Asymmetric Influence in AI-Augmented Systems : Evidence From the Initial Coin Offering Market

The growing consensus that human intelligence and artificial intelligence (AI) are complementary has led to the emergence of AI-augmented systems. In this paper, we investigate whether AI agents influence human decisions in the high-risk environments of initial coin offering (ICO) where most ICO fail. Using panel data from a leading ICO evaluation platform that uses AI to augment human experts, we examine how humans integrate the assessments of AI agents into their assessments when real outcomes are at stake. Our results show that AI agents exert an asymmetric influence on human decisions. When the AI agent rates an ICO low, the decisions of human experts follow it. When the AI agent rates an ICO high, the decisions of human experts deviate from it. Thus, we conclude that human experts use the AI agent’s low score to screen out ICOs and use the agent’s high score as a trigger to investigate an ICO more deeply and come to their own conclusions about its potential. The same human experts that were influenced by low AI ratings were not influenced by other human experts