Authors: Sunday O. Oladejo
This research introduces a novel framework that combines causal uplift modeling with many-objective optimization to improve promotional offer allocation in recommender systems. The study simultaneously optimizes conversion uplift, fairness, budget expenditure, and treatment overlap, enabling organizations to balance business objectives with responsible AI principles.
The framework employs causal machine learning approaches, including S-Learner, T-Learner, X-Learner, and Causal Forest models, together with advanced many-objective optimization algorithms such as NSGA-III and MOEA/D. Experimental results demonstrate that fairness-aware optimization can improve conversion uplift while maintaining fairness constraints and controlling costs.
Key contributions include:
- Fairness-aware uplift optimization for recommender systems.
- Integration of causal machine learning and many-objective optimization.
- Simultaneous consideration of uplift, fairness, cost, and treatment overlap.
- Extensive evaluation using the Hillstrom dataset and Monte Carlo simulations.
- Identification of Pareto-efficient treatment allocation policies.
Published in: IEEE Access, Volume 14, 2026.
