Network Science and Education 2023

KEYNOTE SPEAKER:

Prof. Cesar Hidalgo, University of Toulouse, France.

How time, technology, and language impact collective memory and attention

SLIDES AVAILABLE HERE.

Abstract: From writing to the web, humans have used communication technologies to enhance our collective memory. Yet, much of what was once popular is now forgotten. In this talk, I will present research exploring the roles played by time, language, and technologies on the dynamics of collective memory and attention. Using data on the attention received by biographies, scientific papers, songs, and movies, we will explore the universal decay of collective memory, the role played by languages in global fame, and the biases in attention and collective memory introduced by changes in technology.

Suggested readings:

Candia, Cristian, et al. "The universal decay of collective memory and attention." Nature human behaviour  (2019)

Ronen, Shahar, et al. "Links that speak: The global language network and its association with global fame." Proceedings of the National Academy of Sciences  (2014)

Jara-Figueroa, C., Amy Z. Yu, and César A. Hidalgo. "How the medium shapes the message: Printing and the rise of the arts and sciences." PloS one (2019)

Yu, Amy Zhao, et al. "Pantheon 1.0, a manually verified dataset of globally famous biographies." Scientific data 3.1 (2016)

Bio: César A. Hidalgo is a Chilean-Spanish-American scholar known for his contributions to economic complexity, data visualization, and applied artificial intelligence.  Hidalgo leads the Center for Collective Learning at the Artificial and Natural Intelligence Institute (ANITI) of the University of Toulouse. He is also an Honorary Professor at the University of Manchester and a Visiting Professor at Harvard's School of Engineering and Applied Sciences. Between 2010 and 2019 Hidalgo led MIT’s Collective Learning group. Prior to working at MIT, Hidalgo was a research fellow at Harvard’s Kennedy School of Government. Hidalgo is also a founder of Datawheel, an award-winning company specializing in the creation of data distribution and visualization systems. He holds a Ph.D. in Physics from the University of Notre Dame and a Bachelor's in Physics from Universidad Católica de Chile. His contributions have been recognized with numerous awards, including the 2018 Lagrange Prize and three Webby Awards. He is also the author of three books: Why Information Grows (Basic Books, 2015),  The Atlas of Economic Complexity (MIT Press, 2014), and How Humans Judge Machines (MIT Press, 2021).

INVITED TALK:

Network Science Education in a Global Manufacturing Company 

Toshihiro Tanizawa, Toyota

As a network science researcher in a global manufacturing company, recently moved from academia, I realized that the range of applicability of network science is much more enormous than I ever thought before. Actually it seems to span almost all activities in the company, such as manufacturing processes, quality management of products, flows of decision making or paper works, social networking within the company, let alone R&D activities on Huge Data analyses aiming to establish new mobility business models. I believe that the insights given from network science can be readily useful and beneficial for non-specialists such as general workers in various branches and factories. In this informal talk, I like to share my own humble perspective on these issues to ignite a discussion on "Network Science Education in Industry."

CONTRIBUTED TALKS:

Cognitive network science reveals bias in GPT-3, ChatGPT, and GPT-4, mirroring math anxiety in high-school students 

Katherine Abramski, University of Pisa

Large language models are becoming increasingly integrated into our lives. Hence, it is important to understand the biases present in their outputs in order to avoid perpetuating harmful stereotypes, which originate in our own flawed ways of thinking [1]. This challenge requires developing new benchmarks and methods for quantifying affective and semantic biases [2], keeping in mind that LLMs act as psycho-social mirrors that reflect the views and tendencies that are prevalent in society. One such tendency that has harmful negative effects is the global phenomenon of anxiety toward math and STEM subjects [3]. Here, we investigate perceptions of math and STEM fields provided by cutting-edge language models, namely GPT-3, ChatGPT, and GPT-4, by applying an approach from network science and cognitive psychology. Specifically, we use behavioral forma mentis networks (BFMNs) [4] to understand how these LLMs frame math and STEM disciplines in relation to other concepts. We use data obtained by probing the three LLMs in a language generation task that has previously been applied to humans [4]. Our findings indicate that LLMs have an overall negative perception of math and STEM fields, with math being perceived most negatively. We observe significant differences across the three LLMs (Figure 1). We observe that newer versions, i.e. GPT-4, produce richer, more complex perceptions as well as less negative perceptions compared to older versions and high-school students, N = 159 (Table 1). These findings suggest that advances in the architecture of LLMs may lead to increasingly less biased models that could even perhaps someday aid in reducing harmful stereotypes in society rather than perpetuating them. 

1. Ferrara, E. Should ChatGPT be Biased? Challenges and Risks of Bias in Large Language Models 2023. arXiv: 2304.03738 [cs.CY]. 

2. Hagendorff, T. Machine Psychology: Investigating Emergent Capabilities and Behavior in Large Language Models Using Psychological Methods 2023. arXiv: 2303.13988 [cs.CL]. 

3. Foley, A. E. et al. The math anxiety-performance link: A global phenomenon. Current directions in psychological science 26, 52–58 (2017). 

4. Stella, M., De Nigris, S., Aloric, A. & Siew, C. S. Forma mentis networks quantify crucial differences in STEM perception between students and experts. PloS one 14, e0222870 (2019). 

Unveiling Literary Connections: Integrating Character Networks in English Classes to Encourage Multidimensional Analysis and Engage Diverse Learners 

Tess Stepakoff, The University of Alabama - SLIDES HERE.

The study of networks is often associated with sciences and mathematics, but has strong, longstanding ties with the humanities. While coding and network science are gradually finding their place in K-12 education, humanities-based social networks, particularly uncoded ones, are often overlooked. By incorporating character networks into English classes, we can encourage students to adopt various approaches to understanding literature beyond traditional close reading. Creating a character network can offer a fresh lens through which students can question their comprehension of characters and their interactions. This approach challenges their assumptions and deepens their understanding by asking students to determine what qualifications are required for each character’s inclusion as a node and their interaction as a link within the network. Character networks also uncover hidden patterns that readers may not be able to see while close reading, which could then be compared to other relationships in the same text, other works by the same author, or other texts within the genre. The creation of networks also engages different types of students who might not necessarily excel at close reading, such as those who have a stronger inclination toward math, science, or visual learning. Networks can provide students with a deeper, more interactive understanding of literary texts when integrated into an educational setting, opening doors to different research methods and alternate perspectives of what is traditionally taught. 

Serious Games for Independent Learning in K-12 Education: A Network Overview 

Nidia Lopez Flores, Anna Sigridur Islind, and Marıa Oskarsdottir

SLIDES HERE.

Serious games have potential benefits for learning and engagement and are a growing area of academic research. These are digital games that are used for purposes other than entertainment, such as teaching, training, and improving problem-solving, decision-making and communication skills. In this study, we use a network science approach to analyse game preference patterns and performance in an open educational platform. The platform (Fig. 1) is designed for children between 5 and 13 years old from Spanish and English speaking countries. The platform offers several games, on varied subjects and game mechanics. In this study, we analyse the interactions at a game and user level. At a game level, networks are created by connecting games that are played during the same game session. At a user level, students playing games with similar subject or mechanic are connected. In both cases, we investigate the differences in game preferences between students belonging to different age groups, as well as the relationship between these preferences, the performance achieved, and the time spent playing. In this presentation, we include the exploratory analysis and preliminary results for both game and user level. Furthermore, we will provide insights on patterns and differences between age groups, countries, and game subjects.