NeuroViT: A Brain-Aligned Vision Transformer for Product Image Analysis
Publication Date
1-26-2026
Abstract
This research introduces a brain-aligned approach to image mining grounded in the organization of the human visual system. Although images are central to product evaluation, existing work largely overlooks how consumers' brains process visual information. We address this gap by developing NeuroViT, a vision transformer trained on a large-scale fMRI dataset to predict neural activity in the visual cortex. Using the travel industry as a test case, we validate NeuroViT with hotel images from an original fMRI study, showing that predicted neural activity matches observed brain responses. We then derive two brain-aligned image metrics: retinotopic activity as a domaingeneral indicator of subjective visual complexity, and activity in place-sensitive regions as a measure of navigational affordance. These metrics reliably track human judgments across diverse image datasets and uncover novel drivers of property image effectiveness. We demonstrate the managerial relevance of NeuroViT through two applications: a pre-registered experiment showing that hotel image sets with higher navigational affordance increase booking intentions, and an econometric analysis of Airbnb listings showing that brain-aligned metrics provide additional explanatory power and translate into meaningful revenue differences. Together, this work integrates neuroscience and machine learning to provide an open-source framework for perception-based, brainaligned image analysis in marketing.
Document Type
Article
Keywords
E-Commerce, Image Mining, Machine Learning, Computer Vision, Neuroscience, fMRI
Disciplines
Marketing
Source
SMU Cox: Marketing (Topic)
Language
English
