A Large-scale Time-aware Agents Simulation for Influencer Selection in Digital Advertising Campaigns

TL;DR

A Time-aware Influencer Simulator (TIS) is introduced, helping promoters identify and select the right influencers to market their products, based on LLM simulation, showing that simulating user timelines and content lifecycles over time simplifies scaling, allowing for large-scale agent simulations in social networks.

Abstract

In the digital world, influencers are pivotal as opinion leaders, shaping the views and choices of their influencees. Modern advertising often follows this trend, where marketers choose appropriate influencers for product endorsements, based on thorough market analysis. Previous studies on influencer selection have typically relied on numerical representations of individual opinions and interactions, a method that simplifies the intricacies of social dynamics. In this work, we first introduce a Time-aware Influencer Simulator (TIS), helping promoters identify and select the right influencers to market their products, based on LLM simulation. To validate our approach, we conduct experiments on the public advertising campaign dataset SAGraph which encompasses social relationships, posts, and user interactions. The results show that our method outperforms traditional numerical feature-based approaches and methods using limited LLM agents. Our research shows that simulating user timelines and content lifecycles over time simplifies scaling, allowing for large-scale agent simulations in social networks. Additionally, LLM-based agents for social recommendations and advertising offer substantial benefits for decision-making in promotional campaigns.

Figures (5)

• Figure 1: The interaction-based user timeline and content lifecycle on the internet. "Active" refers to users interacting with the content they encounter. "Inactive" indicates that they are offline or disengaged from the content. "Engaging" content refers to content that users frequently interact with, whereas "Stale" content refers to outdated content that no longer attracts people to interact.
• Figure 2: (a) Temporal Trajectory Simulation identifies agents and content to simulate within specific time frames, focusing on the user timeline and content lifecycle. (b) Agent Simulation uses LLMs to simulate behaviors, including self-awareness, social behavior prediction, and self-assessment. (c) Social Network Simulation leverages the TIS framework to model social network dynamics post-advertisement, offering insights into campaign effectiveness over time.
• Figure 3: (a) User timeline modeling. "GMM Fit" represents the simulated curve, "User Active Counts" indicates the actual interactions engaged in each period of the day, and "Forecast Active Counts" signifies the predicted number of interactions for a certain user. (b) Simulation of the content lifecycle, where each line represents the survival probability distribution of a post or comment over a day, which is 1,440 minutes.
• Figure 4: The illustration of the interaction network after 4 periods’ simulation, starting with the influencer posting an advertisement for Ruby Face Cream. The colored dialogue boxes represent the agent's comments from different periods. The scores for the generated comments reflect the inclination to purchase, starting from no intention (0) to a definite plan to buy (5).
• Figure 5: The changes in the NDCG@10 metric over time in the influencer selection task for Advertising Campaigns, which illustrates the variation curves of four products on the SAGraph dataset. These curves represent the social networks across four different domains.