Neural correlates of evaluative bias against artificial intelligence-labeled versus human-labeled artworks

The rapid advancement of generative artificial intelligence (AI) has enabled machines to produce creative outputs, such as artworks, that rival human creations. However, despite these technological achievements, public acceptance and valuation of AI-generated creative products remain uncertain. This study investigates the bias against AI-generated artworks by examining both behavioral and neural responses using electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS). Across two pre-registered studies (total N = 125), participants evaluated paintings labeled as either human-created or AI-generated, with the artworks being objectively identical. Behavioral results revealed a consistent bias against AI-labeled paintings, with participants rating them lower in different evaluative dimensions. Neural data from EEG indicated relatively higher P300 amplitudes and alpha power during the evaluation of human-labeled paintings, suggesting potentially greater attentional allocation and cognitive engagement. The fNIRS results further demonstrated increased activity in the right angular gyrus and stronger functional connectivity between the inferior frontal gyrus and angular gyrus for human-labeled artworks, which may indicate deeper semantic and emotional processing. These findings highlight that biases against AI-generated artworks are not only evident in explicit evaluations but also manifest in underlying neural processes, reflecting a persistent preference for human creativity. The study underscores the importance of addressing public biases to foster acceptance and trust in AI technologies.