AI and Shared Prosperity

Abstract

Future advances in AI that automate away human labor may have stark implications for labor markets and inequality. This paper proposes a framework to analyze the effects of specific types of AI systems on the labor market, based on how much labor demand they will create versus displace, while taking into account that productivity gains also make society wealthier and thereby contribute to additional labor demand. This analysis enables ethically-minded companies creating or deploying AI systems as well as researchers and policymakers to take into account the effects of their actions on labor markets and inequality, and therefore to steer progress in AI in a direction that advances shared prosperity and an inclusive economic future for all of humanity. CCS CONCEPTS• Social and professional topics → Employment issues; Automation; Economic impact; Codes of ethics; Socio-technical systems.

Six Human-Centered Artificial Intelligence Grand Challenges

“…Moreover, defining AI too narrowly risks alienating critical allies like the U.S., which has traditionally prioritized innovation in the name of expansive growth and national security (Bush, 2020). Other commentators point to the balance between innovating both for economic prosperity and for ecological longevity (Gupta, 2021, Sept 18;Klinova & Korinek, 2021), although the framers of the AI Act try to address possible constraints on innovation by providing for regulatory "sandboxes" that developers can use before deploying an AI product. The draft AI Act appears to be aiming at codification of "trustworthy" AI (see Section 3.…”

Advancing accountability in AI

“…An important strand of literature seeks to implement mathematical fairness metrics to assess a model's impartiality towards subgroups (IDB-OECD, 2021 [21]; Chouldechova, 2017 [23]; Kleinberg, Mullainathan and Raghavan, 2016 [24]; Corbett-Davies et al, 2017 [25]; Koshiyama et al, 2021[26]). Different fairness metrics lead to varying ways to assess bias in a system.…”

“… Fairness vs. performance: The trade-off between fairness/bias and performance is the subject of significant debate (Feldman et al, 2015[66]; Kleinberg, Mullainathan and Raghavan, 2016 [24]; Zafar et al, 2019 [67]). Model designers and developers can define acceptable boundaries of bias and performance, for example, by adopting metrics like statistical parity and accuracy.…”

Defining and classifying AI in the workplace

“…high levels of physical strength, or manual dexterity, or visual acuity etc. For instance, there are AI-powered technologies that describe people, text and objects to accommodate workers with low visual acuity (Henneborn and Eitel-Porter, 2020 [32]). Could the AI under consideration also make the workplace more inclusive for workers with disabilities, by supplementing and complementing their skills, by helping firms to put reasonable accommodations in place more easily and cost-effectively, and by creating an accessible work environment and helping workers with disabilities to interact with that environment?…”