About

Uri Wilensky is a Professor of Computer Science and a Lorraine Morton Professor at Northwestern University, affiliated with the School of Engineering. His research interests focus on computational models, agent-based modeling, and the development of scaffolds and tools to support learning in science and engineering education. Wilensky has made significant contributions to the field of computational thinking, emphasizing the design of emergent systems microworlds and interactive constructionist scaffolds to enhance epistemic learning of biological phenomena and other scientific concepts. He holds a Ph.D. in Media Arts and Sciences from MIT, a Master's in Mathematics from Brandeis University, and a Bachelor's in Mathematics and Philosophy from Brandeis University. Wilensky was awarded the Yidan Prize in 2025 and is the current Yidan Prize Laureate. His work includes supporting the development of agent-based modeling environments such as NetLogo and exploring how computational models can serve as tools for responsive teaching and equitable participation in science education. Wilensky's research impacts a broad range of societal and educational domains, emphasizing the importance of innovative methods in engineering and science education.

Research topics

• Computer Science
• Psychology
• Sociology
• Mathematics education
• Pedagogy
• Artificial Intelligence
• Social Science
• Programming language
• Theoretical computer science
• Engineering
• Data science
• Engineering ethics
• Distributed computing
• Biology
• Epistemology
• Computer architecture
• Mathematics

Selected publications

• Processes Matter: How ML/GAI Approaches Could Support Open Qualitative Coding of Online Discourse Datasets

  ArXiv.org · 2025-04-02

  preprintOpen access

  Open coding, a key inductive step in qualitative research, discovers and constructs concepts from human datasets. However, capturing extensive and nuanced aspects or "coding moments" can be challenging, especially with large discourse datasets. While some studies explore machine learning (ML)/Generative AI (GAI)'s potential for open coding, few evaluation studies exist. We compare open coding results by five recently published ML/GAI approaches and four human coders, using a dataset of online chat messages around a mobile learning software. Our systematic analysis reveals ML/GAI approaches' strengths and weaknesses, uncovering the complementary potential between humans and AI. Line-by-line AI approaches effectively identify content-based codes, while humans excel in interpreting conversational dynamics. We discussed how embedded analytical processes could shape the results of ML/GAI approaches. Instead of replacing humans in open coding, researchers should integrate AI with and according to their analytical processes, e.g., as parallel co-coders.

  PublisherOA PDFDOI

• Computational Modeling in Materials Science and Engineering: Student Responses to a Restructurated Introductory Course

  2025-08-21 · 1 citations

  articleSenior author
  PublisherDOI

• Computational Thinking Without Writing Code: What’s Next for Computational Modeling?

  Proceedings. · 2025-06-10

  articleOpen access

  Generative Artificial Intelligence (AI) introduces exciting new possibilities and challenges to the established field of computational modeling education.The ten posters in this symposium provide different perspectives on the changing landscape and present examples of educational programs, professional development strategies, pedagogical approaches, and digital tools that have helped learners and educators develop the skills needed to interrogate and cocreate scientific computational models with AI.The posters address each stage of computational modeling education, including reading and decoding models, modifying existing models, creating (or co-creating) models, and evaluating models.

  PublisherOA PDFDOI

• From Molecules to Ecosystems: Scientists’ Conceptualizations of Dynamic Equilibrium

  Proceedings. · 2025-06-10

  articleOpen accessSenior author

  The study investigates whether Dynamic Equilibrium (DE) can serve as a "powerful idea" (Papert, 1980) that bridges phenomena across STEM fields.Through semi-structured interviews with five scientists, we identified three themes: DE was rarely used explicitly but appeared across fields; reflection on DE prompted a more generalized framing; and DE spanned scales from molecules to human populations.These findings suggest DE's potential as a crosscutting concept that connects otherwise fragmented content in STEM education.

  PublisherOA PDFDOI

• Dynamic Equilibrium in Science Education: Cross-Disciplinary Perspectives and Representations in Israel and United States Standards and Textbooks

  Proceedings. · 2025-06-10

  articleOpen access

  The study examines how Dynamic Equilibrium (DE) is represented in science national standards and textbooks for high-school biology, chemistry, and physics in the US and Israel.DE, a crucial concept in understanding dynamic systems, is inconsistently represented across educational materials and students encounter difficulties learning about DE.Analyzing 17 textbooks and national standards, the research combines quantitative and qualitative content analysis to assess the frequency and nature of presentation of DE-related phenomena.The study identifies 256 DE-related phenomena, comprising 14% of all phenomena that are studied in science.The primary systems approach used is System Dynamics, which focuses on stocks and rates of flow at one description level.The main representational format is verbal, and computational models are scarcely used.Differences across disciplines and between countries were found.These findings emphasize the need for powerful representations of DE to enhance students' understanding of dynamic systems and improve science education. Introduction, theoretical perspectives, and research goalIn this paper, we focus on Dynamic Equilibrium (DE) in science standards and textbooks, which we are considering as a structuring concept across the sciences in high-school learning.DE is a central concept for understanding patterns in dynamic systems and specific to this work, in the natural sciences.In fact, any semistable pattern in nature might be a result of DE.A system that exhibits DE is in a stable state where opposing influences continuously balance each other, resulting in no net change to the system (Biology Dictionary, n.d.).DE is usually described through the ongoing processes within the system, which return the system to equilibrium, after it is perturbed.Familiar examples are homeostasis in biology, ecological systems, and chemical equilibrium.Despite the importance of this concept, DE is a challenging concept for students at all educational levels, even in universities.A major difficulty stems from the apparent contradiction between the system's stable appearance and the dynamic processes underlying it (Sarayr et al., 2006).For instance, students often misconceive stability solely in terms of constancy or resistance to disturbance and fail to relate it to energetic states, such as the system's tendency toward a minimum potential energy.In chemistry, students incorrectly interpret equilibrium as the complete cessation of reactions rather than a dynamic balance between opposing rates (Nakhleh, 1992;zmen, 2008).Biology students similarly confuse stability with stasis, for example, viewing homeostasis as a fixed, unchanging state rather than an active process of regulation (Zion & Klein, 2015).Moreover, understanding equilibrium as an emergent phenomenon poses additional challenges.Students struggle to recognize the relationship between micro-level interactions and macro-level system properties.Instead, they tend to adopt a deterministic-centralized mindset, attributing system-level patterns to individual components or leaders, as observed in misconceptions about chemical diffusion and traffic flow (Wilensky & Resnick, 1999;Chi et al., 2012).The Samon & Levy (2017) have shown that across with respect to gases, the more different the properties of the two levels are, the more challenging it is for students to understand them.DE can be described in terms of System Dynamics (SD;Forrester, 1961).SD models are used to analyze complex relationships within systems, focusing on how systems evolve over time at one level of description.It employs stocks, flows, and feedback loops to understand dynamic changes and interdependencies.Structure-Behavior-Function (SBF;Hmelo-Silver & Green Pfeffer, 2004) models are normative models of complex systems that represent three core aspects: structure (the components of the system), behavior (the causal mechanisms or processes within the system), and function (the purpose or roles of the system and its components).SBF models are widely used for analyzing, designing, and teaching about complex systems by focusing on how these three elements interact to define system operation and purpose.Agent-Based Modeling (ABM; Wilensky & Rand, 2015) models examine systems through the perspective of multiple agents engaging in dynamic interactions over time and within a spatial context.These models focus on understanding how simple and local interactions among the components of a system at the micro-level lead to complex and sometimes unpredictable behavior at the macro-

  PublisherOA PDFDOI

• Processes Matter: How ML/GAI Approaches Could Support Open Qualitative Coding of Online Discourse Datasets

  Computer-supported collaborative learning/˜The œComputer-Supported Collaborative Learning Conference · 2025-06-10 · 1 citations

  articleOpen access

  Open coding, a key inductive step in qualitative research, discovers and constructs concepts from human datasets.However, capturing extensive and nuanced aspects or "coding moments" can be challenging, especially with large discourse datasets.While some studies explore machine learning (ML)/Generative AI (GAI)'s potential for open coding, few evaluation studies exist.We compare open coding results by five recently published ML/GAI approaches and four human coders, using a dataset of online chat messages around a mobile learning software.Our systematic analysis reveals ML/GAI approaches' strengths and weaknesses, uncovering the complementary potential between humans and AI.Line-by-line AI approaches effectively identify content-based codes, while humans excel in interpreting conversational dynamics.We discussed how embedded analytical processes could shape the results of ML/GAI approaches.Instead of replacing humans in open coding, researchers should integrate AI with and according to their analytical processes, e.g., as parallel co-coders.

  PublisherOA PDFDOI

• Characterizing Integrated Learning of Disciplinary Core Ideas and Science Practices in a Computational Thinking (CT)–Integrated Biology Curriculum

  Journal of Science Education and Technology · 2025-08-23 · 1 citations

  articleSenior author
  PublisherDOI

• Engaging Millions of Worldwide Youth in Informal STEM Learning: Uncovering Open-Ended Design Principles that Drive Physics Lab's Success

  2025-06-23

  articleOpen accessSenior author
  PublisherDOI

• LEAR: LLM-Driven Evolution of Agent-Based Rules

  Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2025-07-14

  articleOpen accessSenior author

  This study investigates the feasibility and effectiveness of integrating Large Language Models (LLMs) as mutation operators within Genetic Programming (GP) frameworks so as to evolve agent behaviors in multi-agent systems (MAS) and provide benchmarks that evaluate the efficacy of LLM-generated code in multi-agent domains. Our approach leverages the sophisticated code-generation capabilities of LLMs to introduce semantically meaningful variations during the evolutionary process. Specifically, we explore and systematically compare these different prompting strategies: zero-shot, one-shot, and two-shot prompting as well as prompting the generation of commented code to assess their impact on the quality of evolved agent behaviors. Additionally, we propose a novel methodology where evolution operates at a higher abstraction level by mutating pseudocode representations of agent behaviors, subsequently converting them into executable code through another LLM-mediated step. This strategy capitalizes on the extensive natural language training data of LLMs, potentially enabling the discovery of more innovative solutions. Our results indicate that LLM-driven mutation with comment generation enhances agent performance while mutating pseudocode representations yields reduced performance. This research contributes valuable insights regarding the integration of LLM-driven GP techniques into MAS, highlighting both the potential and limitations of these approaches. All code is open-sourced at https://github.com/can-gurkan/LEAR.

  PublisherDOI

• Meta-Theoretic Competence for Computational Agent-Based Modeling

  2024-06-20

  articleOpen accessSenior author

  In the U.S. context, science standards encourage educators to engage students in modeling practices, including computational modeling. While much work has investigated the productivity of computational modeling with respect to students’ development of scientific content knowledge, less work has focused on students’ development of knowledge and skills for participation in computational modeling practices. A first step in understanding how these practices develop is examining students’ activity in the context of computational modeling environments with attention to the productive moves they make. These moves can provide insight into the knowledge they bring to their learning, which may be foundational to the development of more sophisticated engagement in computational modeling practices. This paper presents empirical results of an investigation of the knowledge one student brings to her interaction with a computational modeling microworld as she models the spread of disease.

  PublisherOA PDFDOI

Recent grants

• Enabling Modeling and Simulation-Based Science in the Classroom: Integrating Agent-Based Models, Real World Sensing and Collaborative networks

  NSF · $2.8M · 2010–2015

• HCC: Advancing the Science of Agent-based Modeling Through Frameworks, Tools, and Pedagogies

  NSF · $448k · 2007–2010

• EAGER: A Low-Cost Integrated Agent-based Modeling and Physical Computing Platform

  NSF · $144k · 2014–2018

• A Whole-School Model for Integrating Computational Thinking in High School Science and Mathematics

  NSF · $2.5M · 2016–2020

• ITR: Collaborative Research: Procedural Modeling of Urban Activity and Form

  NSF · $1.2M · 2003–2008

Frequent coauthors

• Michael Horn

  Northwestern University

  39 shared

• David Weintrop

  University of Maryland, College Park

  28 shared

• William Rand

  25 shared

• Paulo Blikstein

  Columbia University

  23 shared

• Corey Brady

  19 shared

• Sharona T. Levy

  University of Haifa

  17 shared

• Dor Abrahamson

  University of California, Berkeley

  16 shared

• Arthur Hjorth

  Aarhus University

  15 shared

Education

• M.A, Mathematics

  Harvard University

  1980

• M.A., Mathematics

  Brandeis University

  1977

• B.A., Mathematics & Philosophy

  Brandeis University

  1977

Awards & honors

• Yidan Prize (2025)