Impact of downvotes on bot and user engagement in a blockchain-based social media platform

Steemit, a blockchain-based social media (BOSM) platform, allows users to participate in platform governance and earn token rewards through voting. However, some users manipulate votes using bots. On August 27, 2019, Steemit introduced an additional downvote pool for each user to encourage governance of vote manipulation via downvotes. This study employs Regression Discontinuity in Time (RDiT) to empirically analyze the impact of the policy on user and bot behaviors. After the policy, bot upvotes and transfers from users to bots significantly decreased, suggesting effective curbing of vote manipulation, while the delegation of voting power from users to bots did not exhibit significant changes. However, the policy change led to potential side effects. Human users' upvotes and the total number of posts and comments significantly decreased, reflecting a hesitation in content creation and interaction due to increased negative feedback. Additionally, the policy significantly increased reciprocal downvoting, suggesting that some downvotes were used for retaliatory purposes among users. This study is the first to investigate the effects of voting system changes on a BOSM platform, enhancing our understanding of how platform-level design features influence engagement and interactions among humans and bots.