Posts

It is widely observed that recommendation algorithms influence creators' productivity on content recommendation platforms. We reveal a fundamental trade-off between user satisfaction and content production volume: a relevance-driven recommendation policy with low exploration strength boosts short-term user satisfaction but undermines creators' productivity, whereas a more aggressive exploration policy can increase total content creation volume.

We introduce a competition model generalized from the Tullock contest to analyze the tension between human creators and GenAI. Our theory and simulations suggest that despite challenges, a stable equilibrium between human and AI-generated content is possible.

Our mechanism is deployed on Instagram Reels and the A/B test showed promising results!

We establish a social welfare maximization impossibility theorem in content creation competition game and propose theoretically optimal mechanisms for incentivizing creators in online content platforms.

We measure the user welfare loss under the content supply equilibria induced by relevance-driven recommender systems.

We formalize a model to capture the ‘‘learning user’’ effect in recommendation systems and design learning algorithm to tackle the challenges brought by the non-stationary feedback from a learning user.

We establish a unified framework for strategic classification problems and introduce the notion of strategic VC-dimension (SVC) to capture its PAC-learnability.