The platform employs a participatory model for content verification, allowing users to vote on information trustworthiness and influence the reputation of content creators and disseminators. This approach is supported by a local app that performs content scoring and visualization, ensuring user engagement. EUNOMIA’s development incorporates co-design methodologies, integrating user experiences and trust perceptions in social media use.

EUNOMIA’s utility has been assessed across large communities in social journalism, traditional media, and leading decentralized social media networks. The evaluation demonstrated EUNOMIA’s versatility in enhancing information trustworthiness and democratizing content verification processes.

EUNOMIA represents a significant advancement in tackling misinformation and promoting transparency in social media. By decentralizing the verification process and prioritizing user experience and ethical standards, EUNOMIA offers a viable solution for fostering informed and trustworthy digital communities.

The EUNOMIA architecture comprises of the following components to ensure its secure, decentralised nature:

1. a P2P network
2. a blockchain infrastructure
3. a security and privacy framework

Furthermore, EUNOMIA architecture comprises of the following components that are exposed as tools with the aim of assisting social media users in assessing information trustworthiness:

1. a human-as-trust-sensor component;
2. a social media content and context data analysis component;;
3. trustworthiness scoring
4. a user application in the form of a browser extension known as the Digital Companion (DC).

The Peer-to-Peer (P2P) infrastructure facilitates the storage and communication services within the EUNOMIA ecosystem, catering to two main entities: service providers (EUNOMIA services) and end-users (EUNOMIA digital companion). This P2P network is characterized by several key features:

• It allows any individual to join the decentralized network autonomously, without oversight or control from any centralized organization.
• The network’s capacity and resource availability increase proportionally with the number of users, thanks to their contribution of resources, enhancing the network’s overall efficiency and robustness.
• Due to its distributed architecture, the network experiences significantly fewer failures, enhancing reliability and uptime for all users.
• It supports a range of third-party services, including distributed storage solutions, peer-to-peer messaging (specifically within the Digital Companion), and comprehensive network management functionalities.

These features collectively underscore the network’s ability to provide a resilient, scalable, and user-empowered platform for decentralized communication and storage solutions.

Our research contributions to project EUNOMIA include:

• Blockchain Infrastructure Development: the design and implementation of the blockchain infrastructure that forms the core of EUNOMIA’s decentralized, secure environment. This infrastructure supports transparency, data integrity, and facilitates the trustworthiness scoring mechanism.
• Content and Context Analysis: the development of techniques that enhance the capability to analyze social media content and context, enabling sophisticated sentiment and subjectivity analysis with the use of machine learning models. This work supports the project’s goals of empowering users with the capability to assess information credibility effectively.
• Enhancing User Participation and Security: the orchestration and integration of all technical components that support the EUNOMIA platform.

This section summarizes the technical characteristics and functionalities within EUNOMIA’s blockchain component, outlining its core infrastructure, support for authorization and authentication services, governance implementation, and integration capabilities.

The blockchain infrastructure within EUNOMIA is utilizing the Hyperledger Fabric – a sophisticated enterprise-grade permissioned blockchain framework. The blockchain component in EUNOMIA is essential for ensuring data integrity, supporting decentralized governance, and facilitating secure and transparent interactions within EUNOMIA’s ecosystem. Hyperledger Fabric in EUNOMIA uses the RAFT protocol (a Crash Fault Tolerance (CFT) consensus algorithm). The network setup is permissioned and is governed by a governance body centred around the formation of a foundation. Organizations that participate to the foundation can participate to the decision making of the network by voting.

• Ledger Service
• Instantiates the Blockchain Infrastructure (BI)
• Provides blockchain API
• Blockchain enhances data integrity
• Storage Server
• Consumes blockchain API
• Keeps references of OrbitDB objects only
• Consumes P2P storage API
• P2P storage keeps the actual data
• Voting and governance service

EUNOMIA’s blockchain infrastructure incorporates technical components crucial for implementing comprehensive authorization and authentication services. These services are designed to manage access control and verify the identity of users engaging with the network, ensuring that interactions are secure and that entities are duly authorized. The framework employs a Public Key Infrastructure (PKI) to manage digital certificates and public-key encryptions, facilitating the authentication of users and services within the network. This mechanism ensures that only legitimate users and nodes can participate in the network, enhancing the overall security posture of the EUNOMIA ecosystem.

Privacy-related features:

• Blockchain infrastructure stores metadata (not actual data) for each post.
• Blockchain used as a trust engine that builds on decentralization.
• Anonymization of the data stored on the storage server.
• Users have the right to be forgotten (removal of their private key).
• Data on the storage server are encrypted.

The governance of EUNOMIA’s blockchain network supports the orchestration of consensus among consortium members, facilitating decision-making processes related to network updates, policy changes, and the integration of new participants (e.g., nodes to be added to the blockchain network). The governance framework is operationalized through smart contracts (in chaincode on Hyperledger Fabric), which encode the rules and procedures for network governance, ensuring that changes to the network or its protocols are executed transparently and with consensus from the consortium members.

• EUNOMIA comprises of multiple organizations.
• Orderer nodes participate to the blockchain network consensus
• Each organization can launch as many peers as it wishes.
• Each organization should maintain an Orderer node.
• Network participation is controlled by the Certificate Authority (CA) servers

Orchestrating the integration of EUNOMIA’s blockchain infrastructure with other project components has been a critical aspect of the technical architecture. EUNOMIA’s service node requires interfacing with various components such as: the distributed peer-to-peer (P2P) file storage, the digital companion application, and other service nodes that facilitate user interaction and content verification. The API service layer that implements EUNOMIA’s services has been crucial role in this context, acting as an abstraction layer that simplifies the interaction between the blockchain network and other components of EUNOMIA. This layer enables seamless communication and data exchange, ensuring that the diverse functionalities of the EUNOMIA ecosystem operate cohesively and efficiently.

In conclusion, the technical characteristics of the EUNOMIA project’s blockchain infrastructure represent a practical approach to leveraging blockchain technology for enhancing the integrity, security, and decentralization of social media platforms. EUNOMIA although a research project paves the way for a new paradigm in the fight against misinformation and the promotion of user empowerment in digital media environments.

The EUNOMIA project introduces an innovative Sentiment Analysis Component, a trustworthiness indicator for social media posts utilizing a new deep learning approach that surpasses current methods. It employs an adapted rule-based method for confidence calculation and an ensemble technique that merges deep learning with rule-based methods for enhanced accuracy.

Additionally, the Aspect Terms Extraction Component enables fine-grained sentiment analysis and topic modeling, outperforming existing state-of-the-art solutions. These components collectively enhance the platform’s capability to assess and score the trustworthiness of user-generated content on social media.

References:

– P. Agathangelou, I. Katakis, I. Koutoulakis, F. Kokkoras, and D. Gunopulos, “Learning Patterns for discovering domain-oriented opinion words,” Knowledge and Information Systems,2018. (not part of EUNOMIA)

– P. Agathangelou and I. Katakis, “A hybrid deep learning network for modelling opinionated content,” in Proceedings of the 34th ACM/SIGAPP Symposium on Applied Computing, ser. SAC ’19. New York, NY, USA: ACM, 2019, pp. 1051–1053.

– P. Agathangelou and I. Katakis, “Balancing between Holistic and Cumulative Sentiment Classification”, submitted at the Online Social Networks and Media Journal.

This component is leveraging human intelligence for the assessment of information trustworthiness on social media platforms. The intuition behind this component is predicated on the understanding that human users, through their subjective analysis and interactions, can provide insights into the credibility of content. The HaTS functionality is designed as follows:

1. Data Collection and Structuring for Human Analysis: The HaTS approach necessitates a methodical system for gathering and organizing data in a manner conducive to human appraisal. This involves the aggregation of social media posts, their metadata, and subsequent interactions, structured in a user-friendly format that facilitates easy interpretation. The objective is to present the data in a context that highlights its origin, the trajectory of its dissemination, and any modifications it may have undergone. Such a structured presentation allows users to intuitively assess the reliability of the information, taking into account its provenance and evolution over time.

2. Interactive Visualization and User Participation: A critical component of the HaTS methodology is enabling user interaction with the visualized information, primarily through mechanisms like voting. This participatory model enables users to contribute to the trustworthiness assessment of content by allowing them to endorse or challenge the credibility of a post. The interactive platform not only facilitates this direct engagement but also collects and aggregates these user contributions to generate a comprehensive trustworthiness score for each piece of content.

3. Information Cascade and Trustworthiness Assessment: The collective user interactions and evaluations form an information cascade, a dynamic flow of data that enriches the content with layers of user-generated trustworthiness attributes. This cascade is instrumental for understanding the origin and alterations of posts, offering a multifaceted view of their credibility. Users’ contributions are quantified into metrics that reflect the community’s consensus on the content’s trustworthiness, encapsulating the essence of the HaTS function. These metrics serve as a tangible representation of the content’s perceived reliability, distilled from the collective judgment of the user community.

EUNOMIA’s approach is user-centric, incorporating dynamic algorithms and machine learning to derive trustworthiness assessments to individual user preferences. This methodology is encapsulated in the Trustworthiness Nudge module, a sophisticated tool integrated within the digital companion application, designed to aid users in discerning the reliability of social media content.

Trustworthiness Indicators and User-Centric Configuration: EUNOMIA identifies various trustworthiness indicators that encompass different dimensions of information credibility. Recognizing the subjective nature of trust, the project allows users the flexibility to select which indicators are most significant to them. Users can customize how these selected indicators are weighted in the computation of an aggregated trustworthiness score for each post, a process that occurs locally on the user’s device. This personalized approach ensures that the trustworthiness assessment aligns with the user’s individual criteria and values, enhancing the tool’s relevance and utility.

Evolution of the Trustworthiness Nudge Module: The Trustworthiness Nudge module offers a dynamic technique based on machine learning for automatically learning user trustworthiness preferences. Initially, the module enabled users to manually select trustworthiness features, which the system then used to flag social media posts requiring further scrutiny. This static method provided a foundational level of user engagement and customization.

Integration of Machine Learning for Automatic Feature Extraction: The project’s subsequent phases introduced a machine learning engine to the Trustworthiness Nudge, significantly expanding its capabilities. This engine automatically extracts trustworthiness features related to social media posts and their authors, such as metadata and contextual data. By analyzing a user’s history of trust votes on posts, the machine learning module is capable of locally training a model that reflects the user’s trustworthiness preferences. This dynamic adaptation enables the Trustworthiness Nudge to learn and evolve with the user, offering increasingly personalized and accurate trustworthiness assessments over time.

EUNOMIA is designed to be compatible with any social media platform in theory. For its initial pilot, EUNOMIA was implemented on Mastodon, serving as the primary environment for prototyping. The decentralized nature of Mastodon allowed EUNOMIA to deploy and showcase its misinformation-combating tools without accessing or owning any live user data.