Cecilia Stari

Recent advances in large language models have opened new possibilities for Artificial Intelligence (AI)-assisted grading in higher education. This study explores the use of a GPT-5.4–based system to support the evaluation of introductory physics laboratory reports using a rubric-driven automated workflow implemented in batch mode via API. AI-generated scores were compared with instructor grading, and the feedback was analysed qualitatively across rubric criteria. The results show weak agreement in the ranking of reports and noticeable differences at the level of individual scores. Item-level analysis indicates that the model frequently produced feedback classified as correct application but also generated a non-negligible proportion of reasonable but superficial responses and invalid evaluations. These limitations were mainly associated with restricted access to evidence during text extraction and OCR, particularly for equations, figures, and graphical representations. The findings suggest that batch-based AI grading can provide systematic feedback and help identify recurring patterns in student work, supporting instructors in large-scale courses. However, under the conditions examined, AI-generated scores were not interchangeable with instructor grading and should be understood as a tool to assist, rather than replace, the evaluation process.

We present an experimental study of the Duffing--Holmes oscillator with a double-well potential, implemented as an analog electronic circuit under periodic external forcing. By systematically varying the forcing amplitude and frequency, we characterize the full dynamical landscape of the system through bifurcation diagrams, Poincar\'e maps, and maximum Lyapunov exponent calculations. The observed phenomenology includes period-doubling routes to chaos, periodic windows with multistability, dynamical intermittency, and antiperiodic orbits in which the trajectory recovers the global symmetry of the double-well potential. Multistability-induced discontinuities in the bifurcation diagrams are identified and interpreted as attractor-coexistence signatures arising from the sensitivity of the long-time dynamics to initial conditions, rather than as noise artifacts. These results are synthesized into a high-resolution two-dimensional diagram in parameter space. The close agreement between all experimental diagnostics validates the fidelity of the analog implementation and demonstrates that continuous-time hardware provides a powerful platform for the quantitative study of nonlinear dynamics, free from the discretization artifacts inherent to numerical simulation.

La interpretación de gráficas en cinemática constituye un desafío persistente en la enseñanza de la física. El Proyecto MissionMotion aborda esta dificultad mediante un entorno físico-computacional gamificado que combina sensores de bajo costo, actividad corporal, pensamiento computacional y visualización en tiempo real de gráficos de movimiento. Presentamos el diseño, desarrollo e implementación de la propuesta, con especial énfasis en la fase piloto realizada con estudiantes de educación media en Uruguay. En esta etapa, utilizamos el cuestionario MEEGA+ como instrumento principal para evaluar la experiencia de juego, la usabilidad y la motivación de los participantes. El análisis de los resultados muestra altos niveles de satisfacción, percepción de aprendizaje y compromiso, lo que respalda la viabilidad de la propuesta. Finalmente, proyectamos una evaluación conceptual a gran escala destinada a analizar cómo la propuesta impacta en la comprensión de gráficas cinemáticas, utilizando herramientas de evaluación estandarizadas.

Se presenta una metodología automatizada para detectar habilidades científicas en informes de laboratorio de Física Experimental. La propuesta integra el análisis cualitativo para construir una guía de anotación, fundamentada en las áreas clave de la AAPT y en la Taxonomía Revisada de Bloom, con técnicas de Inteligencia Artificial, en particular de Procesamiento del Lenguaje Natural. El corpus anotado estuvo conformado por 1594 enunciados extraídos de informes de estudiantes y se utilizó en el entrenamiento y validación de modelos de aprendizaje automático supervisado (Multi-layer Perceptron, Regresión Logística y k-Nearest Neighbors) y de grandes modelos de lenguaje (LLM) como Llama3-8B-Instruct y Gemini 2.5 Flash-Lite. Los resultados muestran que los LLM alcanzan un desempeño comparable al de los modelos supervisados en algunas clases, aunque persisten dificultades compartidas con los evaluadores humanos durante la anotación. La metodología propuesta constituye una herramienta para orientar la toma de decisiones pedagógicas fundamentadas en evidencia en los cursos de Física Experimental.

The Mackey–Glass system is a paradigmatic example of a delayed model whose dynamics is particularly complex due to, among other factors, its multistability involving the coexistence of many periodic and chaotic attractors. The prediction of the long-term dynamics is especially challenging in these systems, where the dimensionality is infinite and initial conditions must be specified as a function in a finite time interval. In this paper, we extend the recently proposed basin entropy to randomly sample arbitrarily high-dimensional spaces. By complementing this stochastic approach with the basin fraction of the attractors in the initial conditions space, we can understand the structure of the basins of attraction and how they are intermixed. The results reported here allow us to quantify the predictability giving us an idea about the long-term evolution of trajectories as a function of the initial conditions. The tools employed can result very useful in the study of complex systems of infinite dimension.

We measure the magnetic field produced by a linear octupole formed by four magnets placed in such a way that both their magnetic moments and their quadrupole mo We measure the magnetic field produced by a linear octupole formed by four magnets placed in such a way that both their magnetic moments and their quadrupole mo

The basin entropy is a measure that quantifies, in a system that has two or more attractors, the predictability of a final state, as a function of the initial conditions. While the basin entropy has been demonstrated on a variety of multistable dynamical systems, to the best of our knowledge, it has not yet been tested in systems with a time delay, whose phase space is infinite dimensional because the initial conditions are functions defined in a time interval [−τ,0], where τ is the delay time. Here, we consider a simple time-delayed system consisting of a bistable system with a linear delayed feedback term. We show that the basin entropy captures relevant properties of the basins of attraction of the two coexisting attractors. Moreover, we show that the basin entropy can give an indication of the proximity of a Hopf bifurcation, but fails to capture the proximity of a pitchfork bifurcation. The Hopf bifurcation is detected because before the fixed points become unstable, a oscillatory, limit-cycle behavior appears that coexists with the fixed points. The new limit cycle modifies the structure of the basins of attraction, and this change is captured by basin entropy that reaches a maximum before the Hopf bifurcation. In contrast, the pitchfork bifurcation is not detected because the basins of attraction do not change as the bifurcation is approached. Our results suggest that the basin entropy can yield useful insights into the long-term predictability of time-delayed systems, which often have coexisting attractors.

This article presents a novel method for studying RLC series circuits using two smartphones, one used as a signal generator and the other as an oscilloscope. We measure the voltage at the external resistor as a function of frequency when subjected to a sinusoidal electromotive force. The experimental results demonstrate a remarkable agreement with the theoretical curve for the voltage at the resistor and the resonance frequency, validating the accuracy of the smartphone-based setup. The experiment fills the gap in educational materials related to electrical circuits and provides a portable alternative to traditional, expensive laboratory equipment.