Automatic transparency evaluation for open knowledge extraction systems

Table 1 Differences and commonalities between AI transparency methods

AI transparency methods	Transparency information
	Developers	Model date and version	Model type	Intended use	Training data	Evaluation datasets	Evaluation datasets preprocessing	Model performance measures or decision thresholds	Description of model tests	Most effective attributes	Quantitative analyses	Ethical considerations	Limitations
Model Cards [42]	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	×	×	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)
Nutritional label for ranking AI systems [43]	×	×	\(\checkmark\)	×	×	×	×	×	×	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	×
FactSheets [44]	×	×	\(\checkmark\)	×	\(\checkmark\)	×	×	\(\checkmark\)	\(\checkmark\)	×	\(\checkmark\)	\(\checkmark\)	×

ISSN: 2041-1480