Auftakt-Wochenende

Mit: Volker Quandt, Leiter der IMPRO-AKADEMIE an der Universität Tübingen

Information Transmission, Representing Noise and the Semantic Content of Transmitted Signals

Seminar

Reading: Information or Noise? In: Deacon, T. Incomplete Nature: How Mind Emerged From Matter, 2012

Seminar Guest: Prof. Dr. Ben Jantzen (Virginia Tech, USA)

Information And Meaning Cont. + Prep-Session for next Session on Classical Decision-Making

Seminar

Signal or Noise? The Psychophysics of Decision-Making

Talk & Discussion

Prof. Dr. Rolf Ulrich (Psychologie, Universität Tübingen)

Prep Reading: Blogpost: The Many Schools of the Great Rationality Debate

Statistical Modeling of Signal and Noise

Talk & Discussion

Dr. Sabine Hoffmann (Institut für Statistik, LMU München)

Prep Reading: The multiplicity of analysis strategies jeopardizes replicability: lessons learned across disciplines. Hoffmann et al. 2021. Royal Society Open Science.

Theories of Perception: Unconscious Inferences, Constancy, and Perceptual Organization

Talk & Discussion

Prof. Dr. Hanspeter Mallot (Cognitive Neuroscience, Dept. of Biology University of Tübingen)

Prep Reading: Helmholtz - Physiologische Optik 1867 Paragraph 26 / Helmholtz's Treatise on Physiological Optics - Paragraph 26

Abstract: How can perception tell relevant or "veridical" information about the outside world from "noise" despite accidential  inputs or disturbances produced by the sensory system? In Helmholtz' approach, this is achieved by "unconscious inferences" which are based on analogies with previous acts of perception. Unconscious inferences allow to judge the size of an object based on the visual angle it subtends in the image and its perceived distance. They also allow to judge the color of a surface independent of the color of the illuminant, and so forth. While their nature and neural implementation remain largely elusive, inconscious inferences constitute what has been called the "Laws of Vision" (Wolfgang Metzger) or the "Logic of Perception" (Irvin Rock). The talk will discuss Helmholtz' judgement theory of perception together with related ideas such as perceptual constancy and perceptual organization.

Seminar Prep-Session for K. Munger's Spirals

Reading: 

Blog post „in the belly of the MrBeast“ by Kevin Munger

Vilém Flusser (2022). Communicology — Mutations in Human Relations?, Synopsis (p. 1-7)

Blog post „The Rise of Parasitic AI“ (can be skimmed)

Spirals

Talk & Discussion

Prof. Dr. Kevin Munger (Department of Political and Social Sciences, European University Institute, Florence, Italy)

Brainstorming Project Groups

Brainstorming Project Groups

Signal : Noise :: Information : Randomness

Seminar & Discussion

Prof. Dr. Robert Williamson (Foundations of Machine Learning Systems, Dept. of Computer Science, University of Tübingen)

Abstract: Signals are typically construed to convey information, whereas noise does not, being merely ‘random’. As an indirect approach to the question of understanding the signal vs noise distinction, I will explore some notions of information and randomness, especially as they are relevant in machine learning. The approach will be to adopt what is sometimes called  ‘Grothendieck’s relative point of view', whereby one tries to understand a thing by studying the transformations of the thing. To that end, I will outline the relationship between (many!) different measures of information in a “statistical experiment” and different utilitarian means by which prediction accuracy is measured. I will also look at the (large!) collection of possible notions of randomness, and connect them to notions of fairness and equity. Thus rather than (futilely) trying to answer what information / randomness (or signal / noise) “really is”, we will learn a little about the richness of the concepts, and how central they are to various important tasks. The take-away message is that information and randomness (and equivalently, by adequate arm-waving,  signal and noise) are far from unique, and are necessarily relativised. 

Key references on which I will rely are the following: 

1. Rabanus Derr, Robert C. Williamson (2025). Fairness and Randomness in Machine Learning: Statistical Independence and Relativization. The New England Journal of Statistics in Data Science Vol. 3, no. 1.
2. Robert C. Williamson and Zac Cranko (2024). Information processing equalities and the information-risk bridge . J. Mach. Learn. Res. 25, 1, Article 103.
3. Mark D. Reid, Robert C. Williamson (2011). Information, Divergence and Risk for Binary Experiments. 12 (22).

Sticking to an Information Diet: Significance, Social Media, Signal Processing

Talk & Discussion

Prof. Dr. Philipp Hennig (Methods of Machine Learning, University of Tübingen)

Ontology of the Ambience: Noise, Signal, Information, and War

Talk & Discussion (Online)

Prof. Dr. Chen-Pang Yeang  (Institute for the History & Philosophy of Science & Technology, University of Toronto)

Abstract: In the first half of the twentieth century, the notion of noise was transformed from disturbing sounds into generic errors.  A crucial element of this transformation concerned the emerging information sciences in the 1940s-50s that converted previous theoretical studies of noise into tools for statistical detection, estimation, prediction, and information transmission.  In so doing, they turned noise into an informational concept.  In the received historiography of command, control, communication, computing, and information during World War II and the Cold War, the focus has been on building of artificial systems with “smart” features—e.g., self-regulation and conveying meaning—or understanding why humans, organisms, or organizations possessed similar “intelligence.”  Thus, the researchers on the information sciences often operated under an ontology of an intelligent being as an object of inquiry. 

In this talk, I emphasize the roles of noise in the early development of the information sciences.  In contrast to the ontology of an intelligent object, noise epitomized an “ontology of the ambience” that stood for all kinds of uncertain and disturbing factors from the environment.  These treatments of noise comprised several defense projects in the framework of the rapidly expanding military-industrial-academic complex in the US during World War II: George Uhlenbeck’s radar target detection amidst a noisy background, Norbert Wiener’s statistical prediction and filtering of enemy target under various observational errors in antiaircraft gunfire directing, and Claude Shannon’s theory of information transmission through a noisy channel. At the onset of the Cold War, moreover, engineers utilized insights from the information sciences and turn noise from a plight to be suppressed into an asset for robust information transmission, leading to spread-spectrum communications.    

Generalization in machine learning systems from the behavioral perspective

Talk & Discussion (Online)

Prof. Dr. Max Raginsky  (Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, USA)

Constitution of Project Groups

Bayesianism in Philosophy & Science

2 day-Seminar, program here.

Dr. Tom Sterkenburg  (Munich Center for Mathematical Philosophy (MCMP), LMU München)

Rafael Fuchs  (Munich Center for Mathematical Philosophy (MCMP), LMU München)