SAS Workshop 2018

SAS Workshop 2018 – Epistemic opacity in computer simulation and machine learning

CfP: Opening the black box and finding it – dark …

Epistemic opacity in computer simulation and machine learning
Conference series: Science and Art of Simulation IV (SAS)
28.11. – 30.11.2018, HLRS, Stuttgart, Germany

Mathematics and technology are regarded as paradigms of clarity, traceability, and transparency. Seen from the outside, both may appear as enigmatic black boxes. Although for a long time, the working of the underlying relationships both in mathematics and technology were transparent for the experts; they revealed a well understood deductive system for the trained eye, a chain of necessities. However, their combination in computer simulation and machine learning creates challenges that are summarized under the title “epistemic opacity”. Epistemically opaque machines are said to represent a black box, the internal functioning of which cannot be fully elucidated. It is considered to be a novel phenomenon concerning the state of science. The novelty stems from a non-trivial entrenchment of mathematics and technology within the scientific method.

The fact that nature is opaque and should be elucidated by science is a classic topos Epistemic opacity, however, affects the scientific method by a complex interweaving of mathematics and technology. The procedure becomes nontransparent in a manner to be specified. This also raises the question of justifying results. Their justification is essentially linked to the method by which they were obtained.

The term “epistemic opacity” was brought into the discussion by Paul Humphreys (2004,2009), along with other things to characterize the novelty of the problems that computer simulation poses for philosophy. Since then it has been taken up many times, but systematic and historical clarification is pending. This is the subject of a three-day international interdisciplinary conference, which will take place at the High-Performance Computing Center of the University of Stuttgart (HLRS*) from the 28th to 30th of November 2018. The Science and Art of Simulation (SAS) conference series is intended to bring simulation scientists, philosophers, sociologists, and historians together. In the context of the SAS conference, the interdisciplinary interaction of simulation scientists and humanities, and social sciences experts is of great importance. Theoretical as well as practical questions on opacity and non-transparency in computer simulation and machine learning will be discussed; like which strategies will be applied in practice for dealing with opacity?

Topics

Interested simulation scientists, philosophers, sociologists, and historians can submit contributions on the following topics (non-exclusive):

Historical:

  • Is epistemic opacity a novel phenomenon?
  • What is the relationship between epistemic opacity and the Duhem-Quine thesis, especially confirmation holism?
  • What is the difference between epistemic opacity and classical black boxing?
  • What can we learn from the history of IT failures for epistemic opacity?

Systematic:

  • What is the evidence for epistemic opacity?
  • Are justification strategies changing in the face of methodical opacity?
  • What are the dimensions of opacity (example: social, technical and mathematical) and how do they relate to each other?
  • To what extent is epistemic opacity bound to the properties of computers? What happens if the concept is extended to scientific instruments in general?
  • What role does numerical sensitivity play in the development of opacity (bifurcation, instability, etc.)?
  • How can we know about opacity? Is there transparent and nontransparent opacity? (analog to knowledge and ignorance of ignorance)
  • What strategies are applied in practice for dealing with or resolving opacity?
  • What can we learn from the error detection of software systems (reliability, robustness)?

Abstracts (max. 3,000 characters including spaces without references) can be submitted until 1st of June 2018. Submissions should be prepared for anonymous review (no information identifying the author). Applicants will be notified latest by 30th of June. Accepted papers will be published in a proceeding volume by Springer. The participation fee is 150 Euro.

For submissions: http://philo.hlrs.de/openconf/openconf.php

Keynote Speakers (confirmed):

  • Anouk Barberousse (Université Paris-Sorbonne – Paris IV),
  • Gabriele Gramelsberger (RWTH Aachen),
  • Till Grüne-Yanoff (Royal Institute of Technology  Stockholm),
  • Hans Hasse (Technische Universität Kaiserslautern),
  • Paul Humphreys (University of Virginia),
  • Tarja Knuuttila (University of South Carolina),
  • Johannes Lenhard (University of Bielefeld),
  • Thomas Ludwig (Deutsches Klimarechenzentrum),
  • Julian Newman (Birkbeck Colleague, University of London).

If you have any questions, please contact Andreas Kaminski (kaminski@hlrs.de)

* The HLRS is a research institute and a supercomputer center with one of the fastest computer systems worldwide. HLRS conducts its own research in the field of high-performance computing. Emphasis is placed on the topics of scalability, performance optimization, big data, green IT, and in the application areas of health, environment, energy and mobility. The HLRS also houses a department for the philosophy of science and technology of computer simulation.