The experiment-technology relationship in AI-driven science

09 May 2023 16:00 till 18:00 - Location: TPM, Room C1.060 | Add to my calendar

Presentation in Digital Philosophy Seminar by Sahra Styger, PhD Candidate, University of Konstanz (Germany)

The claim that experimentation constitutes the central link between science and technology offers a natural starting point to analyse the role of machine learning within our scientific pursuits.

In my talk, I will elaborate on three different types of the experiment-technology relationship (RET), where the technology is machine learning, namely „in vivo experiment, in silico data analysis“, „computer experiment and simulation“, and „automated scientific experiment“. This typology represents a continuum of different levels of the autonomy of the AI-system within the RET, namely ‘instrumental’, ‘semi-autonomous’, and ‘autonomous’.

By focussing on how ML techniques are being used in experiments, I will make explicit the  epistemic agents, human and artificial, that are involved in the different RET. I achieve this by  relabelling the three types as practices. Rephrasing the types as practices helps to clarify the role of ML associated with experiments and to understand how ML is used to create different scientific products, i.e. evidence, claims, or novel knowledge.
Furthermore, I will rise the question of how to assess such scientific products by means of scientific values. How can we develop an useful ‘values-approach’ to evaluate AI-systems in science, or more precisely, the types of RET as a practices?

Programme:

  • 16h: seminar hosted by Sahra
  • 17h: drinks