Intepretable AI from start to finish

Purpose and goal
The technology with interpretable AI shall be used to provide transparency in production processes, for AI-based products and services, to understand, verify and validate the functionality of these, which is highly essential for AI development and use in Sweden. Methods and tools must be developed to support a process from identifying the needs of users with AI and data to the delivery and reuse of products containing AI components. The technology must also be compared with AI which is based on so-called black box models to study the pros and cons of the techniques.

Expected effects and results
The project intends to produce knowledge about how functionality in products can be created with AI in a way that can be interpreted and validated by those involved in the product development process. The results will facilitate the changing field of use of AI-based products in re-use and re-manufacturing. The developed technology must enable interpretable optimization of AI-based properties of products that have a direct connection to sustainability-affecting properties described in Agenda 2030.

Planned structure and implementation
The project is organized around two practice cases. Use of vessel progress data to offer better propulsion systems and processing of data from airborne laser scanning of forests to discern interesting details in the stock. Methods and tools must be developed to support a process from identifying the needs of users with AI and data to the delivery and reuse of products containing AI components. Research actors are RISE, Chalmers' departments for mechanics and maritime sciences and Chalmers' department for industry and materials science.