Avatar
FoWParticipants

Imagine all the People and AI in the Future of Work

Crowdsourcing, Job Platforms

Shonan Village Workshop on FoW Sep 8-12, 2019

We envision the Future of Work (FoW) to be a place where humans are empowered with the ability to rely on AI machines in an on-demand fashion, with ability to juggle diverse job opportunities that provide intellectual self-actualization, and enhance their capabilities by continuous knowledge and skill acquisition with a variety of onboarding and training tools. We recognize that the current view of “humans-in-the-loop” tends to see humans as machines, robots, or low-level agents used or exploited in the service of broader AI goals. Our hope is that people in their workplace in the future should be treated as fully human, with respect and dignity with the right to productive employment and the goal of bringing them back in the “frontier” of “humans-in-the-loop” systems. Such an environment will allow everyone everywhere to get a job online and offline, train for a new job, and get help from a mix of humans and AI machines. Ensuring portability between FoW marketplaces, and guaranteeing the protection of workers’ rights, will play a major role in providing a rewarding and safe work environment to all. Existing platforms must rethink their design to empower humans and be at the frontier of FoW.

Today, humans’ relationship to work is changing as online job platforms are blurring the boundaries between physical and virtual workplaces. That is witnessed in freelancing platforms such as CrowdWorks in Japan, TaskRabbit and Fiverr in the USA, and Qapa and MisterTemp in France, crowdsourcing platforms such as Crowd4U in Japan, Wirk, and Prolific Academic in Europe, Amazon Mechanical Turk and Figure Eight in the USA, and platforms that provide help with entrepreneurship such as de Asra in India. Prospective employees can find temporary jobs in the physical world (e.g., a plumber, an event organizer, a gardener, can offer their jobs online), or in the form of virtual gigs (e.g., logo design, web programmer). Job providers can hire one or many individuals to achieve a task. The same person can take on those roles at any point in time. An employer can be a regular citizen who needs to hire a plumber, a social scientist needed to conduct population studies to verify some theories, a data scientist needing to validate a new algorithm, a domain expert seeking to verify how much interest a new product generates. The diversity of needs has given rise to a variety of platforms, all of which act as intermediaries between job providers and job seekers. Platforms differ in their ability to manage physical and virtual jobs, in their support for onboarding, socializing, training, and credentialing for employees, in automating the matching between jobs and workers. They also differ in the tools they offer to workers and job providers to express needs and requirements, and in their compliance with labor-related regulations and their handling of ethical concerns. 

Human Factors at Work. FoW will witness an evolution of humans from being mere agents whose efforts and capabilities are used for the benefit of AI systems (broadly machines or business) to understanding psychological characteristics, attributes, skills, motivations, goals, etc. This will require capturing uncertainty in human behavior and individuals’ evolving needs, and adapting offers to available demand.  Additionally, the attitudes, values, opinions regarding the processes, policies, and outcomes (e.g., perceptions of justice, fairness, bias, etc.) will need to be assessed and considered in the design of FoW ecosystems.

Figure 1: A representation of human predictor variables at work and key outcome metrics

Intellectual challenges of FoW

We envision that the FoW will be increasingly technology-driven and will have the potential to bring human concerns to the center of the design and deployment of job platforms. It is therefore crucial to keep ahead of this trend by addressing the intellectual challenges that arise. 

IC1: Capturing Human Capabilities. The first challenge relates to the ability to understand different types of human roles in future jobs, modeling the inherent uncertainty in human behavior by understanding their evolving characteristics and be able to propose jobs to them by adapting to their changing perceptions and needs.

IC2: Stakeholder Requirement Specification. The ability for all stakeholders, employers, workers, regulators, platforms, involved in the new job ecosystem to specify their requirements by providing declarative and high level tools to express their needs and expectations.

IC3: Social Processes. The third challenge addresses the development of appropriate interaction methodologies and support for the new kind of workplace. This includes social processes relating workers, requesters and platform managers like onboarding, socializing, recognition and training, as well as ways to communicate and delegate work between humans and machines.

IC4: Platform Ecosystems and IC5: Computation Capabilities. FoW will require overhauling the design and engineering of online job platforms to enable the collection, storage, retrieval, analysis, and mining of a wide array of human data across different types of technology-driven work. A fair bit of engineering and testing will be needed to ensure the development of scalable and portable platforms and the integration of multi-stakeholder goals in efficient and effective ways.

IC6: Benchmark and Metrics. Benchmarking and the development of appropriate metrics to measure computing capabilities as well as human aspects such as satisfaction, human capital advancement, and equity, is another intellectual opportunity.

IC7: Ethics. Ethical considerations should drive FoW design and architecture. Privacy, compensation mechanisms and fairness are the main ethical issues of FoW. Privacy refers to the security of the data describing individuals, access to data, how data are being used, and consent. Compensation refers to various mechanisms including insurance, training and bonus, and the possibility for workers to choose among them. Fairness implies the detection and reduction of discrimination in decisions, especially in job allocation through auditing processes and compliance to regulations. Transparency, accountability and explainability must be studied for an ethical FoW.  

FoW Architecture

Human and AI workers operate the different components of the FoW architecture. Their characteristics are captured and updated over time (IC1). The requirements (IC2) come from all subjects, including regulators for policies, AI workers for job deployment and assignment, and human workers who are interested in specifying their needs (e.g., skill improvement and expected compensation). All subjects in the architecture interact and generate human-human, human-machine communication to enable social processes (IC3), as well as machine-machine communication for interoperability and portability (IC4).

The FoW agent, or platform, acts as coordinator/scheduler. Any subject can define work through specifications (IC2). Work is handled by the platform, which generates and assigns opportunities to the workers. Opportunities are executed and generate contributions that can be used in benchmarks and metrics (IC6).

All processes benefit from scalable computing capabilities (IC5). Processes are monitored to create logs of metadata used in two essential ICs: benchmarking purposes (IC6) and feedback to the subjects and verification of the compliance of the process with respect to the requirements (IC7).

Cross-Disciplinary Research

Addressing the intellectual challenges of FoW will require integration and convergence of disciplines across computer science, engineering, education and workforce training, and social, behavioral, and economic sciences, and law. The relevant research communities have been making some attempts to capitalize on human potential and improve workers’ well-being. For example, computational research has focused on developing piecemeal solutions for different components of job deployment such as skill estimation and task allocation. It would benefit from the ability to fully capture and account for changing human characteristics, and to provide support for different needs and preferences of all stakeholders. Psychologists, sociologists, and organization management researchers, on the other hand, have conducted user experiments and surveys, interviews, and physical laboratory experiments that have tested and contributed to a range of social science theories. Their contributions would greatly benefit from computational communities to develop quantitative models. It will be important for these different research communities to engage in collaboration and cross-talk in order to unveil and tap the full capabilities and potential of all subjects in FoW. Effective collaboration and genuine multi-disciplinary research will help approach the goal of optimizing AI systems while at the same time protecting and maximizing worker satisfaction, well-being, and performance potential.

We believe now is the time for a deeper integration of different research disciplines to synthesize the best of every research world to bring a transformative effect on FoW. For example, psychology, management, and organization studies research has to understand to what degree do workers’ perceived fairness and transparency affect the satisfaction of workers and/or employers? This question, and related ideas, can be tested at scale with computational methods, such as broad user surveys and more controlled laboratory experiments. Similar methodologies can be employed across different platforms, adapting to the specifics of each case. Computing communities, such as Data Management and AI researchers, on the other hand, have to focus on how to best design “human-centric” data management systems that leverage theories and practices in the social sciences. This convergence will help develop efficient storage, retrieval, protection, and analysis of human-data that changes over time, as well as design algorithms that adapt to evolving needs and preferences of workers. Such a perspective will be essential to understanding and shaping long-term social and economic outcomes of FoW, to fully exploit the potential of technology-driven FoW while maintaining a humane approach.

We are researchers in a wide variety of disciplines: Data Management, algorithms, Distributed Systems, and Psychology and Management. this blog post is the result of an intense week at Shonan. We would like to thank the Shonan organizers for hosting our workshop and providing ideal conditions for scientific exchanges.

Copyright @ 2019, All rights reserved.

Facebooktwitterlinkedin
874 views

2 Comments

  • Avatar Zakaria Maamar on September 24, 2019

    Hello,

    Is there a proceedings our of the workshop?

    Thank you, Zakaria

    • Avatar Atsuyuki Morishima on September 27, 2019

      We plan to publish several papers and also develop a Web site to record the summary of the workshop at https://humanmachinedata.org/FoW2019/ (it is just the list of participants at the moment).

Comments are closed

Categories