Scientific Understanding and Deep Neural Networks

Lecture Series: Philosophy of Science and Machine Learning

Poster

Description

In an age where artificial intelligence (AI) is transforming science, it becomes increasingly important to reflect critically on the foundations, methodologies, and implications of these advancements. This lecture series will investigate fundamental issues in AI from the vantage point of philosophy of science, which includes topics such as the transparency and interpretability of AI within scientific research, or the impact of AI on scientific understanding and explanation.

This lecture series is a special edition of the AI Colloquium at TU Dortmund University, coorganized by the Lamarr Institute for Machine Learning and Artificial Intelligence, the Research Center Trustworthy Data Science and Security (RC Trust), and the Center for Data Science & Simulation at TU Dortmund University (DoDas). The Lecture Series is organized by the Emmy Noether Group „UDNN: Scientific Understanding and Deep Neural Networks”, and generously funded by the German Research Foundation (DFG; grant 508844757).

Time & Place

Thursdays, 4:15pm, room 303, 3rd floor, Joseph-von-Frauenhofer-Str. 25, 44227 Dortmund

Speakers

17.10.24 André Curtis Trudel (University of Cincinnati): „On finding what you’re (not) looking for: prospects and challenges for AI-driven discovery“

Abstract:

Recent high-profile scientific achievements by machine learning (ML) and especially deep learning (DL) systems have reinvigorated interest in ML for automated scientific discovery (e.g., Wang et al. 2023). Much of this work is motivated by the thought that DL methods might facilitate the efficient discovery of phenomena, hypotheses, or even models or theories more efficiently than traditional, theory-driven approaches to discovery. This talk considers some of the more specific obstacles to automated, DL-driven discovery in frontier science, focusing on gravitational-wave astrophysics (GWA) as a representative case study. In the first part of the talk, we argue that despite these efforts prospects for DL-driven discovery in GWA remain uncertain. In the second part, we advocate a shift in focus towards the ways DL can be used to augment or enhance existing discovery methods, and the epistemic virtues and vices associated with these uses. We argue that the primary epistemic virtue of many such uses is to decrease opportunity costs associated with investigating puzzling or anomalous signals, and that the right framework for evaluating these uses comes from philosophical work on pursuitworthiness.

28.11.24 Tim Räz (University of Bern) „The Concept of Memorization in Machine Learning“ (cancelled)

28.11.23 Leda Berio (RU Bochum): „Social interaction with artificial agents: when, how, why? Perspective taking and social scripts in HRI and interaction with AI“

When a robotic dog wags its tail, we do not hesitate to interpret it as a sign of happiness. We get upset when seeing little robot dinosaurs, barely more than sophisticated toys, mistreated. We hesitate to strike little bug-like robotic objects. Evidence shows, in other words, that our interactions with robots are laden with affect; and this is despite our full awareness that robots do not, ultimately, feel. What are the factors that influence these attributions? What aspects of design can influence the way we interact with artificial agents? I ague that considering interactions with artificial agents in terms of emotionally-loaded scripts can contribute to explaining our attribution of emotional states to social robots as well as our emotional reactions during interactions with them. Moreover, it helps us identify the normative components of such interactions. In this talk, I first explore the ways design features influence basic mechanisms of spontaneous perspective taking with social robots, presenting data that shows how visual appearance modulates these effects. Next, I propose that to explain more sophisticated mental state attribution, we should consider social interactions as activating scripts and schemata (Bicchieri and McNally, 2018) that come with expectations on how agents should behave and feel. Scripts contain information about expected emotional reactions, and their activation prescribe the interpretation of emotions in normative ways, as well as emotional attributions. In this sense, I suggest, when interacting with social robots, our behaviors and emotions, as well as our attributions, are highly normatively regulated. To conclude, I discuss how basic design features relate to the activation of scripts.

12.12.24 Stefan Buijsman (TU Delft): „The impact of epistemic dependence on AI for understanding“

23.01.25 Nina Poth (Radboud University Nijmegen): „Common Sense and the Limits of AI“

Participation

Participation is free, but places are limited. If you are interested in participating online, please register via the following form: https://forms.microsoft.com/r/W3whw0ac3B. If you would like to attend in person, please send an e-mail to udnn.ht@tu-dortmund.de. 

We look forward to your participation and insightful discussions. 

Kind regards, 

The UDNN team (https://udnn.tu-dortmund.de/) 

Main Organizers

Annika Schuster, Frauke Stoll, and Florian J. Boge