Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Sketchpad Series 

Exploring behaviour-adaptive and physiology-sensitive stimuli generation in cognitive neuroscience research (Poster abstract for the Sketchpad Series)

Poster Session D - Monday, March 9, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Paul Compensis¹ (); ¹University of Bamberg

Humans are typically confronted with perceptual input that changes dynamically in response to the environment, and often even adapts directly to the perceiver. Conversely, cognition experiments usually rely on static, perceiver-independent stimuli. To reconcile experimental rigor with ecological validity, recent artificial intelligence tools offer new solutions by enabling the creation of in-situ-adaptive, yet controllable, stimuli. Imagine, for instance, stimuli dynamically adapting to participants’ responses (e.g., ad-hoc generated images prompted by affective ratings) or to participants’ physiology (e.g., a pleasant melody generated in response to heightened electrodermal activity). As a proof-of-concept, I am illustrating the embedding of text-to-image (Stable Diffusion) and text-to-speech (Bark) transformer models into experimental scripts that then create stimuli adapting in real time to participants’ behaviours. Extending previous pilot work, I present two ongoing studies in which participants rate faces or voice sequences that are either (emotionally) intensified or attenuated during the experiment through rating-adaptive, individualised prompt adjustment. This procedure allows for individualised experimental designs and provides insights into how humans process dynamically changing perceptual input, allowing for a closer approximation of real-world conditions. In addition, psychophysiological parameters are recorded using a mobile sensor (EmotiBit) to capture responsiveness to adaptive stimuli and to identify targets for the next methodological step: creating stimuli that are directly physiology-sensitive. While focussing especially on methodological aspects of integrating transformer models, prompted by behavioural response, into experimental scripts, I also outline future steps involving the use of physiological signals to guide stimulus generation, since both approaches offer immense potential for cognitive (neuroscience) research.

Topic Area: METHODS: Other