System and method for coordinated agent learning and explanation using hierarchical factors
A system and method for selection and explanation of solutions is provided. A hierarchy of aggregation factors is maintained for evaluating at least a partial solution. Competing solutions are generated and each solution includes at least a partial solution. Scores are calculated for each of the aggregation factors for each competing solution. A total evaluation score is calculated for each competing solution based on the scores for at least one of the aggregation factors. The competing solution with the best evaluation score is selected and a gist is generated. The gist is a narrative comparing the selected solution with the non-selected solutions based on the at least one aggregation factor. The gist is provided to a user as a rationale for selection of the solution.
1 . A system for selection and explanation of solutions, comprising:
memory to maintain a hierarchy of aggregation factors for evaluating at least a partial solution; and
a server comprising a central processing unit, memory, an input port to receive the hierarchy of aggregation factors from the memory, and an output port, wherein the central processing-unit is configured to:
receive a request for a task or activity;
generate competing solutions for the task for activity via an AI agent, each solution comprising at least a partial solution;
calculate scores for each of the aggregation factors for each competing solution;
calculate an evaluation score based on the scores for at least one of the aggregation factors for each competing solution;
select the competing solution with the best evaluation score;
generate via the AI agent a gist comprising a narrative comparing the selected solution with the non-selected solutions based on the at least one aggregation factor for the task or activity and based on where the at least one aggregation factor falls in the hierarchy of aggregation factors; and
provide the gist to a user as a rationale for selection of the solution to the task or activity.
2 . A system according to claim 1 , wherein the gist identifies one of the aggregation factors as a main factor for selecting the solution with the best evaluation score and one of the aggregation factors as a main penalty that is bypassed during selection of the plan with a worst aggregate score.
3 . A system according to claim 2 , wherein the central processing unit performs the following:
identify the main factor as the aggregate factor that accounts for 40% or more of a penalty difference between the scores for each of the competing solutions; and
identify the main penalty as the aggregation factor that accepts a small penalty disadvantage in exchange for the penalty identified by the main factor, which is larger.
4 . A system according to claim 1 , wherein the central processing unit performs the following:
calculate a further evaluation score for a different set of the aggregation factors for each competing solution;
select the competing solution with a best further evaluation score;
generate a gist comprising a narrative comparing the selected solution with the non-selected solutions based on the at least one aggregation factor and the different set of aggregation factors.
5 . A system according to claim 1 , wherein one or more of the aggregation factors includes sub-factors, which are also assigned scores that are included in the evaluation score.
6 . A system according to claim 1 , wherein the central processing unit performs the following:
create a visualization of the competing solutions; and
display the visualization for each of the solutions on a world model.
7 . A system according to claim 6 , wherein the central processing unit performs the following:
provide visualizations for one or more of the aggregation factors on the world model.
8 . A system according to claim 7 , wherein the visualizations of the competing solutions and the aggregation factors are provided as an overlay on the world model.
9 . A system according to claim 6 , wherein the central processing unit performs the following:
generate a table for the aggregation factors; and
display the table with the world model.
10 . A system according to claim 9 , wherein the central processing unit performs the following
select at least one of the aggregation factors in the table; and
display the selected aggregation factor in the world model.
11 . A method for selection and explanation of solutions, comprising:
maintaining a hierarchy of aggregation factors for evaluating at least a partial solution;
receiving a request for a task or activity;
generating competing solutions for the task or activity via an AI agent, each solution comprising at least a partial solution;
calculating scores for each of the aggregation factors for each competing solution;
calculating for each competing solution an evaluation score based on the scores for at least one of the aggregation factors;
selecting the competing solution with the best evaluation score;
generating via the AI agent a gist comprising a narrative comparing the selected solution with the non-selected solutions based on the at least one aggregation factor for the task or activity and based on where the at least one aggregation factor falls in the hierarchy of aggregation factors; and
providing the gist to a user as a rationale for selection of the solution.
12 . A method according to claim 11 , wherein the gist comprises one of the aggregation factors as a main factor for selecting the solution with the best evaluation score and one of the aggregation factors as a main penalty that is bypassed during selection of the plan with a worst aggregate score.
13 . A method according to claim 12 , further comprising at least one of:
identifying the main factor as the aggregate factor that accounts for 40% or more of a penalty difference between the scores for each of the competing solutions; and
identifying the main penalty as the aggregation factor that accepts a small penalty disadvantage in exchange for the penalty identified by the main factor, which is larger.
14 . A method according to claim 11 , further comprising:
calculating a further evaluation score for a different set of the aggregation factors for each competing solution;
selecting the competing solution with a best further evaluation score;
generating a gist comprising a narrative comparing the selected solution with the non-selected solutions based on the at least one aggregation factor and the different set of aggregation factors.
15 . A method according to claim 11 , wherein one or more of the aggregation factors includes sub-factors, which are also assigned scores that are included in the evaluation score.
16 . A method according to claim 11 , further comprising:
creating a visualization of the competing solutions; and
displaying the visualization for each of the solutions on a world model.
17 . A method according to claim 16 , further comprising:
providing visualizations for one or more of the aggregation factors on the world model.
18 . A method according to claim 17 , wherein the visualizations of the competing solutions and the aggregation factors are provided as an overlay on the world model.
19 . A method according to claim 16 , further comprising:
generating a table for the aggregation factors; and
displaying the table with the world model.
20 . A method according to claim 19 , further comprising:
selecting at least one of the aggregation factors in the table; and
displaying the selected aggregation factor in the world model.