Data Science Lunch and Learn Series
The Data Science Lunch & Learn series was launched by the Academy of Data Science in Fall 2023 to foster connections and collaborations across departments in the data and decision sciences community. Held biweekly, each session in the series features a speaker (either internal or external) who presents on a topic relevant to the field of data science.
Select presentations will be recorded and available for viewing after the event.
Model Interpretability and Explainability
Oliver Schabenberger, Professor of Practice, Academy of Data Science
Model interpretability has been on a collision course with transparency for some time. Measures to improve model performance tend to reduce interpretability, models are increasingly complex, and their consumers might lack understanding or awareness. Many models we work with fall somewhere between being intrinsically understandable and total black boxes.
This situation has led to many approaches to explain what drives a particular model. This presentation distinguishes between interpretability and explainability, introduces model-agnostic local and global methods for explanation, and discusses their pros and cons.
Best Practices for Client Projects
Bill Clark, Adjunct Professor of Business Administration, Duke University Fuqua School of Business
Clark, who has a broad background in consulting and advising in a range of industries, will give an overview of the programs in Duke’s Fuqua School of Business and Pratt School of Engineering and discuss how to keep students and clients on track in experiential learning. He will share best practices for client projects and provide an overview of lessons learned with client-facing projects in capstones, consulting practicums, and core/elective courses using examples such as People Analytics.
MotherDuck: Data Analytics in the Post Big-Data Era
Jordan Tigani, CEO, MotherDuck
Tigani, whose prior roles include Chief Product Officer at SingleStore and product leader, engineering leader and founding engineer on Google BigQuery, will introduce MotherDuck, a startup building a serverless analytics platform based on DuckDB – an open source analytics engine that makes it easy to operate over data locally.
While the modern data analytics toolkit is full of tools that presume that they’re dealing with giant amounts of data, most people don’t have huge data sets – or use just a small subset of their data at once. Enter DuckDB, which is perfect for the sizes of data people actually have.
Complementarity Neglect: When People Don't Recognize Advantages of Human-AI Collaboration
Meng Zhu, Professor of Marketing, Pamplin School of Business
Artificial intelligence (AI) has been shown as an advantageous collaborator in domains that have profound social and personal consequences. Not surprisingly, human-AI collaboration is particularly helpful when the partners involved complement each other, as there are more opportunities for such partners to correct each other’s mistakes.
However, if humans are unable to recognize AI-human complementarity and choose disadvantageous cooperation partners with overlapping mistakes, this complementarity neglect potentially thwarts the advantages of human-AI collaborations.
Generative AI at Virginia Tech
Dale Pike, Associate Vice Provost for Technology-Enhanced Learning, Virginia Tech
Generative artificial intelligence (AI) has dominated Pike’s time and attention of late – from leading workshops and faculty communities of practice to actively exploring the implications of this potentially transformative (and disruptive) technology.
Pike will provide a brief overview of generative AI, explaining what it is, how it works, and why it is becoming increasingly important. The presentation will also delve into some of the implications of generative AI for data scientists, including potential use cases as well as potential challenges and ethical considerations.
Center for Ecosystem Forecasting
Quinn Thomas, Associate Professor of Forest Resources and Environmental Conservation (CNRE)/Biological Sciences (COS), Data Science Faculty Fellow, Virginia Tech
In this discussion, Thomas will introduce the Center for Ecosystem Forecasting along with its data science applications, technology, and education.
The Center for Ecosystem Forecasting is a newly established research-focused center at Virginia Tech, co-directed by Thomas and Cayelan Carey, professor of biological sciences. The vision for the Center is to be an international leader in forecasting research, training, and applications by integrating predictive modeling, cyber infrastructure, and decision support to understand and manage ecosystems across the globe.
The Transformative Role of AI in Retail
Lori Schafer, CEO, Digital Wave Technology
Digital Wave Technology is a software solutions company that transforms retail & CPG business processes through analytics including AI, generative AI, workflow, and automation. During this Lunch & Learn, Schafer will focus on the practical uses and advantages of AI, including generative AI, in the retail and consumer goods sectors.
This discussion will include a live demonstration of solutions and Digital Wave's data science lab, showcasing how its enterprise solutions integrate data science to facilitate collaboration between data scientists, developers, and business users.