IP Library › Granted Patent US 12,619,616
Granted Patent B2
US 12,619,616 · App. 18/931,908 · Granted May 5, 2026

Ethical virtual assistant training platform

Inventors: Jeremy R. Fox (Georgetown, TX); Fernando Luiz Koch (Palm Beach Gardens, FL); Martin G. Keen (Cary, NC); Jessica Nahulan (Markham, CA)
Assignee: INTERNATIONAL BUSINESS MACHINES CORPORATION
G06F16/24575G06F16/2365G06N20/00
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,619,616
App. No.
18/931,908
Granted
May 5, 2026
Kind
B2
Abstract

An embodiment establishes a population database. The embodiment constructs a behavior framework based on population data stored on the population database. The embodiment initiates a training process to train a virtual assistant to generate an output response in response to a query, including training the virtual assistant to generate responses according to the behavior framework. The embodiment monitors an interaction session between a client interacting with the virtual assistant to detect a deviation from the behavior framework. The embodiment generates, upon detection of the deviation from the behavior framework, a correction action to apply to a portion of the interaction session corresponding to the deviation from the behavior framework. The embodiment applies the correction action to the portion of the interaction session corresponding to the deviation from the behavior framework. The embodiment outputs a corrected output response that meets a set of guidelines of the behavior framework.

Claims (58)

1 . A computer-implemented method comprising:

establishing a population database, the population database configured to transmit and store population data;

constructing a behavior framework based on population data stored on the population database;

constructing a second behavior framework based on population data stored on the population database;

initiating a training process to train a virtual assistant to generate an output response in response to a query, wherein the training process comprises training the virtual assistant to generate responses according to the behavior framework;

initiating a second training process to train a second virtual assistant to generate an output response in response to a query, wherein the second training process comprises training the second virtual assistant to generate responses according to the second behavior framework;

monitoring an interaction session between a client interacting with the virtual assistant to detect a deviation from the behavior framework;

generating, upon detection of the deviation from the behavior framework, a correction action to apply to a portion of the interaction session corresponding to the deviation from the behavior framework;

applying the correction action to the portion of the interaction session corresponding to the deviation from the behavior framework;

outputting a corrected output response, wherein the corrected output response meets a set of guidelines of the behavior framework;

monitoring a second interaction session between a second client interacting with the second virtual assistant to detect a deviation from the second behavior framework;

generating, upon detection of the deviation from the second behavior framework, a second correction action to apply to a portion of the second interaction session corresponding to the deviation from the second behavior framework;

applying the second correction action to the portion of the second interaction session corresponding to the deviation from the second behavior framework; and

outputting a second corrected output response, wherein the second corrected output response meets a set of guidelines of the second behavior framework.

2 . The computer-implemented method of claim 1 , wherein the second behavior framework comprises a different set of guidelines than the behavior framework.

3 . The computer-implemented method of claim 1 , wherein the correction action comprises modifying the portion of the interaction session corresponding to the deviation from the behavior framework.

4 . The computer-implemented method of claim 1 , wherein the correction action comprises providing a suggested action to mitigate a potential bias outcome.

5 . The computer-implemented method of claim 1 , wherein the correction action comprises providing additional context surrounding the portion of the interaction session corresponding to the deviation from the behavior framework.

6 . A computer program product comprising one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions executable by a processor to cause the processor to perform operations comprising:

establishing a population database, the population database configured to transmit and store population data;

constructing a behavior framework based on population data stored on the population database;

constructing a second behavior framework based on population data stored on the population database;

initiating a training process to train a virtual assistant to generate an output response in response to a query, wherein the training process comprises training the virtual assistant to generate responses according to the behavior framework;

initiating a second training process to train a second virtual assistant to generate an output response in response to a query, wherein the second training process comprises training the second virtual assistant to generate responses according to the second behavior framework;

monitoring an interaction session between a client interacting with the virtual assistant to detect a deviation from the behavior framework;

generating, upon detection of the deviation from the behavior framework, a correction action to apply to a portion of the interaction session corresponding to the deviation from the behavior framework;

applying the correction action to the portion of the interaction session corresponding to the deviation from the behavior framework:

outputting a corrected output response, wherein the corrected output response meets a set of guidelines of the behavior framework;

monitoring a second interaction session between a second client interacting with the second virtual assistant to detect a deviation from the second behavior framework;

generating, upon detection of the deviation from the second behavior framework, a second correction action to apply to a portion of the second interaction session corresponding to the deviation from the second behavior framework;

applying the second correction action to the portion of the second interaction session corresponding to the deviation from the second behavior framework; and

outputting a second corrected output response, wherein the second corrected output response meets a set of guidelines of the second behavior framework.

7 . The computer program product of claim 6 wherein the program instructions are stored in a computer readable storage device in a data processing system, and wherein the stored program instructions are transferred over a network from a remote data processing system.

8 . The computer program product of claim 6 , wherein the program instructions are stored in a computer readable storage device in a server data processing system, and wherein the program instructions are downloaded in response to a request over a network to a remote data processing system for use in the computer readable storage device associated with the remote data processing system, further comprising:

program instructions to meter use of the program instructions associated with the request; and

program instructions to generate an invoice based on the metered use.

9 . The computer program product of claim 6 , wherein the set of guidelines of the second behavior framework is different from the set of guidelines of the behavior framework.

10 . The computer program product of claim 6 , wherein the correction action comprises modifying the portion of the interaction session corresponding to the deviation from the behavior framework.

11 . The computer program product of claim 6 , wherein the correction action comprises providing a suggested action to mitigate a potential bias outcome.

12 . The computer program product of claim 6 , wherein the correction action comprises providing additional context surrounding the portion of the interaction session corresponding to the deviation from the behavior framework.

13 . A computer system comprising a processor and one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions executable by the processor to cause the processor to perform operations comprising:

establishing a population database, the population database configured to transmit and store population data;

constructing a behavior framework based on population data stored on the population database;

constructing a second behavior framework based on population data stored on the population database;

initiating a training process to train a virtual assistant to generate an output response in response to a query, wherein the training process comprises training the virtual assistant to generate responses according to the behavior framework;

initiating a second training process to train a second virtual assistant to generate an output response in response to a query, wherein the second training process comprises training the second virtual assistant to generate responses according to the second behavior framework;

monitoring an interaction session between a client interacting with the virtual assistant to detect a deviation from the behavior framework;

generating, upon detection of the deviation from the behavior framework, a correction action to apply to a portion of the interaction session corresponding to the deviation from the behavior framework;

applying the correction action to the portion of the interaction session corresponding to the deviation from the behavior framework;

outputting a corrected output response, wherein the corrected output response meets a set of guidelines of the behavior framework;

monitoring a second interaction session between a second client interacting with the second virtual assistant to detect a deviation from the second behavior framework;

generating, upon detection of the deviation from the second behavior framework, a second correction action to apply to a portion of the second interaction session corresponding to the deviation from the second behavior framework;

applying the second correction action to the portion of the second interaction session corresponding to the deviation from the second behavior framework; and

outputting a second corrected output response, wherein the second corrected output response meets a set of guidelines of the second behavior framework.

14 . The computer system of claim 13 , wherein the second behavior framework comprises a different set of guidelines than the behavior framework.

15 . The computer system of claim 13 , wherein the correction action comprises modifying the portion of the interaction session corresponding to the deviation from the behavior framework.

16 . The computer system of claim 13 , wherein the correction action comprises providing a suggested action to mitigate a potential bias outcome.

17 . The computer system of claim 13 , wherein the correction action comprises providing additional context surrounding the portion of the interaction session corresponding to the deviation from the behavior framework.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 30, 2024
From: FOX, JEREMY R.; KOCH, FERNANDO LUIZ; KEEN, MARTIN G.; NAHULAN, JESSICA
To: INTERNATIONAL BUSINESS MACHINES CORPORATION
Reel/Frame 069077/0939 →
Continuity (1)
Related Publication 20260119508A1 · Apr 30, 2026
References Cited (12)
US 11806629B2 · Donovan · 2023 [cited by applicant]
US 20210299576A1 · Donovan · 2021 [cited by applicant]
US 20240071596A1 · Bradley · 2024 [cited by examiner]
US 20240273431A1 · Maikhuri · 2024 [cited by examiner]
CN 111860133A · 2020 [cited by applicant]
IN 202411010843A · 2024 [cited by applicant]
IBM, “AI ethics”, retrieved from web https://www.ibm.com/impact/ai-ethics dated Dec. 12, 2024, 15 pages. [cited by applicant]
IBM, “What is explainable AI?”, retrieved from web https://www.ibm.com/topics/explainable-ai dated Dec. 12, 2024, 11 pages. [cited by applicant]
Hjelmtveit, Kevin, “Moderating Respectful Conversations Using Reinforcement Q-Learning and Markov Decision Process Approaches,” Considering Toxicity in Speech, Dec. 2023, 98 pages. [cited by applicant]
Da Silva et al., “Could a Conversational AI Identify Offensive Language?,” Information, Oct. 12, 2021, 28 pages, vol. 12, No. 10. [cited by applicant]
Vakkuri et al., “ECCOLA—A method for implementing ethically aligned AI systems”, The Journal of Systems and Software, Dec. 2021, 16 pages. [cited by applicant]
Yuan et al., “R-Judge: Benchmarking Safety Risk Awareness for LLM Agents,” arXiv:2401.10019v2 [cs.CL], Feb. 18, 2024, 21 pages. [cited by applicant]