Captured source text (L5)
effective
| | Criteria | | | Practical steps to fulfil criteria | |
| --- | --- | --- | --- | --- | --- |
| Robustness: the degree to which an age assurance method can correctly determine the age of a user in actual deployment contexts. | Robustness: the degree to which an | | Implement age assurance processes that have undergone tests in multiple environments during development. Identify and take appropriate steps to mitigate against methods of circumvention that are easily accessible to children and where it is reasonable to assume that children may use them. | | |
| | age assurance method can correctly | | | | |
| | determine the age of a user in actual | | | | |
| | deployment contexts. | | | | |
| Reliability: the degree to which the age output from an age assurance method is reproducible and derived from trustworthy evidence. | | | | Where age assurance methods forming part of the age | |
| | | | | assurance process rely on artificial intelligence or | |
| | | | | machine learning, take steps to ensure that: | |
| | | | | • the artificial intelligence or machine learning | |
| | | | | method(s) has been suitably tested during the | |
| | | | | development of the age assurance process to | |
| | | | | ensure it produces reproducible results; | |
| | | | | • once deployed, the artificial intelligence or | |
| | | | | machine learning method(s) is regularly | |
| | | | | monitored to ensure it produces reproducible | |
| | | | | results; | |
| | | | | • the outputs of the artificial intelligence or | |
| | | | | machine learning method(s) are assessed against | |
| | | | | key performance indicators designed to identify | |
| | | | | whether the artificial intelligence or machine | |
| | | | | learning produces reproducible results; | |
| | | | | • in circumstances where the artificial intelligence | |
| | | | | or machine learning used is observed to be | |
| | | | | producing unreliable or unexpected results, the | |
| | | | | root cause of the issue is identified and rectified. | |
| | | | | Take steps to ensure that any data relied upon as part of | |
| | | | | the age assurance process comes from a trustworthy | |
| | | | | source. | |
| Fairness: the extent to which an age assurance method avoids or minimises bias and discriminatory outcomes. | | | Ensure that any elements of the age assurance process which rely on artificial intelligence or machine learning have been tested and trained on data sets which reflect the diversity in the target population. For methods reliant on artificial intelligence or machine learning, ensure the age assurance method(s) has been evaluated against the outcome / error parity and the results indicate that the method(s) does not produce significant bias or discriminatory outcomes. | | |
Robustness: the degree to which an Implement age assurance processes that have
age assurance method can correctly undergone tests in multiple environments during
determine the age of a user in actual development.
Reliability: the degree to which the Where age assurance methods forming part of the age
age output from an age assurance assurance process rely on artificial intelligence or
method is reproducible and derived machine learning, take steps to ensure that:
Fairness: the extent to which an age Ensure that any elements of the age assurance process
assurance method avoids or which rely on artificial intelligence or machine learning
minimises bias and discriminatory have been tested and trained on data sets which reflect
outcomes. the diversity in the target population.
4.4We recognise that different kinds of age assurance – or even the same kinds of age
assurance provided by different companies – may perform more strongly in some of these
11
| Reliability: the degree to which the |
| --- |
| age output from an age assurance |
| method is reproducible and derived |
| from trustworthy evidence. |
| Fairness: the extent to which an age |
| --- |
| assurance method avoids or |
| minimises bias and discriminatory |
| outcomes. |
Full captured verbatim text
| | Criteria | | | Practical steps to fulfil criteria | | | --- | --- | --- | --- | --- | --- | | Robustness: the degree to which an age assurance method can correctly determine the age of a user in actual deployment contexts. | Robustness: the degree to which an | | Implement age assurance processes that have undergone tests in multiple environments during development. Identify and take appropriate steps to mitigate against methods of circumvention that are easily accessible to children and where it is reasonable to assume that children may use them. | | | | | age assurance method can correctly | | | | | | | determine the age of a user in actual | | | | | | | deployment contexts. | | | | | | Reliability: the degree to which the age output from an age assurance method is reproducible and derived from trustworthy evidence. | | | | Where age assurance methods forming part of the age | | | | | | | assurance process rely on artificial intelligence or | | | | | | | machine learning, take steps to ensure that: | | | | | | | • the artificial intelligence or machine learning | | | | | | | method(s) has been suitably tested during the | | | | | | | development of the age assurance process to | | | | | | | ensure it produces reproducible results; | | | | | | | • once deployed, the artificial intelligence or | | | | | | | machine learning method(s) is regularly | | | | | | | monitored to ensure it produces reproducible | | | | | | | results; | | | | | | | • the outputs of the artificial intelligence or | | | | | | | machine learning method(s) are assessed against | | | | | | | key performance indicators designed to identify | | | | | | | whether the artificial intelligence or machine | | | | | | | learning produces reproducible results; | | | | | | | • in circumstances where the artificial intelligence | | | | | | | or machine learning used is observed to be | | | | | | | producing unreliable or unexpected results, the | | | | | | | root cause of the issue is identified and rectified. | | | | | | | Take steps to ensure that any data relied upon as part of | | | | | | | the age assurance process comes from a trustworthy | | | | | | | source. | | | Fairness: the extent to which an age assurance method avoids or minimises bias and discriminatory outcomes. | | | Ensure that any elements of the age assurance process which rely on artificial intelligence or machine learning have been tested and trained on data sets which reflect the diversity in the target population. For methods reliant on artificial intelligence or machine learning, ensure the age assurance method(s) has been evaluated against the outcome / error parity and the results indicate that the method(s) does not produce significant bias or discriminatory outcomes. | | | Robustness: the degree to which an Implement age assurance processes that have age assurance method can correctly undergone tests in multiple environments during determine the age of a user in actual development. Reliability: the degree to which the Where age assurance methods forming part of the age age output from an age assurance assurance process rely on artificial intelligence or method is reproducible and derived machine learning, take steps to ensure that: Fairness: the extent to which an age Ensure that any elements of the age assurance process assurance method avoids or which rely on artificial intelligence or machine learning minimises bias and discriminatory have been tested and trained on data sets which reflect outcomes. the diversity in the target population. 4.4 We recognise that different kinds of age assurance – or even the same kinds of age assurance provided by different companies – may perform more strongly in some of these 11 | Reliability: the degree to which the | | --- | | age output from an age assurance | | method is reproducible and derived | | from trustworthy evidence. | | Fairness: the extent to which an age | | --- | | assurance method avoids or | | minimises bias and discriminatory | | outcomes. |