RIGOR - Towards reproducible, transparent, and valid AI methods for buildings and cities

The RIGOR project investigates the actual added value and scientific reliability of AI-based methods in the context of buildings, districts, and cities. It focuses on reproducibility, transparency, and the objective comparison of modern AI approaches with simple, robust baseline models. The goal is to establish an evidence-based foundation for the responsible use of artificial intelligence in energy- and safety-critical applications in the built environment.

Short Description

Background/Motivation

Buildings and cities play a central role in the energy transition and account for a substantial share of global energy consumption and CO₂ emissions. Digital energy services and AI-based methods are widely regarded as key technologies for improving energy efficiency, optimizing operation, and reducing emissions. At the same time, uncertainty is growing as to whether complex AI approaches actually deliver measurable benefits over established, simpler methods in real-world applications.

Methodical Approach

The RIGOR project addresses this challenge by systematically and critically assessing AI-based methods for buildings, districts, and cities. A central research question is whether modern AI models achieve statistically significant performance improvements compared to transparent and easily interpretable baseline models.

In parallel, the project examines to what extent current AI-based research meets fundamental scientific standards of reproducibility and methodological transparency. To this end, RIGOR analyzes scientific publications from recent years, conducts fully reproducible benchmarks, and evaluates data availability, code accessibility, and the clarity of reported results.

Expected Results

All outcomes are documented transparently and published in line with Open Science principles. A complementary mixed-methods approach involving stakeholders from academia, industry, and funding organizations helps identify structural barriers to reproducible research.

By doing so, RIGOR provides the first robust, empirical basis for assessing the real-world value of AI applications in the building sector.

The project supports more targeted use of resources, helps avoid inflated expectations, and promotes AI solutions that demonstrably contribute to energy efficiency, sustainability, and climate neutrality.

Project Partners

Project management

TU Wien, Institute of Building and Industrial Construction

Project or cooperation partners

TU Graz, Institute of Machine Learning and Neural Computation

Contact Address

TU Wien
Univ. Prof. Gerald Schweiger
Karlsplatz 13
A-1040 Wien
E-mail: gerald.schweiger@tuwien.ac.at
Web: https://github.com/tuw-isab