Subcharacteristic definitions

This page contains definitions of all the subcharacteristics. Click here for more details about each subcharacteristic

Characteristic	Subcharacteristics	Definition
Utility	Accuracy	The size of the observational error (systematic and random) of an ML system.
	Effectiveness	The ability of an ML system to produce a desired result on the business task it is being used for.
	Responsiveness	The ability of an ML system to complete the desired task in an acceptable time frame.
	Reusability	The ability to reuse the same ML system without any change, for another business case.
Economy	Cost-effectiveness	The extent to which an ML System achieves the desired relationship between costs and overall impact.
	Efficiency	The ability to avoid wasting resources (computational, human, financial, etc.) in order to perform the desired task effectively.
	Availability	The probability that the system will be up and running and able to deliver useful services to users.
	Resilience	The extent to which an ML system can provide and maintain an acceptable level of service in the face of technical challenges to normal operation.
	Adaptability	The extent to which an ML system can adapt to changes in the production environment, always providing the same functioning level.
	Scalability	The extent of an ML system to handle a growing amount of work by adding resources to the system.
Modifiability	Extensibility	The ease with which a system can be modified, in order to be used for another purpose.
	Maintainability	The ease with which activities aiming at keeping an ML system functional in the desired regime, can be performed.
	Modularity	The extent to which an ML system consists of components of limited functionality that interrelate with each other.
	Testability	The extent with which an ML system can be tested against expected behaviours.
Productionizability	Deployability	The extent to which an ML system can be deployed in production when needed.
	Repeatability	The ease with which the ML lifecycle can be repeated.
	Operability	The extent to which an ML system can be controlled (disable, revert, upload new version, etc.).
	Monitoring	The extent to which relevant indicators of an ML system are effectively observed/monitored and integrated in the operation of the system.
Comprehensibility	Discoverability	The extent to which an ML system can be found (by means of performing a search in a database or any other information retrieval system).
	Readability	The ease with which the code of an ML system can be understood.
	Traceability	The ability to relate artifacts created during the development of an ML system to describe the system from different perspectives and levels of abstraction with each other, the stakeholders that have contributed to the creation of the artifacts, and the rationale that explains the form of the artifacts.
	Understandability	The ease with which the implementation and design choices of an ML system can be understood.
	Usability	The extent to which an ML system can be effectively used by users.
	Debuggability	The extent to which the inner workings of the ML system can be analyzed in order to understand why it behaves the way it does.
Responsibility	Explainability	The ability to explain the output of an ML system.
	Fairness	The extent to which an ML system prevents unjust predictions towards protected attributes (race, gender, income, etc).
	Ownership	The extent to which there exists people appointed to maintaining the ML System and supporting all the relevant stakeholders.
	Standards Compliance	The extent to which applicable standards are followed in the ML system.
	Vulnerability	The ease with which the system can be (ab)used to achieve malicious purposes.