Like Circles in the Water: Responsibility as a System-Level Function

Abstract

What eventually determines the semantics of algorithmic decision-making in not the program artefact, nor-if applicable-the data used to create it, but the preparatory (enabling) and consequent (enabled) practices holding in the environment (computational and human) in which such algorithmic procedure is embedded. The notion of responsibility captures a very similar construct: in all human societies actions are evaluated in terms of the consequences they could reasonably cause, and of the reasons that motivate them. But to what extent does this function exist in computational systems? The paper aims to sketch links between several of the approaches and concepts proposed for responsible computing, from AI to networking, identifying gaps and possible directions for operationalization.

Accounting Value Effects for Responsible Networking

“…A fundamental distinction needs to be drawn between operational uses of responsibility, as in engineering, and responsibility as an attribute of moral or legal agents. In between the technical and sociolegal uses, we refer to the notion of agentive responsibility [10], deemed to take into account how the agent concurs and interacts with a certain outcome, without entering the debate of its "moral" characteristics. The agentive responsibility of an entity results from:…”