Sociotechnical Envelopment of Artificial Intelligence: An Approach to Organizational Deployment of Inscrutable Artificial Intelligence Systems

The paper presents an approach for implementing inscrutable (i.e., nonexplainable) artificial intelligence (AI) such as neural networks in accountable and safe manner organizational settings. Drawing on exploratory case study the recently proposed concept of envelopment, it describes a organization successfully “enveloping” its AI solutions to balance performance benefits flexible models with risks that can entail. authors present several envelopment methods—establishing clear boundaries within which is interact surroundings, choosing curating training data well, appropriately managing input output sources—alongside their influence choice organization. This work makes two key contributions: It introduces sociotechnical by demonstrating ways organization’s successful depends interaction social technical factors, thus extending literature’s focus beyond mere issues. Secondly, empirical examples illustrate how operationalizing enables manage trade-off between low explainability high presented models. These contributions pave way more responsible, implementations organizations, whereby humans gain better control even machine-learning