Virtually join UC Berkeley colleagues for a talk by Dr. Anqi Liu (UC Berkeley) for a seminar entitled “A Conservative Extrapolation Approach to Trustworthy AI”.
Dr. Anqi Liu
Asst. Professor of Computer Science
Friday, March 31st, 12:00–1:15 PM
3505 N. Charles Street
— http://zoom.us/j/6019060976
Abstract:
The unprecedented prediction accuracy of modern machine learning beckons for application in a wide range of real-world situations, including healthcare, education, and hiring. A key challenge is the difficulty to collect data from diverse enough populations. It causes a well-known problem called distribution shift, which means the test cases are not well-represented by the training data and usually leads to inequivalent performance in different subgroups. Overconfident errors in the under-represented group brings harm to social trust in AI-based services. In these cases, we must go beyond the conventional learning paradigm of maximizing average prediction accuracy with generalization guarantees that rely on strong distributional relationships between training and test examples. In this talk, I will describe a distributionally robust learning framework that offers rigorous guarantees under data distribution shift. This framework yields appropriately conservative extrapolations and can be used for producing more equitable prediction results among subgroups. I will also introduce a survey of other real-world applications that would benefit from this framework for future work.
A position is available for a Postdoctoral Research Fellow in the lab of Dr. Daphne Hernandez to study food insecurity. The position offers an opportunity to work on a NIH-funded project with the end goal of creating a culturally appropriate food insecurity scale for Spanish-speaking immigrant parents that have low or limited English proficiency. The Fellow is expected to have studied food insecurity and engaged in qualitative research during graduate school (cognitive interviewing preferred), in addition to having experienced working in resource limited communities with low-income Hispanic families where parents speak primarily Spanish.
The Fellow will also have the opportunity to published from two other community-based projects that focused on food insecurity that used quantitative and qualitative methods. In addition, there will be other publication opportunities based on existing secondary data sets as lead author and co-author. The Fellow will further their research and training agenda through formal mentorship and professional development activities, such as presenting at local and national conferences and participating in seminars through the CSON Postdoctoral Training Program. Twenty percent time can be allocated to publishing dissertation papers. The opportunity to mentor undergraduate students on research projects (implementation of projects, development of posters and manuscripts) is also available.
U.C. Berkeley’s Urban Displacement Project (UDP – urbandisplacement.org) and Eviction Research Network (ERN – evictionresearch.net) at the Institute of Governmental Studies, Berkeley Institute of Data Science (BIDS), and the Department of Sociology has an opening for a 2-year post-doctoral position beginning June 2023 on evaluating eviction outcomes during the pandemic and teaching computational social science. The successful candidate will be appointed part-time as a postdoc, with approximately two-thirds time devoted to research over the course of two-years, and part-time as a lecturer.
The eviction project is a HUD funded grant that investigates how the U.S. Emergency Rental Assistance Program (ERAP) impacted eviction outcomes using a novel longitudinal eviction database. This project will create one of the first longitudinal eviction databases using large public and private datasets, linking records across multiple datasets with personal identifiable information, and utilizing Bayesian and machine learning models to predict neighborhood and household drivers and outcomes of eviction. The purpose of this research is to identify the racial and gender disparities in eviction, displacement, and economic gaps that perpetuate unequal outcomes for different groups and inform policy. The candidate will co-author publications with the PI and team as well as be able to use this database for future research. The remaining candidate time will be devoted to teaching computational social science to graduate students in the social sciences and related fields in a two-semester course sequence that covers reproducibility and transparency, applied machine learning, natural language processing, and causal inference within the Department of Sociology.
The candidate will be part of the Sociology, Urban Planning, and BIDS community, working closely with Dr. Tim Thomas (UDP) and Professor David Harding (Sociology), and will play a vital role in training students in the growing field of computational social science. The team will publish evidence-based research in academic journals and use these findings to directly inform national, state, and local government agencies on how to reduce eviction and housing inequity.
PDB supports research, data collection, and training in demography, behavioral and social science research on reproductive health, and population health.
PDB seeks a program officer with expertise in demography, the social determinants of health, the development and maintenance of longitudinal cohorts, and the collection and analysis of population representative samples. The incumbent will develop a portfolio of interdisciplinary science that supports research, training, and career development in sociology and economics as they relate to health, development, and productivity. The ideal candidate will have significant research experience in causal inference, interactions between macro- and micro-levels of analysis, and the impacts of policies and institutional structures.
This is an exciting time to join NICHD as we implement our strategic plan to guide and advance research in the near- and long-term future. Through this implementation, the successful candidate will be able to shape scientific research opportunities that will directly affect the field, biomedical science, and public health for years to come.
The Associate Dean for Equity, Justice, and Inclusion in the College of Arts and Sciences is seeking to secure the services of a Research Assistant for Spring and Summer quarters to help with the planning, implementation, and analysis of a DEI climate survey as well as putting together a central DEI repository. The proposed .5 FTE appointment is for the entire UW Summer Quarter and Fall Quarters. All salary and benefits will be consistent with the current UAW 4121 Collective Bargaining Agreement.
This is an exciting opportunity to work in cutting-edge research focused on matters of DEI. The result of this work will be the establishment of a series of benchmarks from which the College of Arts and Sciences will be able to evaluate its progress in achieving DEI goals that further the mission of the College.
The Office of Financial Management is seeking a Senior Data Analyst. Information on the position can be found here!
The ARPA-H Dash to Accelerate Health Outcomes, or “ARPA-H Dash,” launched on March 15, 2023 to help identify revolutionary evidence-based ideas to transform health. The ARPA-H Dash is a collaborative online competition open to bold thinkers across health, scientific, and technology communities, as well as the general public. The ARPA-H Dash will use a bracket format and online discussion, debate, and voting to narrow submissions to quarterfinalists, semifinalists, finalists, and a champion idea during March and April 2023.
Nominations for the ARPA-H Dash will be accepted through an online portal for review between March 29 and April 7, 2023. Each submission should include a piece of evidence with citation, published after January 1, 2018, a short descriptive title for display in the brackets, identification of which ARPA-H focus area is targeted for transformation, and answers to three questions about the proposed transformation.
For more information click here!
The Bureau of Economic and Business Research (BEBR) at the University of Florida seeks to hire an applied demographer to develop population and household estimates and projections for Florida. They are looking for someone with training in demographic methods at the graduate or advanced undergraduate level, good programming skills in SAS and/or R, and familiarity with Census data. Information for the application can be found here.
The Barcelona Summer School of Demography (BSSD), offers a four-week course in R. The course is divided into four modules -one per week- covering three major strengths of R: statistical and demographic analysis, data visualization and spatial analysis. Each module consists of 20 hours of teaching, combining theoretical lectures and practical exercises.
More information can be found here.
On April 11, EarthLab and The Population Health Initiative will cohost an Open Space-style event to help to support UW researchers in laying the groundwork for new, interdisciplinary partnerships that address key facets of climate change. Attendees of the event will set the agenda for discussion, offering to convene discussions on possible topics or projects where collaboration is sought. The formal program will be followed by a more informal networking lunch.
RSVP required via https://bit.ly/3ZNAmRS