Artificial Intelligence and Physics Education – ongoing projects at Imperial College London

Artificial Intelligence and Physics Education – ongoing projects at Imperial College London

Date: April 14 2025 - at 11.15 AM (Italian time)

Where: Aula P1A, Complesso Paolotti

Imperial College London has invested in a variety of projects to explore the use of generative artificial intelligence (GAI) and machine learning (ML) in higher education The main project which I will discuss aims to test the effectiveness of Large Language Models (LLMs) in assisting large-scale qualitative research and consequently whether workflows incorporating LLMs can provide useful and timely feedback to instructors on the impact of their teaching. The focus of the work is on student use of scientific argumentation in lab reports. We have constructed a codebook for scientific argumentation to identify elements of argumentation, the relationships between those elements, and their veracity. Through building a training dataset of over 250 lab reports from students in their first- and second-year lab courses, we aim to analyse the whole set of 1899 lab reports from two cohorts of students by utilizing fine-tuning of open-source LLMs to automate the coding process (Fussell et al., 2025). The specific purpose here is to test how changes to teaching between the two cohorts affected students’ scientific argumentation skills, and, therefore, help us to evaluate the impact of the changes.

In addition to this, I will provide an overview of the projects that are currently being undertaken across Imperial related to AI and education. These institutionally funded projects include: (1) the use of GAI in providing feedback to students through the in-house problem sheet web-interface Lambda Feedback; (2) the use of LLMs to analyse and provide in-the-moment feedback to students on the structural elements of their lab reports; (3) using ML to identify at-risk students through analysis of interaction data from Lambda feedback; and (4) collecting data on students’ views about how instructors use GAI in their teaching.

conference speakers

Michael Fox

Michael has worked on a wide range of projects in physics education research, from workforce development for the quantum industry through to analysis of the process of curriculum and culture change in a physics department. His core interest is in what and how students learn in physics teaching laboratories. This has been informed by his own experience as a student in undergraduate teaching labs through to his PhD research on analysing data on plasma turbulence in nuclear fusion reactors. How to teach experimental physics effectively came to the forefront when he was teaching high-school physics, leading to post-doctoral work on assessing student learning in teaching labs using the E-CLASS and MAPLE surveys during his time in the Lewandowski group in Boulder, Colorado. He has recently been appointed as the head of the teaching laboratories in the Department of Physics at Imperial College London, where he has started to implement evidence-based practices.

https://profiles.imperial.ac.uk/michael.fox

by Admin

PER-Grounded di Faculty Learning Community

PER-Grounded di Faculty Learning Community

Date: December 16 2024 - from 4.30 PM to 5.30 PM (Italian time)

Where: Aula P4C, Complesso Paolotti

The primary goal of physics teaching is to educate human beings to science and its nature. However, as technology continues to evolve, artificial intelligence (AI) is gaining prominence not only in technical and scientific domains but also in education. Recently, the rise of AI has sparked both concern and enthusiasm among educators, affording new opportunities and challenges for teaching and learning.
As we navigate this fast-evolving landscape, several questions arise: What competences, practices, and aspects of human development do we aim to foster, and how does the availability of AI-based systems impact their teaching and learning? How should we, as educators, frame AI-based systems and our – and our students’ – interactions with them? Is it even possible to “collaborate” with AI in meaningful ways? The panelists, all experts in physics education research, will bring a variety of expertise and perspectives, from technical considerations to ethical implications.

conference speakers

Magdalena Kersting

Department of Science Education, University of Copenhagen

Olivia Levrini

Dipartimento di Fisica e Astronomia “Augusto Righi”, University of Bologna

Giulia Polverini

Department of Physics and Astronomy, Uppsala University

Christopher Robin Samuelsson

Department of Physics and Astronomy, Uppsala University

by Admin

Modelling of brain stimulation to unveil signal propagation and network dynamics

Modelling of brain stimulation to unveil signal propagation and network dynamics

Date: December 13 2023 - at 2.45 PM (Italian time)

Where: Aula Magna del VIMM, Via Giuseppe Orus, 2, 35129 Padova PD

The human brain comprises distinct resting-state networks (RSNs) characterized by spontaneous activity patterns. Despite this highly structured functional pattern, its laws of motion and principles of organization have proven challenging to understand with currently available measurement techniques. In such epistemic circumstances, an extremely elegant modus operandi to investigate brain complexity with high spatial and temporal resolution entails the administration of precise and synchronized external stimulation, followed by a meticulous examination of the resulting induced propagation dynamics that emerge in response to these perturbations. In this framework, a combination of empirical stimulus-evoked data analyses and whole-brain, connectome-based neurophysiological modelling provide an elegant scaffold to investigate questions around the physiological basis and spatiotemporal network dynamics of RSNs’ activity. Deciphering this evoked propagation pattern is essential for a comprehensive understanding of the brain’s response to stimulation and therefore for personalized and targeted interventions, with potential applications ranging from therapeutic treatments to cognitive enhancement. Dr. Momi will also give a more technical lectures about the mathematical modelling and computational pipelines for the students of Physics of Data on Thursday 14 December at 2.30 pm, in P2B room, Paolotti building. This lecture are open also for PNC Ph.D. students.

conference speakers

Davide Momi

Davide is a second year Post-Doctoral Research Fellow at Krembil Centre for Neuroinformatics (Toronto). He completed his Ph.D. at the Department of Neuroscience, Imaging and Clinical Sciences at the University of Chieti. As part of his PhD, Davide attended a period abroad at the Martinos Center for Biomedical Imaging in Boston. Prior to his doctoral studies, he obtained a Master’s degree in “Neurosciences and Neuro-Psychological Rehabilitation” from the University of Bologna and a Bachelor’s in “Psychology” from the University of Perugia. Davide has experience with multimodal neuroimaging and electrophysiological data, brain stimulation, quantitative structural MRI assessment, machine learning, simulations of macroscale brain dynamics. His research is mainly focused on personalized simulations of brain dynamics to provide a better understanding of neurological and psychiatric disorders.

by Admin

The transformers revolution: how an algorithm shaped the current AI landscape and society

The transformers revolution: how an algorithm shaped the current AI landscape and society

Date: November 8 2023 - at 4.30 PM (Italian time)

Where: Aula Rostagni, Physics and Astronomy Dept, Via Paolotti 9, 35131, Padova

The original Transformers architecture by Vaswani et al. dates only to 2017. In a very short time, it revolutionized our understanding of the capabilities of Artificial Intelligence. This session explores how this idea revolutionized natural language processing, enabling machines to comprehend and generate human-like text. From their inception to contemporary applications, we elucidate the profound influence of Transformers in diverse fields, including chatbots, and multimodal content generation. We investigate the societal effects, scrutinizing the ethical implications, biases, and job market transformations induced by this technology.

conference speakers

Carlo Nicolini

Carlo Nicolini, PhD is currently Senior Scientist at Ipazia SpA, a company of artificial intelligence services for industry and banks. After completing a PhD in applied physics focused on community detection in brain fMRI networks, he pursued postdoctoral studies exploring the statistical physics of complex systems. His research spanned various fields, including neuroscience and finance. Transitioning to industry, he joined a leading data science consulting firm in Italy, where he served as a senior analyst in the financial services sector. He is the creator of scikit-portfolio, a library for financial portfolio optimization. Currently, his research interests lie at the crossroads of deep learning, statistical physics, and finance.

by Admin

From the Transformer architecture to ChatGPT

From the Transformer architecture to ChatGPT

Date: October 17, 2023 - at 4.30 PM (Italian time)

Where: Aula Voci, Physics and Astronomy Department, Via Francesco Marzolo 8, 35131, Padova

Large Language Models (LLMs) have risen to prominence in recent years thanks to their ability to accurately model human Natural Language, solve reasoning tasks and effectively assist humans in several tasks via a chat interface.
In my talk I will introduce the core neural network architecture behind LLMs, the Transformer. I will further present the main line of research and insights that lead to the successful scaling of Transformer Language Models up to hundreds of billions of parameters. Finally I will relate the main ideas that let us turn a Large Language Model into a useful language assistant such as ChatGPT.

conference speakers

Nicola Dainese

Nicola Dainese graduated in Physics of Data in 2020. He is now PhD Candidate in Computer Science (Deep Learning and Artificial Intelligence) at Aalto University, Finland. His expertises are reinforcement learning and language models.

by Admin

A magnifying glass to elucidate their rise in photovoltaics: atomistic simulations of organic halide perovskites

A magnifying glass to elucidate their rise in photovoltaics: atomistic simulations of organic halide perovskites

Date: June 20, 2023 - at 11.00 AM (Italian time)

Where: Aula P1A, Complesso Paolotti, Via Giambattista Belzoni 7, 35131, Padova

Organic inorganic hybrid halide perovskites (PSCs) have emerged as promising class of materials for solar cells. Thanks to their remarkable optical properties such as high absorption coefficient, tunable bandgap, high charge carrier mobility and low exciton binding energy, PSCs facilitate high power conversion efficiencies (PCE). Indeed, the PCEs of PSCs have rapidly risen since 2009 from 3.8% to a certified 26.2% in 2022. Most of these rapid advances are due to changes in the preparation conditions, however the connection to the underlying processes at the atomic level is basically missing.  Computer simulation can in principle elucidate the atomistic mechanisms behind these phenomena, but the systems involved are highly complex involving multiple species and high activation barriers. Our approach entails the development and application of theoretical and multiscale computational modeling spanning the electronic and atomistic length scales. Classical molecular dynamics (MD) is embedded with ab-initio MD and density functional theory simulations to have an outright description of the physics behind the mechanisms responsible for long-term stability and improvement of the optical properties for PSCs.

conference speakers

Virginia Carnevali

Virginia Carnevali is a scientist in Ursula Röthlisberger’s group at the Institute of Chemical Sciences and Engineering Ecole Polytechnique Federale de Lausanne EPFL, Switzerland. She has research experience in the theory and simulations of structural, electronic, magnetic, optical, transport, and thermal properties of 2D materials, topological materials, thermoelectric materials, and solar cells. Virginia has also developed an interest in researching useful applications of quantum computing to solved condensed matter problems. Her research has focused on understanding the physics and crucial mechanisms that make a material attractive for renewable energy applications. She recently demonstrated – for the first time –that coupling of rotating lone pairs and the vibrational motion is an effective mechanism to achieve ultralow thermal conductivity in crystalline materials. Virginia is currently studying perovskite-based solar cells, one of the most promising photovoltaic solutions because of their efficient solar-power conversion, low cost and simple manufacturing process. Her efforts are aimed at improving the optical performance of this class of materials by seeking effective techniques for defect passivation, phase stabilization, and improved of charge carrier transport at the interfaces.

by Admin

Modelling and quantifying robustness of interconnected systems: at the crossroad of network science and AI

Modelling and quantifying robustness of interconnected systems: at the crossroad of network science and AI

Date: June 16, 2023 - at 11.00 AM (Italian time)

Where: Aula B, Via Paolotti 9, 35131, Padova

Over the last years, network theory has been revealed to be a convenient and flexible framework to characterize and model the structure and dynamics of complex systems. A wide range of disciplines, from socio-technical to natural sciences, face common questions when their objects of study can be represented as networks. One of them is the robustness and resilience of interconnected systems to perturbations and malfunctions. In this seminar, I will provide a general and comprehensive overview of the main strands of research on network dismantling. Emphasis will be put on how to quantify robustness by combining mathematical techniques and AI-assisted tools.

conference speakers

Oriol Artime

Oriol Artime is a “Margarita Salas” postdoctoral researcher at the University of Barcelona, Spain. Prior to this, he has been a postdoctoral researcher at the Bruno Kessler Foundation, and he holds a PhD from the Institute for Cross-Displinary Physcs and Complex Systems – IFISC. His research primarily deals with topics at the interface between statistical physics and complex systems, with a strong focus on complex network robustness and resilience.

by Admin

Before and after chatGPT: chances, changes, and opportunities of large language models in research

Before and after chatGPT: chances, changes, and opportunities of large language models in research

Date: May 23, 2023 - at 16.30 (Italian time)

Where: Aula B, Via Paolotti 9, 35131, Padova

We have all come to realize recently that there is a before and after ChatGPT, the algorithms created for the new large “language model”. This is especially true for the fields of AI, CS, Data Science, and ML, but in other ways also for social sciences, humanities, and industry. Many of these changes are relevant to the work we do, and LLMs will change it forever.

In his talk, Dirk will explain the revolutionary ideas behind ChatGPT and their history. He will also give his take on how these new tools change possibilities for research and society, and what to be aware of when using them. He will discuss the worries and concerns as well as the new possibilities emerging every day.

conference speakers

Dirk Hovy

Dirk (left image) is an associate professor at Bocconi’s department of computing sciences in Milan, and director of the Data and Marketing Insight research unit at the Bocconi Center for Data Science and Analytics.

Andrea Pin

Associate Professor of Comparative Public Law, University of Padova.

Read full bio

by Admin

Threats to Sustainability under Climate and Societal Change

Threats to Sustainability under Climate and Societal Change

Date: May 12, 2023 - at 11.15 AM (Italian time)

Where: Aula Rostagni, Via Paolotti 9, 35131, Padova

The increasing water demand by human societies raises concerns about the extent to which it is possible to feed the world with the limited freshwater resources of the planet. The emergent competition for water between human uses and environmental needs limits the development of suitable water security scenarios for a sustainable future. Human appropriation of water resources is for large part instrumental to the enjoyment of human rights to food. To what extent can such rights be reconciled with other human needs as well as the needs of Nature? Here we show how humanity is placing unprecedented pressure on the global agricultural system and its water resources while reshaping the patterns of water dependency through teleconnections between consumers and production areas. Through a suite of ecohydrological and socio-environmental analyses, we evaluate the biophysical and social justice limits to the sustainable use of water resources through a variety of perspectives accounting for hydrologic constraints, environmental needs, societal resilience, food security, and globalization.

conference speaker

Paolo D’Odorico

Paolo D’Odorico is Professor of Hydrology at the Department of Environmental Science, Policy, and Management at UC Berkeley. He has research expertise in Ecohydrology, Surface Hydrology, Ecosystem Ecology, Aeolian Processes, Desertification, Stochastic Nonlinear Environmental Dynamics and Water and Food Security, with about 300 publications in these topics. Prof. D’Odorico’ research focuses on the role of hydrological processes in the functioning of terrestrial ecosystems and his work has highlighted important nonlinearities in the coupling between soil moisture dynamics and plant water stress, biogeochemical cycling, land-atmosphere interactions, plant community composition, and soil susceptibility to wind erosion. Prof. D’Odorico is currently investigating the globalization of water through virtual water trade and international land investments, and its impact on water equity, societal resilience, environmental stewardship, and food security. Prof. D’Odorico has been awarded the Hydrological Sciences Medal (American Meteorological Society), the Paul Witherspoon Lecture Award (American Geophysical Union), the Sustainability Science Award (Ecological Society of America). He is a Fellow of the American Geophysical Union, American Association for Advancement of Science,  and The John Simon Guggenheim Memorial Foundation.

by Admin

Modelling weather & climate: challenges and applications.

Modelling weather & climate: challenges and applications.

Date: March 10, 2023 - at 02.30 PM (Italian time)

Where: Aula Rostagni, Via Paolotti 9, 35131, Padova

Understanding and predicting the weather and climate variability is fascinating as much as challenging. Here we highlight some of the key aspects associated with the study of this subject, and describe the hierarchy of models that are currently employed to disentangle such a complex problem. The potential applications will also be discussed, with a specific focus on the energy sector.

conference speaker

Dr. Giacomo Masato, Ph.D

Giacomo graduated in physics at the University of Padua in 2006, with a thesis on the Rossby waves vertical propagation and their role in the onset of the Sudden Stratospheric Warmings. He then moved to the UK where he continued his studies at the Department of Meteorology (University of Reading). He spent many years studying the large-scale circulation at the mid-latitudes and he collaborated, among others, with Prof Sir Brian Hoskins, one of the world leading figures in atmospheric physics. In 2015 he moved to the private sector, first in the UK and subsequently in Italy. He now works for Illumia, a retailer company within the energy space, and leads the analytical department of the firm, where his team develops statistical and fundamental models to predict the weather and energy behaviour.

by Admin