How Our Project Will Contribute To The Growth Of The Decentralized AI Platform

Our goal is to empower anyone and everyone to contribute to AI ethics governance, as well as provide resources towards developing responsible AI, such as financing, developer time, and computing and data resources.

Our Team

We have worked together cohesively on several projects (e.g., maintaining the SNET Archives;  building tooling such as https://archive-oracle.netlify.app/ & https://archives-dashboard.netlify.app/). Our complementary skills (developing knowledge graphs; open-source building: community engagement) & our shared skills (deep understanding of AI ethics; project management; research/documentation; strong connections in the sNET ecosystem) together make us eminently suited to delivering this proposal.

Company Name (if applicable)

The core problem we are aiming to solve

This project addresses the need for practical methods to ethically audit AI systems that use graph-based data and vectorisation techniques. It aims to detect and mitigate biases, enhance transparency, and ensure compliance with ethical standards and legal regulations (e.g., GDPR). Although it will be created and tested on a specific dataset, and will give particular focus to issues that arise around recordkeeping, it will develop a didactic, reproducible auditing framework applicable to community graph datasets in any other context, ensuring they are fair, accountable and ethical.

Our specific solution to this problem

We will develop an Ethical AI Auditing framework, with data & code samples provided at each stage, to audit the ethical effects of migrating a dataset to a graph-based system. Such a migration enables vector manipulation using generative models, allowing analysis of a records corpus without the need for exhaustive cataloguing and tagging; we will audit how possible it is to do this while maintaining the ethical integrity of the data. 

Our dataset is the sNET Archives https://snet-ambassadors.gitbook.io/singularitynet-archive , an open-source record of meetings and decisions in the SingularityNET Ambassador Program, that we have maintained since 2022. 

• Sample data in JSON formats will be stored in GitHub folders corresponding to each audit method to be applied, such as degree distribution, to check if node connections follow expected patterns; centrality measures, to identify influential nodes; & clustering coefficients, to assess community structures. Each method and dataset will be documented & tested in an associated GitHub folder for easy access.

• A Graph database (Neo4j) will be used to host the data model. The associated setup & Cypher queries will be documented on the project repo.

• The audit framework will be developed iteratively using the project's audit methods & community participation, and documented on a repo wiki for easy reference.

• Graph data will be optimized for vector models, with techniques such as counterfactual testing and human-in-the-loop anomaly detection.

Project details

Scope

This project will prepare community-sourced data for migration to a graph-based system and vector manipulation, developing an Ethical AI Auditing framework throughout the process. Auditing will focus on fairness, transparency, and compliance using human-driven techniques and Non-Reinforcement Learning from Human Feedback*, ensuring explainability and ethical alignment in AI models.

(* These methods focus on structured, manual, or algorithmic techniques to assess AI behavior for fairness, bias, transparency, and ethical compliance.)

Impact

1. Social Impact: We aim to ensure AI fairness, mitigate biases, and improve trust in automated decision-making and data analysis , particularly in the diverse communities that comprise the SingularityNet ecosystem.
2. Why It Matters: AI systems often reinforce systemic biases; this project provides a structured, human-audited approach to correct and prevent such ethical risks.

Ethics

The project prioritizes fairness, transparency, and compliance through manual auditing techniques (supported by algorithms), stakeholder engagement, and bias detection. It actively prevents discriminatory patterns by using visual inspection, statistical analysis, and counterfactual testing to ensure AI systems treat all groups equitably.

Safety

To mitigate risks, the framework incorporates privacy-preserving techniques, regulatory compliance (GDPR/CCPA), and stakeholder validation. A human-in-the-loop approach ensures AI outputs are continually reviewed and adjusted to prevent harmful recommendations.

Decentralized AI

By leveraging open-source tools and community-driven validation, the project promotes decentralization, allowing diverse stakeholders to contribute to and review AI decision-making systems, reducing reliance on centralized, opaque algorithms.

Innovation

Unlike traditional AI audits relying on RLHF or automated fairness metrics, this project introduces a multi-layered manual auditing framework that includes graph analysis, domain-specific rules, and counterfactual reasoning—offering a more adaptable and human-centered approach.

Technical Feasibility

The project is grounded in established graph-based auditing methods, using Neo4j and statistical techniques such as degree distribution, centrality, and clustering coefficients to evaluate AI decisions efficiently.

Community engagement and co-production

Throughout the project we plan to engage the BGI Nexus and SingularityNET communities in learning about our approach and giving their input on ethical practice, to ensure our audit framework co-produced. For example, in Milestone 3, the sNET Ambassador community (who created much of the Archives data by submitting summaries of their meetings, and therefore know the corpus quite well) will be invited to help with conducting initial auditing using visual inspection and rule-based checks. In Milestone 4, we will use open meetings to finalise our auditing framework and integrate stakeholder feedback and domain-specific rules, and ensure that our work addresses the community’s concerns on transparency, fairness and ethical alignment. And during Milestone 6, BGI Nexus and sNET community members will be invited to open workshops where they can learn how to use our ethical audit framework. In this way, we will ensure that our work addresses stakeholders’ concerns, and will offer opportunities for people to explore and learn about the issues raised, thereby increasing accessibility.

Team

Stephen Whitenstall (Project and Development Lead) :

Twitter: https://twitter.com/qa_dao 

Stephen is the co-founder of Quality-Assurance DAO, https://quality-assurance-dao.github.io/.

He founded the SingularityNet Ambassador Archive Workgroup in 2022.

He has provided project management consultancy for many Cardano Catalyst governance projects since Fund 4 including Catalyst Circle, Audit Circle, Community Governance Oversight, Training & Automation (with Treasury Guild) and Swarm. A Catalyst Circle V2 representative for funded proposers. He has 30 years experience in development, test management, project management, social enterprises in Investment Banking, Telecoms and Local Government. A philosophy honours graduate with an interest in Blockchain governance and benevolent AI.

Vanessa Cardui (Workshop co-ordination and Documentation): 

Vani is a community archivist, with 20+ years of experience in supporting marginalised communities to preserve and manage their own records - see for example https://creationofacommunity.wordpress.com/ and https://homeless.omeka.net/. She is deeply interested in the ethics of using AI to help us maintain and analyse records. In DeepFunding, she is a member of the DF Focus Group and the Onboarding Circle; and in the SingularityNET Ambassador Program, she leads the AI Ethics Workgroup and is a core member of the  Archives WorkGroup. She also has extensive experience of community engagement, peer education, and co-production.

André Diamond (Open Source Administration)

Andre is a developer who has worked in Cardano Catalyst since 2021 and in Singuarity Net since 2022. He has helped build the Treasury Tool used in the SNet Ambassador Programme and the SNet Archive WG dashboard. He maintains a wide range of Open Source GitHub Organisations and repos across SNet and Cardano. And has produced a range of technical and educational documentation.

Existing resources

• SingularityNET Archives Gitbook (https://snet-ambassadors.gitbook.io/singularitynet-archive)

• SingularityNET Archives dataset (see sample datasets here https://github.com/SingularityNET-Archive/SingularityNET-Archive/tree/main/Data)

Open Source Licensing

n/a

Links and references

• SingularityNET Archives Gitbook (https://snet-ambassadors.gitbook.io/singularitynet-archive)

• SingularityNET Archives dataset (see sample datasets here https://github.com/SingularityNET-Archive/SingularityNET-Archive/tree/main/Data)

Additional videos

Stalnaker’s Concept of Context and Its Relevance to AI Ethics - https://youtu.be/VhYfA0ehw3w

In this presentation, we discuss how Robert Stalnaker’s ideas—like the role of presuppositions, the dynamic nature of context, and the importance of common ground—can guide us in designing ethical and context-aware AI systems.

Was there any event, initiative or publication that motivated you to register/submit this proposal?

Describe the particulars.

Opportunity to leverage the SinguarityNet Ambassdor's Archive