Research
Our ongoing research and the students responsible for it. Every project has a direct contact for partnerships and inquiries. To learn what each group studies, see About.
Ongoing research
If a project is of interest to your company, laboratory or institution, use the button on that project. The message reaches our team, who will forward it to the students responsible.
AI Safety
In progressHow UFSCar faculty members perceive the advance of artificial intelligence, its risks and safety practices in the biological, forestry and applied sciences.
AI Safety
In progressA joint study with FEA.dev, at the University of São Paulo, of two behaviors of language models in Portuguese. The first is positional bias: whether a model favors an answer only because of where it sits in a multiple-choice question. The second is sycophancy: whether a model that has pointed out a user’s poor financial decision gives in when the user insists on it. Open-ended answers are graded by an LLM judge with our own rubric, validated against grading by two people.
FEA.dev
Academic
In progressSocial analysis applied to the development of artificial intelligence: the project examines whether AI models reproduce patterns of inequality related to gender, race and political minorities.
AI Safety
In progressAn in-person meeting at UFSCar to discuss the Hugging Face case and the questions it raises about the safety and governance of artificial intelligence.
Academic
In progressA Closed-Economy DSGE Model for Macroeconomic Analysis and Out-of-Sample Forecast Evaluation
A closed-economy dynamic stochastic general equilibrium (DSGE) model of the Brazilian economy, used to analyze current macroeconomic conditions and evaluate out-of-sample forecasts. The model is being calibrated to forecast the data set well and may later support policy simulations.
Academic
In progressAn Out-of-Sample Assessment of New Keynesian Microfoundations and Heterodox Macroeconomic Mechanisms During the 2014–2016 Recession
An agent-based model of the Brazilian business cycle, evaluated out of sample on the 2014–2016 recession to compare New Keynesian microfoundations with heterodox macroeconomic mechanisms.
Contact
Write to the team with the topic, the objective and the timeline.
