Assessing Fair Machine Learning Strategies Through a Fairness-Utility Trade-off Metric

Due to the increasing use of artificial intelligence for decision making and the observation of biased decisions in many applications, researchers are investigating solutions that attempt to build fairer models that do not reproduce discrimination. Some of the explored strategies are based on adversarial learning to achieve fairness in machine learning by encoding fairness constraints through an adversarial model. Moreover, it is usual for each proposal to assess its model with a specific metric, making comparing current approaches a complex task. In that sense, we defined a utility and fairness trade-off metric. We assessed 15 fair model implementations and a baseline model using this metric, providing a systemically comparative ruler for other approaches.

Correction to: Reduced-form mechanism design and ex post fairness constraints

A Correction to this paper has been published: 10.1007/s40505-021-00211-1

Reduced-form mechanism design and ex post fairness constraints

This paper incorporates fairness constraints into the classic single-unit reduced-form implementation problem (Border in Economet J Econ Soc, 59(4):1175–1187, 1991, Econ Theory 31(1):167–181, 2007; Che et al. in Econometrica 81(6): 2487–2520, 2013; Manelli and Vincent in Econometrica, 78(6):1905–1938, 2010) with two agents. To do so, I use a new approach that utilizes the results from Kellerer (Math Ann 144(4):323–344, 1961) and Gutmann et al. (Ann Prob 19:1781–1797, 1991). Under realistic assumptions on the constraints, the conditions are transparent and can be verified in polynomial time.

Matchings with Group Fairness Constraints: Online and Offline Algorithms

We consider the problem of assigning items to platforms in the presence of group fairness constraints. In the input, each item belongs to certain categories, called classes in this paper. Each platform specifies the group fairness constraints through an upper bound on the number of items it can serve from each class. Additionally, each platform also has an upper bound on the total number of items it can serve. The goal is to assign items to platforms so as to maximize the number of items assigned while satisfying the upper bounds of each class. This problem models several important real-world problems like ad-auctions, scheduling, resource allocations, school choice etc. We show that if the classes are arbitrary, then the problem is NP-hard and has a strong inapproximability. We consider the problem in both online and offline settings under natural restrictions on the classes. Under these restrictions, the problem continues to remain NP-hard but admits approximation algorithms with small approximation factors. We also implement some of the algorithms. Our experiments show that the algorithms work well in practice both in terms of efficiency and the number of items that get assigned to some platform.

Data-Driven Methods for Balancing Fairness and Efficiency in Ride-Pooling

Rideshare and ride-pooling platforms use artificial intelligence-based matching algorithms to pair riders and drivers. However, these platforms can induce unfairness either through an unequal income distribution or disparate treatment of riders. We investigate two methods to reduce forms of inequality in ride-pooling platforms: by incorporating fairness constraints into the objective function and redistributing income to drivers who deserve more. To test these out, we use New York City taxi data to evaluate their performance on both the rider and driver side. For the first method, we find that optimizing for driver fairness out-performs state-of-the-art models in terms of the number of riders serviced, showing that optimizing for fairness can assist profitability in certain circumstances. For the second method, we explore income redistribution as a method to combat income inequality by having drivers keep an $r$ fraction of their income, and contribute the rest to a redistribution pool. For certain values of $r$, most drivers earn near their Shapley value, while still incentivizing drivers to maximize income, thereby avoiding the free-rider problem and reducing income variability. While the first method is useful because it improves both rider and driver-side fairness, the second method is useful because it improves fairness without affecting profitability, and both methods can be combined to improve rider and driver-side fairness.

Competitive-Blockchain-Based Parking System with Fairness Constraints

Parking slot detection is one of the most popular applications of Vehicular ad Hoc Network re-search field. Proposing smart algorithms for fast parking is crucial not only to facilitate drivers, but also to reduce traffic congestion, pollution, and vehicle energy consumption. Typically, an urban area has several competitive car parks and, in order to make the parking process automatic, a mechanism to ensure a fair competition among them is needed. Among all the methods able to guarantee transparency and equity in a system, blockchain is a robust technology. It has been success- fully applied in many different research fields, from financial to health. In this work, we provide an automaticparking system in which vehicles are allocated among several competitive parking areas (called competitors), through a blockchain-based approach, by applying a consensus mechanism to manage the system modifications. To this aim, two classes of fairness constraints are defined, according to which any new operation on the parking consortium must be approved by the members. Such an approach brings benefits for different reasons, starting from traffic condition improvement, up to driver stress and pollution decrease.

Closed-Form Results for Prior Constraints in Sum-Product Networks

Incorporating constraints is a major concern in probabilistic machine learning. A wide variety of problems require predictions to be integrated with reasoning about constraints, from modeling routes on maps to approving loan predictions. In the former, we may require the prediction model to respect the presence of physical paths between the nodes on the map, and in the latter, we may require that the prediction model respect fairness constraints that ensure that outcomes are not subject to bias. Broadly speaking, constraints may be probabilistic, logical or causal, but the overarching challenge is to determine if and how a model can be learnt that handles a declared constraint. To the best of our knowledge, treating this in a general way is largely an open problem. In this paper, we investigate how the learning of sum-product networks, a newly introduced and increasingly popular class of tractable probabilistic models, is possible with declared constraints. We obtain correctness results about the training of these models, by establishing a relationship between probabilistic constraints and the model's parameters.