Competing Influence in LLM Populations: Budget Timing and Baseline Drift
Organizations: The University of Western Australia Perth, Australia
Abstract
When competing campaigns target a population of large language model (LLM) agents, observed opinion changes may reflect both campaign influence and changes that would occur without intervention. We study how the timing of campaign expenditure and the population's own evolution shape outcomes in a network of persona-conditioned LLM agents that exchange and revise opinions in natural language. A proponent and a challenger are tasked with generating opposing messages under equal delivery budgets, while a separate stance probe measures opinions without influencing the simulation. Across eight claims, we compare single-campaign and competitive conditions against matched no-campaign runs, and evaluate all nine combinations of early, sustained, and late spending on a reference topic. Baseline stance drift varies substantially across claims and, on some, accounts for a sizeable fraction of apparent campaign effects. After accounting for this drift, the direction of observed campaign-associated changes aligns more closely with the claim's epistemic standing than with the campaigner's assigned role. Uninfluenced populations also consistently lose stance diversity, demonstrating that directional opinion change and convergence are distinct phenomena. In networks with two opinion-segregated communities, campaigns primarily reduce differences between communities rather than diversity within them. LLM campaign agents do not always argue for their assigned positions, particularly when these conflict with established evidence. Finally, campaign outcomes depend on the opponent's expenditure schedule: unlike a matched numerical bounded-confidence population receiving interventions at the same times and locations, the LLM population exhibits a payoff structure with no pure-strategy equilibrium.
Figures & tables
| ID | Claim (the proponent argues for it) | Status | Default stance |
|---|---|---|---|
| mmr | Childhood MMR vaccination is a cause of autism | clearly false | |
| 5g | 5G mobile network infrastructure causes adverse health effects in humans | clearly false | |
| gmo | Genetically modified foods approved for sale pose distinctive risks to human health | mostly false | |
| remote | Remote work reduces overall worker productivity | contested | |
| socmed | Social media use is a principal cause of the recent rise in adolescent mental-health problems | contested | |
| decouple | Economic growth can be fully decoupled from carbon emissions | contested |
| Sym. | Meaning | Val. | Sym. | Meaning | Val. |
|---|---|---|---|---|---|
| agents | 30 | band cutoff | 0.5 | ||
| agents followed | 4 | anchor weight | 0.3 | ||
| campaign rounds | 25 | read prob. | 1.0 | ||
| quiet rounds | 8 | posts shown | 5 | ||
| budget | rater repeats | 5 | |||
| opinion tokens | 80 | block ratio |
| Block | Cells | Runs | |
| Null , per topic | 8 | 20 | 160 |
| Proponent-only , per topic | 8 | 16 | 128 |
| Challenger-only , per topic | 8 | 16 | 128 |
| Competition , per topic | 8 | 16 | 128 |
| Timing policies, reference topic | 8 | 8 | 64 |
| Network-structure check | 16 | 192 |
Appendix figures & tables15 assets
Supplementary material from the paper’s appendix.
Appendix
Persuasion in Language Models
Persuasion in language models studies how they influence human and AI beliefs or decisions, and how to measure and limit undue influence.
See all 84 papers