Fairness Finder is a collaborative hackathon project with the purpose to develop a prototype application designed to assess and promote fairness for data workers in participative projects, volunteering initiatives, and online gigs. The solution will leverage the FINDHR toolkit to ensure non-discriminatory practices.

Current Situation

Data workers often face challenges related to unfair working conditions, unequal treatment, or discriminatory behavior in participative projects. The current landscape includes manual monitoring processes, which are subjective and not always effective. The exploration of machine learning and AI can bring about solutions to automatically monitor and penalize discriminatory practices, but these also carry risks of bias.

The toolkit can give us a framework to develop a system that ensures ethical data processing while being technologically grounded to promote fairness. Solution by Fairness Finder is aimed at providing a proof of concept that can offer adhoc, operational fairness metrics for participatory data work interactions. Based on it, we can develop a prototype app that assesses and promotes fairness in participative projects, volunteering initiatives, and small-to-medium enterprise (SME) job opportunities, leveraging the FINDHR project's toolkit to ensure non-discriminatory practices.

The FINDHR (Fairness and Intersectional Non-Discrimination in Human Recommendation) project has developed a comprehensive toolkit to address algorithmic discrimination in hiring processes. This toolkit includes methods, algorithms, and training materials to create fairer AI-assisted recruitment systems. The challenge is to extend these principles beyond traditional hiring to various participative activities.

Goals

Participants will create a user-friendly mobile or web application with the following features:

1. Fairness Assessment Module:
   • Design an algorithm that evaluates the fairness of participative projects, volunteer programs, or SME job listings. Consider factors like diversity, inclusion, and equal opportunities.
   • Utilize the FINDHR toolkit's fairness metrics and bias detection techniques to analyze project descriptions, volunteer requirements, or job postings for potential discrimination.
2. Recommendation Engine:
   • Develop a system that suggests improvements to make these opportunities more inclusive. For instance, provide recommendations to project organizers or employers on how to attract a diverse range of participants or applicants.
   • Implement the FINDHR pre-processing fairness interventions to ensure the app's suggestions are unbiased.
3. User Feedback and Reporting:
   • Create a feedback mechanism for users to report any perceived discrimination or unfair practices in the listed opportunities.
   • Incorporate a secure multi-party computation (SMPC) system to handle sensitive user data, ensuring privacy and compliance with data protection regulations.
4. Educational Component:
   • Include an interactive tutorial or guide within the app to educate users about algorithmic fairness, discrimination, and the importance of diversity in participative projects and workplaces.
   • Draw content from the FINDHR training materials and research reports to provide practical examples and insights.

Further outputs can include:

• A working prototype of the app.
• A presentation demonstrating the app's functionality and its potential impact on promoting fairness.
• A technical documentation outlining the implementation of FINDHR principles and algorithms.
• Data collection: Gather and preprocess diverse datasets from online labor platforms, community projects, and volunteering databases.
• System design: Use the FINDHR toolkit for fairness metrics. Develop a web application or an API that can integrate with platform APIs.
• Iterative prototyping: Test the application with a diverse set of users to ensure that the fairness metrics are useful and actionable.
• Feedback loops: Establish a system for feedback from users and stakeholders to understand which fairness metrics are most valuable and how to prioritize them.
• Communication: Validate the prototype through discussions with relevant stakeholders, including users, to ensure that the application effectively promotes fairness in data work environments.
• Documentation: Write clear guidelines and best practices based on lessons learned, enabling future developers and users to understand and adopt the technology.

This challenge aims to encourage participants to apply the FINDHR project's insights to real-world scenarios, fostering innovation in fairness-aware technology and contributing to a more inclusive digital society.

Resources

• AlgorithmWatch Toolkit (PDF)
• Weval: Latent Discrimination in Hiring Score
• FINDHR Toolkit: https://findhr.eu/toolkit
• FINDHR Research Reports: https://findhr.eu/findhr-expert-reports/
• FINDHR Python Package for bias mitigation: https://github.com/findhr/findhrAPI

Besides this we can take advantage of common resources from the hackathon:

• AI Models and Infrastructure: Leveraging access to pre-trained fairness models and cloud computing resources can facilitate rapid prototyping and testing.
• Platform APIs: Accessing APIs from various platforms (e.g., crowdsourcing, open volunteer platforms, gig economy platforms) for near-real-time data retrieval and analysis.
• Data: Utilizing publicly available datasets that reflect diverse work environments and jobs (like anonymized user data from large platforms, or custom research data from related fields).

Outcomes

This project promotes open science and social responsibility by releasing a prototype that can be extended, customized, and improved upon by others. Fairness Finder aims to catalyze broader discussions and projects focused on ethical AI, fair labor practices, and inclusive development of technologies.

• Open Science: The project will release the prototype and source code, along with documentation on design choices and assumptions used, under open licenses.
• Public Policy: By promoting fairness in digital workspaces, Fairness Finder aligns with Swiss AI Initiative goals to harness AI for societal good, contributing to conversations around digital labor rights and inclusive innovation.
• Strategic Importance: For Switzerland, this project contributes to the development of AI that respects human rights and actively counteracts discrimination. Globally, it offers a model that can be adapted to various contexts, particularly where digital platforms mediate work and contributions.

Develop clear interfaces and explain the fairness measures and their implications. Users should understand how their data is being used, especially in the context of an application aimed at promoting fairness.

By starting with these steps, the Fairness Finder team will be well-positioned to tackle this challenge effectively. The project stands to make a tangible contribution to fairness in digital work by integrating cutting-edge AI fairness tools into practical solutions.

Further steps should also consider knowledge from the ethics and legal guidelines provided in the Swiss {ai} Weeks and engaging with legal experts where necessary to ensure the project's fairness and respect for ethical standards from the outset.

The goal is to create a solution that is not only technically sound but also ethically sound and practically usable in real world contexts, enhancing fairness and safeguarding against unintentional bias in how data work is organized and valued.

The prototype should be delivered in a way that is open and inviting to future development and possibly integration into actual platforms, making fairness assessment a standard, visible part of participatory projects and digital workplaces.

In conclusion, Fairness Finder represents a critical stepping stone towards enabling a more equitable and ethical use of AI in mediating and managing digital work, by creating practical tools anchored in advanced AI fairness frameworks. The project will not only benefit current practices in Switzerland but also inspire similar initiatives worldwide, addressing the growing need for equitable and transparent platforms in the digital economy.

🅰️ℹ️ Written with help of APERTUS-70B-INSTRUCT