Clustering, Explainability, and Automated Decisions in Computer-Based Reasoning Systems

• US 20200134484A1
• Filed: 10/22/2019
• Published: 04/30/2020
• Est. Priority Date: 10/30/2018
• Status: Active Grant

• Current Assignee: Howso Incorporated
• Original Assignee: Diveplane Corporation
• Inventors: Hazard, Christopher James, Fusting, Christopher, Resnick, Michael
• Granted Patent: US 11,494,669 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 5/045 Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence

Abstract

The techniques herein include using an input context to determine a suggested action and/or cluster. Explanations may also be determined and returned along with the suggested action. The explanations may include (i) one or more most similar cases to the suggested case (e.g., the case associated with the suggested action) and, optionally, a conviction score for each nearby cases; (ii) action probabilities, (iii) excluding cases and distances, (iv) archetype and/or counterfactual cases for the suggested action; (v) feature residuals; (vi) regional model complexity; (vii) fractional dimensionality; (viii) prediction conviction; (ix) feature prediction contribution; and/or other measures such as the ones discussed herein, including certainty. The explanation data may be used to determine whether to perform a suggested action.

Claims

1. A method comprising:

receiving a request for a suggested action based on an input context, in a computer-based reasoning system, wherein the computer-based reasoning system includes a computer-based reasoning model;

determining one or more candidate cases based on the input context in the computer-based reasoning system, wherein the one or more candidate cases include respective one or more candidate actions;

determining the suggested action based on the respective one or more candidate actions;

determining a suggested cluster based on the one or more candidate cases,

determining a compatibility score for the suggested action and the suggested cluster based at least in part on the suggested cluster and the suggested action;

when there the compatibility score is not beyond a threshold, responding to the request for the suggested action with the suggested action and the suggested cluster;

when there the compatibility score is beyond the threshold;

causing the performance of the suggested action by a control system,

wherein the method is performed by one or more computing devices.

2. The method of claim 1, wherein determining the compatibility score for the suggested action and the suggested cluster comprises determining whether there is a known compatibility between the suggested action and the suggested cluster based on a compatibility lookup of the suggested action and the suggested cluster.

3. The method of claim 1, wherein determining the compatibility score for the suggested action and the suggested cluster comprises determining whether there is a known incompatibility between the suggested action and the suggested cluster based on an incompatibility lookup of the suggested action and the suggested cluster.

4. The method of claim 1, wherein the compatibility score for the suggested action and the suggested cluster represents a likelihood of fit between an action and a cluster.

5. The method of claim 1, wherein the compatibility score for the suggested action and the suggested cluster represents a probability of fit between an action and a cluster.

6. The method of claim 1, further comprising:

determining a certainty score for the suggested cluster, and

when there the compatibility score is not beyond a threshold or the certainty score is not beyond a second threshold, responding to the request for the suggested action with the suggested action, the suggested cluster, and the certainty score for the suggested cluster;

when there the compatibility score corresponds is beyond the threshold and the certainty score is beyond the second threshold;

causing the performance of the suggested action by the control system.

7. The method of claim 6, wherein determining the certainty score for the suggested cluster comprises determining the certainty score is based on a conviction function associated with:

removing a suggested case associated with the suggested cluster from the computer-based reasoning model;

adding the suggested case into the computer-based reasoning model.

8. The method of claim 7, wherein the conviction function is a measure of how much information the suggested case distorts the computer-based reasoning model.

9. The method of claim 7, wherein the conviction function is a measure of how much information the suggested case distorts the computer-based reasoning model

10. A system for performing a machine-executed operation involving instructions, wherein said instructions are instructions which, when executed by one or more computing devices, cause performance of a process comprising:

receiving a request for a suggested action based on an input context, in a computer-based reasoning system, wherein the computer-based reasoning system includes a computer-based reasoning model;

determining one or more candidate cases based on the input context in the computer-based reasoning system, wherein the one or more candidate cases include respective one or more candidate actions;

determining the suggested action based on the respective one or more candidate actions;

determining a certainty score based on the suggested action, wherein the certainty score is determined based on a conviction function associated with;

removing a suggested case associated with the suggested action from the computer-based reasoning model;

adding the suggested case into the computer-based reasoning model,

wherein the conviction function is a measure of how much information the suggested case distorts the computer-based reasoning model;

when the certainty score is beyond a certain threshold, responding to the request for the suggested action with at least the suggested action;

when the certainty score is not beyond the certain threshold;

determining one or more explanation factors for the suggested action determined based at least in part on the input context;

responding to the request for the suggested action with the suggested action and the one or more explanation factors.

11. The system of claim 10, wherein determining the suggested action comprises determining a cluster associated with each of the one or more candidate cases, determining a suggested cluster based on the cluster associated with each of the one or more candidate cases, and determining the suggested action based at least in part based on the suggested cluster.

12. The system of claim 10, wherein determining the suggested action comprises determining a cluster associated with each of the one or more candidate cases, determining a suggested cluster based on the cluster associated with each of the one or more candidate cases, and the suggested cluster is returned as the suggested action.

13. The system of claim 10, the process further comprising:

when the certainty score is not beyond the certain threshold, after providing the suggested action and the one or more explanation factors in response to the request for the suggested action, receiving an indication whether to perform the suggested action;

when the received indication indicates performance of the suggested action, causing control of a controllable system based on the suggested action.

14. The system of claim 10, wherein determining the suggested action comprises:

determining a weighting for each action of the respective one or more candidate actions based on a function of a distance between the input context and each of the one or more candidate cases;

determining the suggested action based on the weighting for each action of the respective one or more candidate actions.

15. The system of claim 10, wherein responding to the request for the suggested action with at least the suggested action comprises when the certainty score is beyond the certain threshold, comprises responding to the request for the suggested action with at least the suggested action and the certainty score when the certainty score is beyond the certain threshold.

16. A system for performing a machine-executed operation involving instructions, wherein said instructions are instructions which, when executed by one or more computing devices, cause performance of a process comprising:

receiving a request for a suggested action based on an input context, in a computer-based reasoning system, wherein the computer-based reasoning system includes a computer-based reasoning model;

determining one or more candidate cases based on the input context in the computer-based reasoning system, wherein the one or more candidate cases include respective one or more candidate actions;

determining the suggested action based on the respective one or more candidate actions;

determining a certainty score based on the suggested action, wherein the certainty score is determined based on a conviction function associated with;

removing a suggested case associated with the suggested action from the computer-based reasoning model; and

adding the suggested case into the computer-based reasoning model,

wherein the conviction function is a measure of how much information is required to describe a position of the suggested case relative to existing cases in the computer-based reasoning model;

when the certainty score is beyond a certain threshold, responding to the request for the suggested action with at least the suggested action;

when the certainty score is not beyond the certain threshold;

determining one or more explanation factors for the suggested action determined based at least in part on the input context;

responding to the request for the suggested action with the suggested action and the one or more explanation factors.

17. The system of claim 16, wherein determining the suggested action comprises determining a cluster associated with each of the one or more candidate cases, determining a suggested cluster based on the cluster associated with each of the one or more candidate cases, and determining the suggested action based at least in part based on the suggested cluster.

18. The system of claim 16, wherein determining the suggested action comprises determining a cluster associated with each of the one or more candidate cases, determining a suggested cluster based on the cluster associated with each of the one or more candidate cases, and the suggested cluster is returned as the suggested action.

19. The system of claim 16, further comprising:

when the certainty score is not beyond the certain threshold, after providing the suggested action and the one or more explanation factors in response to the request for the suggested action, receiving an indication whether to perform the suggested action;

when the received indication indicates performance of the suggested action, causing control of a controllable system based on the suggested action.

20. The system of claim 16, wherein determining the suggested action comprises:

determining a weighting for each action of the respective one or more candidate actions based on a function of a distance between the input context and each of the one or more candidate cases;

determining the suggested action based on the weighting for each action of the respective one or more candidate actions.

21. The system of claim 16, wherein responding to the request for the suggested action with at least the suggested action comprises when the certainty score is beyond a certain threshold, comprises responding to the request for the suggested action with at least the suggested action and the certainty score when the certainty score is beyond the certain threshold.