceptance Scale and the Artificial Intelligence Attitude Scale, the study examined performance expectancy, effort expectancy, facilitating conditions, social influence, and general attitudes toward Artificial Intelligence. Quantitative data were analyzed descriptively and inferentially, and qualitative open-ended responses were analyzed using reflexive thematic analysis. Results: A total of 110 respondents from 14 countries participated. Most reported moderate to high familiarity and comfort with GenAI, with frequent users demonstrating significantly greater confidence and perceived proficiency. Performance expectancy was high, particularly regarding efficiency and productivity, but lower for complex problem-solving. A proportion of respondents expressed undecided positions, indicating transitional stages of adoption. Qualitative analysis revealed three overarching themes: (1) adoption and utilization, (2) personal perspectives, and (3) perceived challenges and benefits. Discussion: CPD professionals view GenAI as an augmentative tool that enhances efficiency and creativity rather than replacing human expertise. While individual readiness is high, sustainable adoption depends on institutional support, structured training, and ethical integration. This study provides novel insights into GenAI adoption at the CPD system level and highlights the importance of engaging the “undecided” to move from incremental use to meaningful transformation.

  • Top Cited Papers

    Using locally-hosted Small Language Models (SLMs) to protect student, patient and research subject data in Health Professions Education

    by Ken Masters, Sofia Valanci-Aroesty, Jennifer Benjamin, Neil Mehta, Heather MacNeill

    Health Professions Education (HPE) scholarship is crucial to advancing HPE, and frequently relies on student, patient, or research participant data (qualitative or quantitative). The public release...


  • Open Access

    AI-Enhanced Continuing Professional Development as an Evolving Sociotechnical System: Multimethod Theoretical Framework Development Study

    by Vjekoslav Hlede, Sofia Valanci, G Robert D'ANtuono, Heather Dow, Ronan O'BEirne, Richard Wiggins

    Background: Artificial intelligence (AI) is changing continuing professional development (CPD) in health care and its interactions with the broader health care system. However, current scholarship lacks an integrated theoretical model that explains how AI impacts CPD as a complex sociotechnical system. Existing frameworks usually focus on isolated phenomena, such as ethics, literacy, or learning theory, leaving unaddressed the dynamics of how those phenomena interact in the complex sociotechnical AI-enhanced CPD system, as well as the new roles that AI-empowered patients and society play. Objective: The objective of this study is to propose a comprehensive, theory-driven framework that provides insight into how AI transforms CPD systems. The goal was to integrate established AI constructs with Complexity Theory (CT) and Actor-Network Theory (ANT) to develop a model that guides practice, research, and policy. Methods: We conducted a multimethod theory construction. The process started with identifying the AI-enhanced CPD as an established yet evolving phenomenon. Through a structured literature review, the main building blocks of AI-enhanced CPD were identified, as well as the ontological base (CT and ANT). The model was developed through iterative human-led and AI-assisted abductive analysis. The final model was abductively validated on a case study of a national organization pioneering AI use, demonstrating the theoretical model makes sense in practice. All conceptual decisions were reviewed collaboratively by the author group. Results: The ALEERRT-CA framework is made of 6 pillars: AI literacy, explainability, ethics, readiness, reliability, and learning theories, and 2 theoretical lenses: CT and ANT. CT elucidates macro-level system behaviors in the AI-enhanced CPD system. Those behaviors include emergence, feedback loops, adaptation, and reality made of nested complex systems. ANT explains how localized interactions among human and nonhuman actors shape AI-enhanced CPD. Together, these lenses illustrate how AI redistributes agency, amplifies tensions, and generates emergent learning dynamics within CPD and the broader health care system. Conclusions: This study presents a novel conceptual model of AI-enhanced CPD as a sociotechnical system. The integration of CT and ANT with AI constructs improves explanatory power of the ALEERRT-CA framework. Educators, program leaders, and policymakers can use the framework as a structured toolset to evaluate AI readiness, design responsible AI-enhanced CPD practices, and plan future empirical research. The framework provides a theoretical lens for observing the rapidly evolving field of AI-enhanced CPD and health care practice.


  • Open Access

    Promise or Peril: Evaluating the Performance of OpenEvidence in Clinical Decision Making in the Global Context. (Preprint)

    by Jennifer Benjamin, Ken Masters, Brenda Kunga, Sofia Valanci, Anoop Agrawal, Jen-Feng Liang, Alvaro Margolis, Adhari Al Zaabi, Sudha Rajan, Neil Mehta

    BACKGROUND Clinicians face the impossible task of keeping up with evidence-based literature to inform clinical decision making. OpenEvidence holds great promise by providing latest evidence to inform patient care.AI solutions are increasingly used to support physicians during clinical decision making. Clinicians need to critically analyze and evaluate outputs and be aware of biases that exist prior to application, especially when caring for patients in global contexts. OBJECTIVE To evaluate performance of OpenEvidence (OE) in global context. METHODS In this study a sample of 11 clinicians from the United States (n=3) and six other countries (n=8) performed standardized searches for 10 common clinical conditions. Search results were evaluated using the Rubric Assessment Scale and additional questions compared U.S. based results versus other countries. Data were analyzed using descriptive statistics, and Wilcoxon rank-sum test for comparisons of median differences. All statistical tests were two-tailed p-value < .05 was considered statistically significant. RESULTS A total of 80 evaluations showed 88% accuracy and 89% efficiency with significant geographic disparities. U.S. based searches demonstrated alignment with management guidelines (p=0.02), higher perceived overall value (p=0.02) and inclusion of local experts(p<0.0001). The global group reported wider, less relevant differential diagnoses (p=0.03) and demonstrated lack of regional and socio-cultural contexts. CONCLUSIONS OE is a valuable tool but exhibits significant geographic bias, favoring U.S. clinical contexts with limitations for global applicability. There is critical need for training AI models on diverse and representative datasets to ensure accurate performance. Clinicians should critically appraise AI outputs prior to application to model responsible use of AI. CLINICALTRIAL NA

  • Addressing educational overload with generative AI through dual coding and cognitive load theories

    by Neil Mehta, Jennifer Benjamin, Anoop Agrawal, Sofia Valanci, Ken Masters, Heather MacNeill

    Health professions education faces a critical challenge: the volume and complexity of medical knowledge has outpaced the cognitive limits of learners. Cognitive load theory indicates that traditional text-heavy instruction overloads the working memory. Current educational materials fail to leverage the dual coding theory which proposes that content should be presented through both verbal and visual channels. Generative AI tools enable the creation of multimodal educational content. Platforms such as ChatGPT, Gemini, NotebookLM, and HeyGen enable the creation of multimodal educational content - audio summaries, interactive mind maps, infographics, narrated videos, and voice-based interactions - designed to improve information processing in the working memory. Multimodal AI-generated content potentially enhances comprehension, retention and learning efficiency, and provides scalable, easily updated resources. These tools align with Generation Z learning preferences and can be integrated into existing curricula to transform text heavy content into multimodal engaging content. These tools can be used by faculty and learners,with minimal technical expertise or time investment. To ensure success, institutions must provide faculty and learners training, with consistent access, and evaluate outcomes. Pilot programs with iterative feedback can guide thoughtful implementation. By aligning educational strategies with cognitive science principles, generative AI can play a transformative role in addressing educational overload and creating more effective, engaging learning environments.


  • Open Access

    Integrating Artificial Intelligence in Healthcare Continuing Professional Development: A Theoretical Framework (Preprint)

    by Vjekoslav Hlede, Sofia Valanci, G. Robert D'Antuono, Heather Dow, Richard Wiggins, Todd Dorman, Ronan O'Beirne

    UNSTRUCTURED Artificial intelligence (AI) enhanced continuing professional development (CPD) has great potential. It promises to transform the learning experience, improve learning outcomes, and ultimately help us improve patient care. Yet, AI is not a magical solution but a new element in the complex socio-technical systems that form our society. We need to understand the system to understand the power of AI. This position paper examines the need for a theoretical framework to help us better understand and guide interactions between AI and healthcare CPD. The 5 step theory construction methodology outlined by Borsboom et al. has been used to create the framework made of six foundational AI pillars: Literacy, Explainability, Ethics, Readiness, Reliability, and Learning Theories, and two complementary theoretical lenses: Complexity Theory and Actor-Network Theory.

  • Open Access

    Evolving Maintenance of Certification in Canada: A Collaborative Journey

    by Kate Runacres, Jill Garner, Chanelle Goulet, Farah Wissanji, Rhonda St Croix, Kate Marsden, Lucie Filteau, Guylaine Lefebvre, Sofia Valanci-Aroesty

    Continuous professional development (CPD) is crucial for physicians to maintain and enhance their skills. In response to the changing context of CPD and health care, this study applied a design thinking approach to transform and modernise the Royal College of Physicians and Surgeons of Canada’s Maintenance of Certification (MOC) Program. A member-wide survey and co-design sessions with physicians, CPD leaders, and patient representatives were conducted, emphasising the importance of their insights and experiences. The data revealed key themes for the programme such as fostering meaningful learning, addressing barriers to CPD, supporting collaboration, and responding to the need for modern, flexible CPD delivery methods. Using “empathy”, “define”, “ideate”, “prototype”, and “test” phases, we continuously refined the MOC framework of CPD activities based on comprehensive user experiences and needs insights. The revised framework was iteratively prototyped and validated to ensure it was user-friendly and aligned with professional and regulatory requirements. The findings underscore the effectiveness of the design thinking approach in creating a dynamic, responsive MOC framework that supports CPD and meets the evolving needs of medical professionals. This approach not only demonstrates the effectiveness of design thinking but also the importance of engaging users in the development process, making them feel valued and integral to the transformation of the MOC Program.

  • Establishing validity evidence for device-assisted advanced laparoscopic suturing tasks using simulation

    by Elif Bilgic, Allan Okrainec, John Paige, Saba Balvardi, Sofia Valanci, Caterina Masino, Pepa Kaneva, Mohammed Alabri, Katherine M. McKendy, Yusuke Watanabe, et al.

    Background Laparoscopic suturing (LS) is an essential technique required for a wide range of procedures, and it is one of the most challenging for surgical trainees to master. We designed and collected validity evidence for advanced LS tasks using an automated suturing device and evaluated the perceived educational value of the tasks. Methods This project was a multicentre prospective study involving McGill University, University of Toronto (UofT), and Louisiana State University (LSU) Health New Orleans. Novice (NS) and experienced (ES) surgeons performed suturing under tension (UT) and continuous suturing (CS) tasks. ES performed the tasks twice to establish proficiency benchmarks, and they were interviewed to develop formative feedback tools (FFT). Participants were assessed on completion time, error, Global Operative Assessment of Laparoscopic Skills (GOALS), and FFT. Data were analyzed using descriptive and inferential statistical methods. Results Twenty-seven participants (13 ES, 14 NS, median age 34 years; 85% male) completed the study. Eight were attending surgeons, 7 fellows, 4 PGY5, 5 PGY4, and 3 PGY3 (18 from McGill, 5 UofT, and 4 LSU). Comparing ES and NS, for UT task, ES had significantly greater task scores (570 [563–648] vs 323 [130–464], p value 0.00036) and GOALS scores (14 [13–16] vs 10 [8–12], p value 0.0038). Similarly, for CS, ES had significantly greater task scores (976 [959–1010] vs 785 [626–856], p value 0.00009) and GOALS scores (16 [12–17] vs 12.5 [8.25–15], p value 0.028). After FFTs were developed, comparing ES and NS, for both UT and CS tasks, ES had significantly greater FFT scores (UT 25 [24–26] vs 17 [14–20], p value 0.0016 and CS 30 [27–32] vs 22[17.2–25.8], p value 0.00061). Conclusion In conclusion, preliminary validity evidence was provided for the tasks. Once further validity evidence is established, incorporating the tasks into the training curricula could improve trainee skills and help to meet the need for better advanced suturing models.