Narrate | Trustworthy and ethical conversational AI

Region:

Europe

Making AI conversations more transparent, responsible and user-centred

Conversational AI is rapidly transforming how people access information, interact with services and make decisions. However, concerns around bias, transparency, privacy and reliability continue to limit adoption in many organisations. To address these challenges, Narrate develops ethical, trustworthy and personalised conversational AI solutions that combine advanced language technologies with strong user-centered and privacy-preserving principles.

Target group

The project results are relevant for organisations looking to deploy trustworthy conversational AI systems, including:

Media organisations and content providers
Human resources and talent management services
Engineering and software service providers
Developers of AI-powered digital services and platforms
Public and private organisations seeking responsible AI adoption

Context

The emergence of large language models (LLMs) and generative AI (GenAI) tools has made conversational AI accessible to a broad audience. Users increasingly expect personalised, interactive and natural interactions with digital systems.

However, current GenAI systems rely on probabilistic models that can generate inaccurate, biased or non-transparent responses. While these technologies offer significant opportunities, they still struggle to consistently uphold values such as fairness, accountability, honesty and privacy. For organisations, this creates uncertainty around the adoption of conversational AI. Many are looking for solutions that not only provide relevant and personalised responses but also meet ethical requirements and regulatory expectations.

At the same time, different application domains have highly specific requirements that general-purpose AI models cannot always address effectively. This creates a need for domain-specific approaches that combine technological performance with trustworthiness and responsible AI practices.

Objectives & results

Within the Narrate project, supported by the Brussels Capital Region and Innoviris, Sirris and 3E investigate AI-driven pipelines for the reliable (semi-)automatic extraction of technical information from complex, heterogeneous industrial documents. Narrate brings together 14 partners from 5 countries with the goal to develop trustworthy, ethical AI-driven human interaction interfaces with multi-modal data integration and automated knowledge updates.

The project will:

Assess the feasibility and effectiveness of domain-specific language models for targeted applications
Investigate how specialised AI models can be combined with large-scale general-purpose language models
Develop a flexible platform that enables the integration and interoperability of AI components and conversational agents
Design personalised conversational interfaces tailored to specific user needs and contexts
Implement mechanisms that promote ethical AI behaviour, transparency and user trust
Validate the approach in the domains of media, human resources and engineering software services

Approach

The project is structured around four main workstreams.

1. Domain-specific AI models

The consortium evaluates the benefits of specialised language models and investigates how they can complement existing large-scale AI systems in different application domains.

2. Platform development

A scalable and flexible platform is developed to enable the integration, orchestration and interoperability of conversational AI technologies.

3. Ethical and personalised conversational interfaces

The project develops user-centered conversational interfaces and investigates narrative design approaches that enable personalised, context-aware and ethical interactions.

4. Validation and business development

The developed technologies are validated through real-world use cases in media, human resources and engineering software services. The project also explores sustainable business models and deployment strategies.