Causal Effects with Unobserved Unit Types in Interacting Human-AI Systems

We study experiments on interacting populations of humans and AI agents, where both unit types and the interaction network remain unobserved. Although causal effects propagate throughout the system, the goal is to estimate effects on humans. Examples include online platforms where human users interact alongside AI-driven accounts. We assume a human-AI prior that gives each unit a probability of being human. While humans cannot be distinguished at the unit level, the prior allows us to compute the average human composition within large subpopulations. We then model outcome dynamics through a causal message passing (CMP) framework and analyze sample-mean outcomes across subpopulations. We show that by constructing subpopulations that vary in expected human composition and treatment exposure, one can consistently recover human-specific causal effects. Our results characterize when distributional knowledge of population composition (without observing unit types or the interaction network) is sufficient for identification. We validate the approach on a simulated human-AI platform driven by behaviorally differentiated LLM agents. Together, these results provide a theoretical and practical framework for experimentation in emerging human-AI systems.

• Publication year

  2026

• Venue

  Unknown venue

• Publication date

  2026-03-02

• Fields of study

  Mathematics, Computer Science, Psychology

• Identifiers
  arXiv 2603.01339
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Validating Causal Message Passing Against Network-Aware Methods on Real Experiments

  2026cited by this paper
• On Evolution-Based Models for Experimentation Under Interference

  2025cited by this paper
• Estimating Total Effects in Bipartite Experiments with Spillovers and Partial Eligibility

  2025cited by this paper
• Simulating and Experimenting with Social Media Mobilization Using LLM Agents

  2025cited by this paper
• Experimentation under Treatment Dependent Network Interference

  2025cited by this paper
• Differences-in-Neighbors for Network Interference in Experiments

  2025cited by this paper
• Can We Validate Counterfactual Estimations in the Presence of General Network Interference?

  2025influential reference
• AgentSociety: Large-Scale Simulation of LLM-Driven Generative Agents Advances Understanding of Human Behaviors and Society

  2025cited by this paper
• LLMs generate structurally realistic social networks but overestimate political homophily

  2024cited by this paper
• Higher-Order Causal Message Passing for Experimentation with Complex Interference

  2024cited by this paper
• Data-Driven Switchback Experiments: Theoretical Tradeoffs and Empirical Bayes Designs

  2024cited by this paper
• Generative Agents: Interactive Simulacra of Human Behavior

  2023cited by this paper
• Causal clustering: design of cluster experiments under network interference

  2023cited by this paper
• Causal message-passing for experiments with unknown and general network interference

  2023influential reference
• Simulating Social Media Using Large Language Models to Evaluate Alternative News Feed Algorithms

  2023cited by this paper
• Design of Panel Experiments with Spatial and Temporal Interference

  2023cited by this paper
• Correcting for Interference in Experiments: A Case Study at Douyin

  2023cited by this paper
• Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?

  2023cited by this paper
• Network Synthetic Interventions: A Framework for Panel Data with Network Interference

  2022cited by this paper
• Network Interference in Micro-Randomized Trials

  2022cited by this paper
• Reducing Marketplace Interference Bias Via Shadow Prices

  2022cited by this paper
• Markovian Interference in Experiments

  2022cited by this paper
• Exploiting neighborhood interference with low-order interactions under unit randomized design

  2022cited by this paper
• Switchback Experiments under Geometric Mixing

  2022cited by this paper
• Estimating the total treatment effect in randomized experiments with unknown network structure

  2022cited by this paper
• Detecting Interference in A/B Testing with Increasing Allocation

  2022cited by this paper
• Neighborhood Adaptive Estimators for Causal Inference under Network Interference

  2022cited by this paper
• Average Direct and Indirect Causal Effects under Interference

  2021cited by this paper
• Treatment Effects in Market Equilibrium

  2021cited by this paper
• Random graph asymptotics for treatment effect estimation under network interference

  2020cited by this paper
• Experimental Design in Two-Sided Platforms: An Analysis of Bias

  2020cited by this paper
• Design and Analysis of Switchback Experiments

  2020cited by this paper
• Reducing Interference Bias in Online Marketplace Pricing Experiments

  2020cited by this paper
• Experimenting in Equilibrium

  2019cited by this paper
• Variance Reduction in Bipartite Experiments through Correlation Clustering

  2019cited by this paper
• Metalearners for estimating heterogeneous treatment effects using machine learning

  2017cited by this paper
• Detecting Network Effects: Randomizing Over Randomized Experiments

  2017cited by this paper
• Generalized random forests

  2016cited by this paper
• Estimation and Inference of Heterogeneous Treatment Effects using Random Forests

  2015cited by this paper
• Design and Analysis of Experiments in Networks: Reducing Bias from Interference

  2014cited by this paper
• The rise of social bots

  2014cited by this paper
• Graph cluster randomization: network exposure to multiple universes

  2013cited by this paper
• An Iterative Construction of Solutions of the TAP Equations for the Sherrington–Kirkpatrick Model

  2012cited by this paper
• Estimating Average Causal Effects Under General Interference

  2012cited by this paper
• Identification of treatment response with social interactions

  2012cited by this paper
• The dynamics of message passing on dense graphs, with applications to compressed sensing

  2010cited by this paper
• Message-passing algorithms for compressed sensing

  2009cited by this paper
• Principal Stratification in Causal Inference

  2002cited by this paper
• Finite Mixture Modeling with Mixture Outcomes Using the EM Algorithm

  1999cited by this paper
• Probabilistic reasoning in intelligent systems: Networks of plausible inference

  1991cited by this paper
• Fusion, Propagation, and Structuring in Belief Networks

  1986cited by this paper
• Reverend Bayes on Inference Engines: A Distributed Hierarchical Approach

  1982cited by this paper
• Randomization Analysis of Experimental Data: The Fisher Randomization Test Comment

  1980cited by this paper
• Planning of Experiments

  1958cited by this paper

• No citing papers are available for this paper.