Misuse index for explainable artificial intelligence in computing environments

1. A non-transitory computer readable storage medium comprising instructions which, when executed by a data processing machine, cause the data processing machine to perform operations comprising:

mapping training data with inference uses in a machine learning environment, wherein the training data is used for training a machine learning model and the inference uses correspond to application of the machine learning model;

detecting, based on one or more policy/parameter thresholds, one or more discrepancies between the training data and the inference uses;

identifying, based on the one or more discrepancies, that one or more misuses of the machine learning model occurred;

determining, based on the one or more discrepancies, a misuse index value corresponding to a misuse index; and

estimating, based on the misuse index value, a severity of the one or more misuses of the machine learning model and identify one or more responses to the one or more misuses.

2. The non-transitory computer readable storage medium of claim 1 , wherein the inference uses are based on field data captured using one or more capturing devices including one or more of a camera, a microphone, and one or more sensors, wherein the training data and the field data are labeled with metadata to identify detailed contents relating to one or more of the training data and the field data, wherein the machine learning environment is part of an explainable artificial intelligence (XAI) environment.

3. The non-transitory computer readable storage medium of claim 1 , wherein the operations further comprise:

determining the one or more misuses to be one or more errors or one or more policy/parameter-altering acts;

modifying the misuse index to distinguish between the one or more errors and the one or more policy/parameter-altering acts; and

proposing actionable tasks to respond to the one or more policy/parameter-altering acts, wherein the actionable tasks include one or more of modifying the one or more policy/parameter thresholds, pausing the one or more policy/parameter-altering acts, and issuing one or more warnings to prevent future occurrences of the one or more policy/parameter-altering acts.

4. The non-transitory computer readable storage medium of claim 2 , wherein the one or more policy/parameter thresholds are based on one or more features including one or more of governmental laws, administrative rules and regulations, organizational policies, cultural norms, societal customs, and ethical expectations, wherein the one or more policy/parameter thresholds are dynamically modified based on changes in the one or more features, wherein the one or more features are extracted from one or more of the training data, the field data, the metadata, and user inputs.

5. The non-transitory computer readable storage medium of claim 4 , wherein the operations further comprise:

transmitting one or more surveys to one or more computing devices over one or more communication mediums, wherein the one or more surveys are used to collect the user inputs from one or more users having access to the one or more computing devices, wherein the user inputs includes one or more of policy contexts, policy reasons, and policy recommendations;

evaluating the one or more policy/parameter thresholds based on the user inputs; and

modifying the one or more policy/parameter thresholds based on the user inputs.

6. The non-transitory computer readable storage medium of claim 5 , wherein the operations further comprise adjusting the detection of the one or more discrepancies based on the one or more modified policy/parameter thresholds.

7. The non-transitory computer readable storage medium of claim 1 , wherein the data processing machine comprises one or more processors including one or more of a graphics processor and an application processor, wherein the one or more processors are co-located on a common semiconductor package.

8. A method comprising:

mapping training data with inference uses in a machine learning environment, wherein the training data is used for training a machine learning model and the inference uses correspond to application of the machine learning model;

detecting, based on one or more policy/parameter thresholds, one or more discrepancies between the training data and the inference uses;

identifying, based on the one or more discrepancies, that one or more misuses of the machine learning model occurred;

determining, based on the one or more discrepancies, a misuse index value corresponding to a misuse index; and

estimating, based on the misuse index value, a severity of the one or more misuses of the machine learning model and identify one or more responses to the one or more misuses.

9. The method of claim 8 , wherein the inference uses are based on field data captured using one or more capturing devices including one or more of a camera, a microphone, and one or more sensors, wherein the training data and the field data are labeled with metadata to identify detailed contents relating to one or more of the training data and the field data, wherein the machine learning environment is part of an explainable artificial intelligence (XAI) environment.

10. The method of claim 8 , further comprising:

determining the one or more misuses to be one or more errors or one or more policy/parameter-altering acts;

modifying the misuse index to distinguish between the one or more errors and the one or more policy/parameter-altering acts; and

proposing actionable tasks to respond to the one or more policy/parameter-altering acts, wherein the actionable tasks include one or more of modifying the one or more policy/parameter thresholds, pausing the one or more policy/parameter-altering acts, and issuing one or more warnings to prevent future occurrences of the one or more policy/parameter-altering acts.

11. The method of claim 9 , wherein the one or more policy/parameter thresholds are based on one or more features including one or more of governmental laws, administrative rules and regulations, organizational policies, cultural norms, societal customs, and ethical expectations, wherein the one or more policy/parameter thresholds are dynamically modified based on changes in the one or more features, wherein the one or more features are extracted from one or more of the training data, the field data, the metadata, and user inputs.

12. The method of claim 11 , further comprising:

transmitting one or more surveys to one or more computing devices over one or more communication mediums, wherein the one or more surveys are used to collect the user inputs from one or more users having access to the one or more computing devices, wherein the user inputs includes one or more of policy contexts, policy reasons, and policy recommendations;

evaluating the one or more policy/parameter thresholds based on the user inputs;

modifying the one or more policy/parameter thresholds based on the user inputs; and

adjusting the detection of the one or more discrepancies based on the one or more modified policy/parameter thresholds.

13. The method of claim 8 , wherein the method is facilitated by one or more processors including one or more of a graphics processor and an application processor, wherein the one or more processors are co-located on a common semiconductor package.

14. An apparatus comprising:

one or more processors to:

map training data with inference uses in a machine learning environment, wherein the training data is used for training a machine learning model and the inference uses correspond to application of the machine learning model;

detect, based on one or more policy/parameter thresholds, one or more discrepancies between the training data and the inference uses;

identify, based on the one or more discrepancies, that one or more misuses of the machine learning model occurred;

determine, based on the one or more discrepancies, a misuse index value corresponding to a misuse index; and

estimate, based on the misuse index value, a severity of the one or more misuses of the machine learning model and identify one or more responses to the one or more misuses.

15. The apparatus of claim 14 , wherein the inference uses are based on field data captured using one or more capturing devices including one or more of a camera, a microphone, and one or more sensors, wherein the training data and the field data are labeled with metadata to identify detailed contents relating to one or more of the training data and the field data, wherein the machine learning environment is part of an explainable artificial intelligence (XAI) environment.

16. The apparatus of claim 14 , wherein the one or more processors are further to:

determine the one or more misuses to be one or more errors or one or more policy/parameter-altering acts;

modify the misuse index to distinguish between the one or more errors and the one or more policy/parameter-altering acts; and

propose actionable tasks to respond to the one or more policy/parameter-altering acts, wherein the actionable tasks include one or more of modifying the one or more policy/parameter thresholds, pausing the one or more policy/parameter-altering acts, and issuing one or more warnings to prevent future occurrences of the one or more policy/parameter-altering acts.

17. The apparatus of claim 15 , wherein the one or more policy/parameter thresholds are based on one or more features including one or more of governmental laws, administrative rules and regulations, organizational policies, cultural norms, societal customs, and ethical expectations, wherein the one or more policy/parameter thresholds are dynamically modified based on changes in the one or more features, wherein the one or more features are extracted from one or more of the training data, the field data, the metadata, and user inputs.

18. The apparatus of claim 17 , wherein the one or more processors are further to:

transmit one or more surveys to one or more computing devices over one or more communication mediums, wherein the one or more surveys are used to collect the user inputs from one or more users having access to the one or more computing devices, wherein the user inputs includes one or more of policy contexts, policy reasons, and policy recommendations;

evaluate the one or more policy/parameter thresholds based on the user inputs; and

modify the one or more policy/parameter thresholds based on the user inputs.

19. The apparatus of claim 18 , wherein the one or more processors are further to adjust the detection of the one or more discrepancies based on the one or more modified policy/parameter thresholds.

20. The apparatus of claim 14 , wherein the one or more processors include one or more of a graphics processor and an application processor, wherein the one or more processors are co-located on a common semiconductor package.