Ezzat Elokda

Ezzat Elokda, Ruiting Wang, Karl H. Johansson et al.

Motivated by the need to develop fair and efficient schemes to facilitate the electrification of transport, this paper proposes a non-monetary karma economy for flexible Electric Vehicle (EV) charging, managing the intertemporal allocation of limited power capacity. We consider a charging facility with limited capacity that must schedule arriving EVs to charge in real-time. For this purpose, the facility adopts online karma auctions, in which each EV user is endowed with non-tradable karma tokens, places a karma bid in each time interval it is present in the facility, and capacity is allocated to the highest bidders, who must pay their bids. These payments are subsequently redistributed to the users to form a closed, indefinitely sustainable economy. The main contribution is to extend previous karma Dynamic Population Game (DPG) formulations to this setting which features novel State of Charge (SOC) dynamics and private trip deadlines in addition to urgency. A Stationary Nash Equilibrium (SNE) of the EV charging karma economy is guaranteed to exist, and it is demonstrated to provide pronounced benefits with respect to benchmark scheduling schemes as it balances between meeting deadlines and prioritizing high urgency.

Nicolò Pagan, Joachim Baumann, Ezzat Elokda et al.

Prediction-based decision-making systems are becoming increasingly prevalent in various domains. Previous studies have demonstrated that such systems are vulnerable to runaway feedback loops, e.g., when police are repeatedly sent back to the same neighborhoods regardless of the actual rate of criminal activity, which exacerbate existing biases. In practice, the automated decisions have dynamic feedback effects on the system itself that can perpetuate over time, making it difficult for short-sighted design choices to control the system's evolution. While researchers started proposing longer-term solutions to prevent adverse outcomes (such as bias towards certain groups), these interventions largely depend on ad hoc modeling assumptions and a rigorous theoretical understanding of the feedback dynamics in ML-based decision-making systems is currently missing. In this paper, we use the language of dynamical systems theory, a branch of applied mathematics that deals with the analysis of the interconnection of systems with dynamic behaviors, to rigorously classify the different types of feedback loops in the ML-based decision-making pipeline. By reviewing existing scholarly work, we show that this classification covers many examples discussed in the algorithmic fairness community, thereby providing a unifying and principled framework to study feedback loops. By qualitative analysis, and through a simulation example of recommender systems, we show which specific types of ML biases are affected by each type of feedback loop. We find that the existence of feedback loops in the ML-based decision-making pipeline can perpetuate, reinforce, or even reduce ML biases.