The follwing guidelines are applied to the design, development, and maintenance of INFORM’s systems, regardless of which AI methods they rely on. Thereby, we aim to build trustworthy software. We aim to uphold high standards in engineering the algorithms underlying our systems and engage in their responsible application and design.

As part of this, it is relevant to distinguish between different applications of AI. As an example, recommender systems which are often used in e-commerce are usually uncritical. Decision support systems often fall into a similar category. In contrast, in consequential applications of AI, an AI system’s decisions will have a more significant impact, e.g. affecting people at a personal level. For example, when AI-based algorithmic decisions are not only used as recommendations or decision support, they may also directly impact workers’ lives. Some AI systems even need to be considered life-critical, e.g., in medical applications or autonomous driving. Most INFORM systems can be categorized as recommender or decision support systems, but some of them have a higher impact.