Date: 20 February 2026, Friday
Place: C Block Amphitheater
Time: 12:30
“Towards Scalable Computational Principles of Perception: Modeling Human Behavior Under Realistic Task Complexity“
by
Can Oluk, Brain Mind Institute at EPFL- École Polytechnique Fédérale de Lausanne
Abstract:
How does the human mind transform chaotic sensory input into a coherent understanding of the world? My research investigates the computational mechanisms of perception, with a focus on vision and its interplay with cognitive processes. Specifically, I aim to identify principles that generalize to real-world conditions, overcoming the limitations of models developed in simplified settings. To do this, I combine normative, interpretable, image-computable modeling with behavioral experiments that involve realistic complexity.
In this talk, I illustrate this approach with two studies. First, I increase stimulus (input) complexity by allowing both target and background properties to vary unpredictably across trials. Under these conditions, human target detection is best explained by normalization-based heuristics. Second, I increase response (output) complexity by collecting eight-alternative motion-direction judgments concurrently with confidence ratings. This joint analysis reveals internal processing dynamics, such as competition and gain control, that are obscured in standard two-choice tasks. I conclude by discussing why this framework is a promising strategy and my future plans to refine and extend this framework to realistic target identification tasks.
Biography:
Can Oluk is a postdoctoral researcher at the Brain Mind Institute at EPFL, working with Prof. Michael Herzog. Before joining EPFL, he was a postdoctoral fellow at ENS-PSL in Paris with Prof. Pascal Mamassian. He earned his Ph.D. in Psychology at the University of Texas at Austin, where he completed the Center for Perceptual Systems interdisciplinary training program in the lab of Prof. Wilson Geisler. He holds a B.A. in Psychology, with a minor in Philosophy, from Bilkent University. His research focuses on characterizing computations in the human perceptual system and how cognitive processes build upon them. He develops normative, interpretable models and evaluates their predictions across a diverse set of visual tasks to identify principles that generalize to real-world complexity.