Discovering the Causal Structure of Haptic Material Perception
by , ,
Abstract:
The sensory signals that occur when we touch or interact with objects carry the information necessary to perceive and reason about object properties. Research on material perception has provided evidence that humans can categorize materials and assess their similarity based solely on haptic information. This evidence is based on the performance on classification tasks and correlation analyses, which, by definition, provide no information on the causes of the observed behavior. This paper explores the use of causal discovery methods to analyze human haptic perception of material categories. Causal discovery algorithms analyze statistical patterns in the data, such as conditional (in)dependence relationships, and then determine causal relationships between the variables that are compatible with these patterns. The result is a set of causal graphs with nodes representing the variables and directed edges representing empirically plausible causal relationships. In this paper, a causal discovery algorithm is used to analyze material category ratings and vibratory signals from haptic exploration. The goal is to understand the underlying cause-effect structure linking material samples, vibration signals, and category similarity ratings. The identified causal structure indicates that the information represented by the slope of the vibratory signal plays a key role in rating a material's similarity to different categories, but in parts, it is only an indirect cause. The practical use of causal discovery methods for analyzing haptic perception data is demonstrated.
Reference:
Discovering the Causal Structure of Haptic Material Perception (Jaime Maldonado, Christoph Zetzsche, Vanessa Didelez), In Haptics: Understanding Touch; Technology and Systems; Applications and Interaction (Hiroyuki Kajimoto, Pedro Lopes, Claudio Pacchierotti, Cagatay Basdogan, Monica Gori, Betty Lemaire-Semail, Maud Marchal, eds.), Springer Nature Switzerland, 2025.
Bibtex Entry:
@InProceedings{maldonado2024discoveringcausalhaptic,
author={Maldonado, Jaime and Zetzsche, Christoph and Didelez, Vanessa},
editor={Kajimoto, Hiroyuki and Lopes, Pedro and Pacchierotti, Claudio and Basdogan, Cagatay and Gori, Monica and Lemaire-Semail, Betty and Marchal, Maud},
title={Discovering the Causal Structure of Haptic Material Perception},
booktitle={Haptics: Understanding Touch; Technology and Systems; Applications and Interaction},
year={2025},
publisher={Springer Nature Switzerland},
address={Cham},
pages={171--184},
abstract={The sensory signals that occur when we touch or interact with objects carry the information necessary to perceive and reason about object properties. Research on material perception has provided evidence that humans can categorize materials and assess their similarity based solely on haptic information. This evidence is based on the performance on classification tasks and correlation analyses, which, by definition, provide no information on the causes of the observed behavior. This paper explores the use of causal discovery methods to analyze human haptic perception of material categories. Causal discovery algorithms analyze statistical patterns in the data, such as conditional (in)dependence relationships, and then determine causal relationships between the variables that are compatible with these patterns. The result is a set of causal graphs with nodes representing the variables and directed edges representing empirically plausible causal relationships. In this paper, a causal discovery algorithm is used to analyze material category ratings and vibratory signals from haptic exploration. The goal is to understand the underlying cause-effect structure linking material samples, vibration signals, and category similarity ratings. The identified causal structure indicates that the information represented by the slope of the vibratory signal plays a key role in rating a material's similarity to different categories, but in parts, it is only an indirect cause. The practical use of causal discovery methods for analyzing haptic perception data is demonstrated.},
isbn={978-3-031-70058-3},
keywords = {EASE-H1}
}