BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.iacapconf.org//iacap-2026//talk//BGJSHQ
BEGIN:VTIMEZONE
TZID:US/Central
BEGIN:DAYLIGHT
DTSTART:20250715T000000
TZNAME:CDT
TZOFFSETFROM:-0500
TZOFFSETTO:-0500
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251102T020000
RDATE:20261101T020000
TZNAME:CST
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260308T030000
RDATE:20270314T030000
TZNAME:CDT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Deep Learning Models in Science: Explanation without Representatio
 n? - Conny Knieling
DTSTART;TZID=US/Central:20260715T164000
DTEND;TZID=US/Central:20260715T171000
DTSTAMP:20260726T082708Z
UID:pretalx-iacap-2026-BGJSHQ@pretalx.iacapconf.org
DESCRIPTION:Deep learning models (DLMs) now automate core stages of scient
 ific inquiry\, from protein-structure prediction to medical image reconstr
 uction. Yet their opacity has led many philosophers of science to argue th
 at they cannot genuinely contribute to scientific explanation unless they 
 qualify as scientific models standing in fine-grained representational rel
 ations to their targets. I argue that this assumption is mistaken—and th
 at it obscures how the deployment of machine learning techniques is actual
 ly transforming scientific practice.\nThe dominant view treats opacity as 
 a threat because it allegedly blocks assessment of a model’s representat
 ional status. But this presupposes that explanatory contribution depends o
 n representation. I challenge that presupposition. Contemporary DLMs neith
 er function as mechanistic surrogates nor map onto the causal structure of
  their targets in the way traditional representational models do. Demandin
 g such mapping misconstrues their epistemic role.\nInstead\, I argue that 
 DLMs are better understood in comparison to scientific instruments in thei
 r automated epistemic role. Their contribution to explanation lies not in 
 mirroring structure but in reliably generating results that stabilize patt
 erns of dependence and support counterfactual reasoning. Drawing on the ep
 istemology of experimentation\, I propose robustness—established through
  benchmarking\, error management\, and cross-validation—as the appropria
 te normative standard.\nOpacity\, on this view\, is not a failure of repre
 sentation but a challenge of validation. Once reframed in terms of robustn
 ess rather than representational fidelity\, the explanatory role of automa
 ted deep learning systems becomes intelligible without requiring explanati
 on by representation.
LOCATION:Apollo Auditorium
URL:https://pretalx.iacapconf.org/iacap-2026/talk/BGJSHQ/
END:VEVENT
END:VCALENDAR