“Distinctive from What? And for Whom?” Deep Learning-Based Product Distinctiveness, Social Structure, and Third-Party Certifications
利用深度学习分析现代艺术家的作品独特性，发现艺术家与竞争对手及自身过去的差异能提升第三方认证，但社会结构地位会调节这种效果。
How do producers’ distinctiveness and social structure influence third-party certifications? We argue that producers compete against prior and current competitors, as well as against their past selves. In the context of 153 artists active during a key period of the emergence of modern art (1905-1916), we use a convolutional neural network used in computer vision to extract feature vectors of artworks, and then measure quantitative distance of these artists’ works from canonical reference points. We find that artists are rewarded for distinctiveness from prior and current competitors and their past selves (up to a point). However, the artists’ autonomy to differentiate themselves depends on their position in social structure, which we divide into the supply-side of artist-to-artist networks, and the demand side of artist-to-gallerist networks. Artists with high or low supply-side status receive higher rewards for distinctiveness from current competitors than do artists with middle supply-side status. Artists with higher demand-side status receive higher rewards for distinctiveness from their own past, but lower rewards for distinctiveness from current competitors. These results show that peers strive to constrain each other to conform to positions of gravity within product space, and that market audiences deploy either higher or lower constraints on a producer’s identity depending on the reference point.