About the ETIA project

ETIA is a European research and innovation project developing a new generation of Causal AI tools for data-driven insight and decision-making.

Its goal is to make advanced causal discovery, modelling and reasoning accessible beyond a small community of specialists.

Most data analysis today is built around prediction. It can identify patterns, associations and likely outcomes. But when a decision needs to be made, another question becomes critical:

Why is this happening?

ETIA is being developed to help researchers, analysts and organisations move from identifying correlations to understanding plausible causal relationships, exploring what-if scenarios, finding root causes and making better-informed decisions.

The name ETIA comes from the Greek word αιτία: cause or reason.

Correlation can tell us what moves together. Causality asks why.

Modern machine learning can be extraordinarily good at prediction. But predictive relationships are not automatically causal relationships.

Two variables may appear strongly related because they are both influenced by something else. Acting on one of them may therefore have little or no effect on the outcome we actually want to change.

This distinction matters whenever data is used to support decisions.

A healthcare researcher may want to understand what contributes to an outcome. A telecommunications company may want to know which network parameters actually affect performance. An industrial organisation may need to identify the root cause of a failure. A business may want to understand which actions genuinely influence customer behaviour.

In each case, prediction is useful, but it does not answer the whole question.

Causal models can support a different class of analysis:

  • identifying plausible causal relationships;
  • estimating the effects of possible actions;
  • exploring what-if scenarios;
  • finding root causes;
  • comparing alternative decisions;
  • reasoning about what might have happened under different circumstances.

The challenge is that causal analysis remains technically difficult. It often requires specialist knowledge of statistics, machine learning, causal theory and software engineering.

ETIA aims to change that.

Machine learning systems typically learn how variables are associated with an outcome.

ETIA is being developed to go further by automating the process of discovering and analysing causal relationships in data.

Instead of requiring users to manually select algorithms, configure pipelines and interpret complex causal models, ETIA will bring these capabilities together within a unified platform.

The project combines three main elements:

Automated Causal Discovery

The ETIA AutoCD engine is designed to automate key stages of causal analysis, from identifying relevant variables to selecting causal models and reasoning over their relationships.

Causal reasoning

ETIA will support analyses such as causal effect estimation, what-if simulations, decision optimisation, root-cause analysis and counterfactual reasoning.

Accessible interfaces

Different users will be able to interact with ETIA through a graphical interface, programmatic APIs and a Causal AI Agent capable of supporting natural-language interaction and explaining results.

The ambition is simple:

to make sophisticated causal analysis easier to use, easier to interpret and applicable to real-world scientific and industrial problems.

ETIA is the result of a long research trajectory in causal discovery and automated machine learning.

Its foundations were developed through two previous European Research Council projects.

CAUSALPATH

The CAUSALPATH ERC Consolidator Grant advanced the theory and algorithms required for causal discovery across increasingly complex datasets and real-world conditions.

Among its contributions was work addressing one of the fundamental barriers to automated causal discovery: how to select and tune an appropriate causal discovery pipeline when there is no known answer against which the resulting model can simply be validated.

This research led to the development of the Out-of-Sample Causal Tuning (OCT) approach.

AutoCD

Building on these foundations, the AutoCD ERC Proof-of-Concept project moved from research toward implementation.

The project designed, constructed and validated the first ETIA prototype, including its core software architecture and fundamental automated causal discovery capabilities.

ETIA was subsequently made available as an open-source software library, accompanied by synthetic data generation tools and documentation.

The prototype has also been applied through collaborations involving real-world challenges in areas including telecommunications, automotive safety and data-driven decision-making.

The current ETIA project takes the next step: transforming this body of research and the existing prototype into a robust, accessible and validated Causal AI platform.

Making causal analysis accessible

ETIA’s long-term vision is to help turn causal discovery and reasoning from a highly specialised analytical process into a broadly accessible capability.

The project is working towards a platform that can support both technical experts and users who may have little or no specialist knowledge of causal discovery.

Rather than asking users to understand which algorithms to select or how to configure complex analytical pipelines, ETIA is being designed so that users can focus on the questions that matter to them.

Questions such as:

What is causing this outcome?

What could happen if we change this?

Which decision is most likely to produce the outcome we want?

What is the root cause of this failure?

What might have happened if a different decision had been taken?

By making these questions easier to explore with data, ETIA aims to support better scientific insight and more informed decision-making across multiple domains.

ETIA combines technological development, validation and commercial preparation through six core objectives.

01. Build the ETIA AutoCD engine

Develop a general-purpose automated causal discovery engine capable of learning plausible causal models from different types of data and supporting a wide range of causal questions.

The project will extend ETIA to work across increasingly realistic conditions, including time-series data, mixed data types, observational and interventional data, and existing causal knowledge.

02. Make causal analysis accessible

Develop different ways of interacting with ETIA for different types of users.

These will include:

  • a Causal AI Agent for natural-language interaction;
  • a graphical user interface for visual analysis;
  • APIs for developers, analysts and power users.

The aim is to make advanced causal analysis usable without requiring every user to become a causal discovery specialist.

03. Build for technical robustness

Create continuous testing and benchmarking capabilities to assess the quality, reliability and performance of ETIA as the technology evolves.

The project will develop a dedicated causal benchmark corpus and mechanisms for tracking data, models, algorithms and software versions to strengthen reproducibility and provenance.

04. Validate ETIA in real-world environments

Test and refine the technology through real industrial and scientific use cases.

ETIA will be evaluated across multiple application areas, generating feedback about its usefulness, usability, analytical performance and requirements for future development.

05. Prepare ETIA for the market

Develop the foundations required to take ETIA beyond the research project.

This includes work on:

  • the business model;
  • go-to-market strategy;
  • product roadmap;
  • intellectual property strategy;
  • investment readiness;
  • regulatory and certification considerations.

The project also foresees the creation of a new legal entity to support future commercialisation.

06. Build an open community around causal discovery

ETIA will actively engage researchers, developers, analysts, organisations and other potential users.

Part of the technology will be made available to support experimentation and community participation, while the project will encourage contributors to develop new algorithms, modules and extensions.

Scientific publications, software releases, events, workshops and other dissemination activities will help grow the wider community around Automated Causal Discovery.

By the end of the project, ETIA aims to have moved significantly beyond its current proof-of-concept stage.

Expected outcomes include:

A validated ETIA platform

An advanced AutoCD engine, Causal AI Agent, graphical interface and developer API tested across realistic application scenarios.

A causal benchmarking framework

A growing corpus of causal datasets and models, together with infrastructure for continuously evaluating algorithms and system performance.

Real-world validation

Application of ETIA to industrial and scientific problems to demonstrate its capabilities, identify limitations and guide product development.

New scientific knowledge

New causal discovery algorithms, scientific publications and research outputs generated through the development and application of ETIA.

An active user and contributor community

A growing ecosystem of researchers, developers and organisations using, testing and contributing to ETIA.

A pathway to commercialisation

A business model, product roadmap, go-to-market strategy, intellectual property strategy and investment plan to support the next phase of ETIA’s development.

A new European deep-tech venture

The project foresees the creation of a new legal entity to take ETIA towards the market and continue its development after the EU-funded project concludes.

ETIA is a 36-month project, progressing from technology development through validation and preparation for market deployment.

Months 1–12

Build the foundations

Development of the initial AutoCD engine, Agent, GUI and API, technical robustness infrastructure, project website, dissemination programme and first business plan.

Months 12–24

Expand, test and prepare

Further development of ETIA’s causal capabilities, benchmarking and technical robustness, intellectual property planning and establishment of the new legal entity.

Months 24–30

Validate in real-world environments

Testing and validation of ETIA across industrial use cases, combined with user feedback and continued technological refinement.

Months 30–36

Advance towards market readiness

Completion of the advanced ETIA platform, final technical assessment, industrial validation, business model, commercialisation strategy and investment preparation.

ETIA is designed as a general-purpose technology.

Causal discovery can potentially support any field in which organisations or researchers collect data and need to understand not simply what is associated with an outcome, but what may be driving it.

Potential applications span areas including:

Telecommunications · Automotive safety · Healthcare · Bioinformatics · Marketing · Industrial systems · Scientific research

Across these domains, the underlying question remains the same:

What causes what, and what can we learn from knowing why?

ETIA: Causal AI for Data-Driven Insights and Optimal Decision Making is developed within the Horizon Europe Programme under the European Innovation Council (EIC) Transition framework.

Grant Agreement: 101291665
Project Coordinator: University of Crete

ETIA builds upon research developed through:

CAUSALPATH
ERC Consolidator Grant 617393

AutoCD
ERC Proof-of-Concept Grant 101069394

The project brings together research in causal discovery, automated machine learning, artificial intelligence, software engineering, technology transfer and commercialisation to advance European deep-tech research towards real-world application.

Join the Community

The ETIA project is building an open community of researchers, developers, industry experts and curious minds working to make causal discovery more accessible, robust and useful in the real world. Whether you want to contribute ideas, test the technology, explore new applications or simply follow the journey, there is a place for you here.

Join us in moving from correlation to understanding.