Polls, Discussion topics, Debates

A.E.: "a discipline of applying Ethical Decisions in all life-cycle phases of automated and intelligent systems including the design, development, operation, maintenance and end-of-life phase." I included"all life-cycle phases" and added the critical end-of-life phase. The term "Ethical Choice" is not intuitively clear. I like to imply with "Ethical Decisions" that it is a "must" to apply ethics. It also carries in its name, that there is not a unique solution, which fits all AI systems.

Suggestion for rephrasing the definition, in order to avoid saying 'ethics is a discipline of applying':

"Algorithm Ethics is the systematic application of Ethical Choice during the design, deployment and maintenance phase of automated and intelligent systems"

In the disciplines, I would add "explainability"

I'm concerned about the use of 'discipline' and 'phases' and 'context'. I think we can simplify while still covering more ground.

I suggest the inclusion of decommissioning in amongst the algorithm lifecycle. Provided Ethical Choice includes a consciousness of technical, social and cultural bias and an understanding of the context of such "choice" or at least the application domain of the algorithm, then this definition should work.

A. Why "in the context of"? It makes sense in the ethical choice definition, but here it doesn't. They're stages of the process. B. Are they all the stages? Shouldn't we just write all stages? C. I'd add an Oxford comma D. I'm not sure about the sub-disciplines, are we finalizing that now as well? Can we discuss it separately?

I'm only skeptical about the term 'discipline'. That sounds academic to me. I might be mistaken. What about 'subject area'? I also struggle to see a red thread in the list of 'disciplines'. Proportionality is a principle, Risk Analysis is an activity, etc... If at all, these should be called 'sub-disciplines' rather than disciplines if they are part of algo ethics; but I'd prefer to talk about 'key issues' of algorithm ethics or something along these lines, anyway.

In the Algorithm Ethics, I would add KRI design (Key Risk Indicator). KPI tends to be more about performance of the business whereas KRI is more about what might prevent the Entity from achieving its goals (likelihood of future adverse impact).