- From: Harshvardhan Pandit <me@harshp.com>
- Date: Mon, 21 Apr 2025 13:29:05 +0100
- To: public-dpvcg@w3.org
Hi All. I am proposing that we model "operating factors" such as data, instruments, environmental conditions which affect or are relevant to the operation of the technology. These are recommended/necessary in AI Act and in documentation efforts such as Model Cards where such factors are described in context of necessary conditions for operation, known problematic/unsuitable conditions, etc. The full proposal with thought process and examples is available at: https://harshp.com/dev/dpv/tech-operating-factors The summary is to add tech:OperatingFactor with subclasses: - tech:NecessaryOperatingFactor - tech:SuitableOperatingFactor with subclass tech:IdealOperatingFactor - tech:UnsuitableOperatingFactor with subclasses tech:HazardousOperatingFactor and tech:UntestedOperatingFactor To describe what is in the factor, both model cards and the AI Act refer to groups (people and data categories), technologies, environmental conditions, locations, etc. -- for which the existing DPV concepts and properties are available for reuse, except for environment. Therefore, I propose we add `tech:OperatingEnvironment` to describe the broader environment of operation (as needed in AI Act), and `tech:EnvironmentalCondition` to represent the conditions such as temperature etc. (there should be good industrial ontologies which expand upon this). Together with these, the blog post outlines 22 properties which can be used to define the factor. To distinguish from the earlier proposal on use-cases as both are similar and are specified as categories of `dpv:Process`: Here, the 'factor' is a description of conditions that affect the technology, whereas 'use-case' is a scenario which represent how something might be used. They are _very similar_ concepts when considering how they are used, but they have _distinct_ uses in documentation. For example, manuals contain descriptions of factors as suitable/unsuitable to describe the technical nature of the technology, whereas use-cases are used as hypotheticals and scenarios to show what could happen or will happen e.g. in impact assessments. So both share an conceptual and modelling overlap, but they have their uses in different contexts. Regards, Harsh
Received on Monday, 21 April 2025 12:29:12 UTC