Quantitative Fairness -- A Framework For The Design Of Equitable Cybernetic Societies

TL;DR

This work proposes a quantitative, transactional, distributive fairness framework, which enables systematic design of socially feasible decision-making systems and emphasizes the importance of fairness and transparency when designing algorithms for equitable, cybernetic societies.

Abstract

Advancements in computer science, artificial intelligence, and control systems of the recent have catalyzed the emergence of cybernetic societies, where algorithms play a significant role in decision-making processes affecting the daily life of humans in almost every aspect. Algorithmic decision-making expands into almost every industry, government processes critical infrastructure, and shapes the life-reality of people and the very fabric of social interactions and communication. Besides the great potentials to improve efficiency and reduce corruption, missspecified cybernetic systems harbor the threat to create societal inequities, systematic discrimination, and dystopic, totalitarian societies. Fairness is a crucial component in the design of cybernetic systems, to promote cooperation between selfish individuals, to achieve better outcomes at the system level, to confront public resistance, to gain trust and acceptance for rules and institutions, to perforate self-reinforcing cycles of poverty through social mobility, to incentivize motivation, contribution and satisfaction of people through inclusion, to increase social-cohesion in groups, and ultimately to improve life quality. Quantitative descriptions of fairness are crucial to reflect equity into algorithms, but only few works in the fairness literature offer such measures; the existing quantitative measures in the literature are either too application-specific, suffer from undesirable characteristics, or are not ideology-agnostic. Therefore, this work proposes a quantitative, transactional, distributive fairness framework, which enables systematic design of socially feasible decision-making systems. Moreover, it emphasizes the importance of fairness and transparency when designing algorithms for equitable, cybernetic societies.

Figures (5)

• Figure 1: The Fairness of Decision-Making Systems. Decision-Making is depicted as a resource allocating process, that takes an initial situation as an input, makes a decision, and provides a resource allocation as an output. The fairness discussion of decision-making therefore covers procedural and distributive fairness. For procedural fairness, aspects such as (i) the transparency of the process, and (ii) what chances individuals have to influence the process are important questions to answer. For distributive fairness, aspects such as (i) what resources are fairness-relevant and distributed to (ii) which groups and (iii) how decisions are made (based on which guiding principles) are important questions to answer.
• Figure 2: A Transactional View on Distributive Fairness. Aristotle distinguishes dianemetic and diorthotic, distributive fairness. Dianemetic fairness is concerned with the fair distribution of resources from top to bottom (endownment) from usually one central decision maker (institution, government) to the population. Diorthotic fairness is concerned with the fair distribution of resources in a decentralized way as a result of transactions between individuals (markets).
• Figure 3: Philosophic Guiding Principles For Fairness. Six guiding principles for distributive fairness that cover different perspectives on fairness. The difference principle advocates a situation in which the least-advantaged (poorest) are in the best possible situation. The equality principle advocates an equal distribution of outcomes. The equality-of-opportunity principle advocates an equal distribution of the initial situation to provide every individual with the same chances to make their own luck. The greater-good principle is the basis for Utilitarianism and argues that the suffering of the few is acceptable if it serves the greater good of the majority. The proportion principle advocates distributions that stand in proportional relation to the contributions and status of individuals. The sufficiency principle advocates distributions in which everyone receives a certain sufficient minimum to satisfy basic needs, inequalities above that are not an issue.
• Figure 4: Case Study: Fair cake-cutting problem & dianemetic fairness. A given cake that can be cut into three heterogeneous pieces only shall be distributed across two individuals, that differ in their contributions to paying or making the cake, and their preferences on the toppings. Eight allocations of the three pieces to the two agents are possible. Different guiding principles on fairness result in different, fairness-optimal recommendations for how to distribute the cakes.
• Figure 5: Case Study: Fishermen & diorthotic fairness. Two fishermen go fishing at a pond every day, and at this specific day they fish seven fishes together. They differ in the working hours and catching rates due to different fishing techniques. Their outcome is stochastic. How should they distribute the total fish catch per day? Assuming fish is divisible, there is a continuous allocation space of the fish ranging from zero to seven fishes for each fisher. The Pareto-efficient frontier (blue line) displays this solution space. Different guiding principles on fairness result in different heatmaps and social welfare functions (contour plots of the heatmaps), with different recommendations on how to distribute the daily fish catch.