EXPLAIN: Explanatory Artificial Intelligence for Industry

The aim of EXPLAIN is design and evaluate an interactive and explainable end-to-end Machine Learning (ML) lifecycle for industrial domain experts and end users. The research focus on human factors and ML algorithmic challenges, including:

1. How to investigate and establish a good understanding of experts’ mental models to identify relevant explanation mechanisms specific to the stage of the ML life cycle, the user role, and application context.
2. How to enable users to effectively interact with explanations and the high dimensionality of the industrial data.
3. How to generate and provide high-quality, consistent, machine amenable feedback to operators, taking into consideration workflows perceptions, effort, incentive, and trust.
4. Derive robust explanation methods that produce meaningful explanations in a reliable and reproducible manner based on the data types relevant for industrial applications.
5. Exploring novel approaches to ML training that allow domain expert feedback based on explanations.

EXPLAIN will in several industrial uses cases develop demonstrators of AI support systems that incorporate ML models and explanatory algorithms. In Sweden, the two industrial uses cases are provide by Boliden in mining, and Södra in pulping. Novel interface prototypes will be designed to facilitate effective interaction between ML algorithms and industry experts. Several innovation cycles will drive prototype development, with each cycle being validated with target users. The project will deliver human factors and algorithmic guidelines, and best practices for the design of ML solutions to the industrial domain, focusing on how to effectively engage with domain experts and end users.

The EXPLAIN project is an international collaboration with 18 partners from three countries (Sweden, Netherlands, Germany). The project runs for three years between 2022-2025. EXPLAIN is a EUREKA Clusters AI research project funded by VINNOVA in Sweden.