Dear All,
This is a reminder for the upcoming guest lecture in the field of Philosophy of Science, featuring Professor Jun Otsuka from Kyoto University.
We are pleased to announce that in the winter of 2024, Seoul National University will host a guest lecture series presented by international scholar Jun Otsuka, organized by the Institute of Science Data Innovation and the Institute of Philosophy and Thought.
The theme of the lecture is “Thinking About Statistics in the Era of AI,” and the lectures will take place over two days, December 17 and 18. This is an opportunity to deeply explore the philosophical foundations of statistics with Professor Otsuka. We look forward to your participation and interest.
To attend, please register via the link below:
[https://forms.gle/yDAHU6z69pwYqAfj7]
Lecture Theme: Thinking About Statistics in the Era of AI
Speaker: Professor Jun Otsuka (Kyoto University)
Moderator: Professor Hyun-Deug Cheon (Seoul National University)
Program
Lecture 1: Thinking About Statistics I: Bridging the gap between statistics and epistemology
- Date: Tuesday, December 17, 2024, 4:00 PM–6:00 PM
- Venue: Room 309, Sinyang Academic Information Center [Gwanak 4-Dong], Seoul National University
Summary:
In this lecture series, we will explore the philosophical foundations of statistics. The first session focuses on the epistemological characteristics of Bayesian and frequentist statistics. Inferential statistics makes inductive inferences by modeling the “uniformity of nature” underlying the data through statistical models and estimating their parameters. Bayesian and frequentist approaches represent distinct epistemologies, each espousing a different conception as to what counts as “justified inference.” Specifically, Bayesian statistics is characterized by an internalist epistemology, justifying the posterior distribution as a derivation from prior beliefs, whereas frequentist statistics aligns with an externalist epistemology, justifying conclusions (e.g., rejection of the null hypothesis) through reliable methods. By bridging philosophy and statistics, we clarify the nature of each statistical approach and gain insights into issues such as the problem of prior probabilities in Bayesian statistics and the p-value problem in frequentist statistics.
Lecture 2: Thinking About Statistics II: Ontological implications of model selection, machine learning, and causal inference
- Date: Wednesday, December 18, 2024, 4:00 PM–6:00 PM
- Venue: Room 309, Sinyang Academic Information Center [Gwanak 4-Dong], Seoul National University
Summary:
In the second session, we will discuss the philosophical implications of machine learning methods developed since the late 20th century, including model selection, deep learning, and causal inference, with a particular focus on their ontological characteristics. From an ontological perspective—namely how the underlying “uniformity of nature” is modeled—model selection theory, exemplified by AIC (Akaike Information Criterion), and deep learning models can be understood as methods for identifying “real patterns” (à la Dennett) that support effective predictions and extrapolations based on data. In contrast, causal inference, which aims to go beyond mere prediction and evaluate intervention outcomes, introduces an additional ontological layer that models possible worlds and the laws governing transitions between these worlds. These ontological considerations clarify the strengths and challenges of each inductive methodology.
About the Speaker:
Jun Otsuka is a philosopher of science specializing in the philosophy of statistics, machine learning, and evolutionary biology. He is an Associate Professor of Philosophy at Kyoto University and a visiting researcher at the RIKEN Center for Advanced Intelligence Project in Japan. His work has been published in leading scientific and philosophical journals, including Philosophy of Science, British Journal for the Philosophy of Science, Philosophical Transactions of the Royal Society, and Proceedings of Machine Learning Research.
Organizers: Institute of Science Data Innovation, Institute of Philosophy and Thought
Supported by: SNU G-LAMP Project Group
Contact: Bonjin Koo (Researcher) at koobon1998@snu.ac.kr